API documentation

Meeting minutes

meeting_minutes

Input and limits

Plain UTF-8 text only, up to 12,000 characters; each non-empty source line is limited to 1,200 characters. options.items accepts the strings "5", "10" or "20" (default "10"). Source language is preserved. Uploaded PDFs, remote URLs and images are not accepted.

Result and export

The result contains reviewed draft candidates, exact source quotations, line numbers and byte offsets. Missing owners, deadlines or answers remain missing. result.exports.text and result.exports.csv contain filename, mime and UTF-8 content. CSV cells receive formula protection. Learning quizzes also provide result.exports.moodle_xml for Moodle short-answer import. Questions and answers still need subject review.

JSON request

{
    "tool": "meeting_minutes",
    "text": "Decision: use the blue design.\nAnna will send the draft by 7 November 2026.\nBudget remains open.",
    "options": {
        "items": "10"
    }
}

Tender requirements

tender_requirements

Input and limits

Plain UTF-8 text only, up to 12,000 characters; each non-empty source line is limited to 1,200 characters. options.items accepts the strings "5", "10" or "20" (default "10"). Source language is preserved. Uploaded PDFs, remote URLs and images are not accepted.

Result and export

The result contains reviewed draft candidates, exact source quotations, line numbers and byte offsets. Missing owners, deadlines or answers remain missing. result.exports.text and result.exports.csv contain filename, mime and UTF-8 content. CSV cells receive formula protection. Learning quizzes also provide result.exports.moodle_xml for Moodle short-answer import. Questions and answers still need subject review.

JSON request

{
    "tool": "tender_requirements",
    "text": "A signed declaration is mandatory.\nThe proposal must include two references.\nThe submission deadline is 9 November 2026.",
    "options": {
        "items": "10"
    }
}

Work instructions

work_instructions

Input and limits

Plain UTF-8 text only, up to 12,000 characters; each non-empty source line is limited to 1,200 characters. options.items accepts the strings "5", "10" or "20" (default "10"). Source language is preserved. Uploaded PDFs, remote URLs and images are not accepted.

Result and export

The result contains reviewed draft candidates, exact source quotations, line numbers and byte offsets. Missing owners, deadlines or answers remain missing. result.exports.text and result.exports.csv contain filename, mime and UTF-8 content. CSV cells receive formula protection. Learning quizzes also provide result.exports.moodle_xml for Moodle short-answer import. Questions and answers still need subject review.

JSON request

{
    "tool": "work_instructions",
    "text": "Prerequisite: a clean surface.\nOperator: technician.\nSwitch off the device before cleaning.\nCheck that the surface is dry.",
    "options": {
        "items": "10"
    }
}

Learning quiz

learning_quiz

Input and limits

Plain UTF-8 text only, up to 12,000 characters; each non-empty source line is limited to 1,200 characters. options.items accepts the strings "5", "10" or "20" (default "10"). Source language is preserved. Uploaded PDFs, remote URLs and images are not accepted.

Result and export

The result contains reviewed draft candidates, exact source quotations, line numbers and byte offsets. Missing owners, deadlines or answers remain missing. result.exports.text and result.exports.csv contain filename, mime and UTF-8 content. CSV cells receive formula protection. Learning quizzes also provide result.exports.moodle_xml for Moodle short-answer import. Questions and answers still need subject review.

JSON request

{
    "tool": "learning_quiz",
    "text": "Water freezes at zero degrees Celsius.\nA triangle has three sides.\nPhotosynthesis uses light energy.",
    "options": {
        "items": "10"
    }
}

CSV and XLSX cleaner

clean_table

Input and limits

Send CSV in text (up to 24,000 bytes), or text beginning D4XLSX1: followed by strict base64 of an XLSX file (up to 1,000,000 decoded bytes). Limits: one sheet, 201 rows including the header, 32 columns, 3,000 cells and 20,000 cell-text bytes. Formulas, macros and external links are rejected. Styles and spreadsheet features are not preserved. options.delimiter: auto, comma, semicolon or tab. options.decimal: keep or comma.

Result and export

The review identifies exact duplicate rows and conservative value corrections. To export, submit the same original text and options again, with options.office_action="export", options.export_format="csv" or "xlsx", and selected as a JSON-encoded string list of reviewed IDs. Only selected changes apply. An empty selection applies no cleanup. CSV formula protection remains active; XLSX stores literal strings. The worker recomputes IDs rather than trusting a client report.

selected is a string containing a JSON list, for example "[\"C1\"]"; use actual IDs from the review. Maximum 5,000 bytes and 3,000 distinct IDs. Export jobs return result.type="office_export" and result.export containing filename, mime and content_base64. Decode base64 as file bytes. No client-supplied report is accepted.

JSON request

{
    "tool": "clean_table",
    "text": "Name,Date\nAlpha,2026-11-02\nAlpha,2026-11-02",
    "options": {
        "delimiter": "comma",
        "decimal": "keep"
    }
}

Text anonymiser

anonymize

Input and limits

Plain UTF-8 text up to 12,000 bytes. options.scope accepts all, email, phone, iban or ip. Optional terms is a string of up to 30 exact, case-sensitive terms on separate lines, at most 2,000 bytes. Name and address detection requires these custom terms. IP recognition covers IPv4; phone candidates need review and IBAN candidates require a valid checksum.

Result and export

Review each occurrence and its proposed placeholder. Repeated identical values use stable placeholders within that text. For export, resubmit the original text and options with options.office_action="export", options.export_format="txt", and selected as a JSON-encoded string list of occurrence IDs. Unselected occurrences remain unchanged. This assists redaction and does not certify irreversible anonymisation.

selected is a string containing a JSON list, for example "[\"C1\"]"; use actual IDs from the review. Maximum 5,000 bytes and 3,000 distinct IDs. Export jobs return result.type="office_export" and result.export containing filename, mime and content_base64. Decode base64 as file bytes. No client-supplied report is accepted.

JSON request

{
    "tool": "anonymize",
    "text": "Contact: alex@example.org; reference AB-104.",
    "terms": "AB-104",
    "options": {
        "scope": "all"
    }
}

Subtitle check and translation

subtitles

Input and limits

Send plain SRT or WebVTT in text, up to 65,536 bytes, 200 cues and 16,000 cue-text characters. Styling, embedded markup and unsupported metadata are rejected. options.mode is check (default) or translate; options.format is auto, srt or vtt. options.target selects a supported language for translation. Integer fields max_line_chars (10–100, default 42), max_lines (1–6, default 2) and max_cps (5–40, default 20) are top-level fields.

Result and export

The result reports timing errors, overlaps, line lengths and reading speed. Checking keeps the original language. Translation changes cue text while retaining source timings and cue correspondence; it does not retime or split cues. Downloadable text is in result.file with result.filename and result.mime. Invalid timing may leave file null. A successful check is not a guarantee of accessibility or linguistic quality.

JSON request

{
    "tool": "subtitles",
    "text": "1\n00:00:01,000 --> 00:00:04,000\nThe workshop opens at nine.\n",
    "max_line_chars": 42,
    "options": {
        "mode": "check",
        "format": "srt"
    }
}

E-invoice validation

invoice_validate

Input and limits

Send UTF-8 invoice XML in text, up to 262,144 bytes. Accepted inputs are UBL Invoice, UBL CreditNote and UN/CEFACT CII with an EN 16931 or XRechnung 3 profile. options.profile is auto, en16931 or xrechnung. ZIP, PDF, embedded ZUGFeRD PDF extraction, DTDs and external entities are not supported. Extracted XML may be submitted independently if it meets the accepted profile.

Result and export

Validation uses official, pinned XSD and Schematron artefacts: KoSIT 1.6.3, XRechnung 3.0.2 configuration dated 2026-08-31, EN 16931 rules 1.3.16 and XRechnung rules 2.6.0. The result records the versions, document hash, findings, severity, rule IDs and acceptance status. not_checked means no conformance decision was obtained. A technical pass does not guarantee recipient acceptance, accounting correctness or legal compliance.

JSON request

The compact XML example below is deliberately incomplete and should return validation errors. A full business invoice must include all mandatory fields; the example demonstrates the request format, not a conformant invoice.

{
    "tool": "invoice_validate",
    "text": "<Invoice xmlns=\"urn:oasis:names:specification:ubl:schema:xsd:Invoice-2\" xmlns:cbc=\"urn:oasis:names:specification:ubl:schema:xsd:CommonBasicComponents-2\"><cbc:CustomizationID>urn:cen.eu:en16931:2017</cbc:CustomizationID></Invoice>",
    "options": {
        "profile": "auto"
    }
}

Blog writing assistant

blog_assistant

Input and limits

The briefing phase requires topic (300 bytes), audience (500 bytes) and goal (500 bytes). Optional text supplies up to 12,000 bytes of source material, in no more than 12 paragraphs of 1,800 bytes each. An optional source_url must be a public HTTP(S) address of at most 2,000 characters; analyse_url must be the JSON boolean true to authorise analysis. Private targets and disallowed pages remain blocked.

Result and export

First submit phase="briefing". Review the returned outline, questions and sources. The owner-bound result supplies approval_token and briefing_hash. Only after explicit review, submit phase="article", approved=true, sources_approved=true and both unchanged values. The approval expires after 30 minutes and belongs to the authenticated account. Article requests cannot replace topic, audience, goal, text or sources. The article remains a draft with source-linked paragraphs; nothing is published automatically.

JSON request

{
    "tool": "blog_assistant",
    "phase": "briefing",
    "topic": "Bicycle inspections",
    "audience": "Local commuters",
    "goal": "Explain the appointment process",
    "text": "An inspection requires an appointment.\n\nThe workshop checks brakes, tyres and lights."
}

Article request after approval

{
    "tool": "blog_assistant",
    "phase": "article",
    "approved": true,
    "sources_approved": true,
    "approval_token": "TOKEN_FROM_REVIEWED_BRIEFING",
    "briefing_hash": "HASH_FROM_REVIEWED_BRIEFING"
}

Receipt digitiser

receipt_extract

Input and limits

The envelope contains files: one object with name and content_base64. PDF, PNG, JPEG or TXT; maximum 1 MB, PDFs up to three pages. options.action is review or export; options.export_format is csv or xlsx. For export, resubmit the original files plus approved=true, source_sha256 and reviewed with corrected fields and items.

Result and review

The review returns extracted fields, line items, exact source evidence and arithmetic warnings. Exported values are customer-reviewed drafts, not verified accounting records. OCR can introduce errors.

Request example

{
    "tool": "receipt_extract",
    "text": "{\"files\":[{\"name\":\"receipt.txt\",\"content_base64\":\"U2FtcGxlIFN0b3JlCjIwMjYtMTEtMDIKTmV0OiAxMDAuMDAKVGF4OiAxOS4wMApUb3RhbDogMTE5LjAwIEVVUg==\"}]}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "files"
    ],
    "properties": {
        "files": {
            "type": "array",
            "minItems": 1,
            "maxItems": 1,
            "items": {
                "type": "object",
                "additionalProperties": false,
                "required": [
                    "name",
                    "content_base64"
                ],
                "properties": {
                    "name": {
                        "type": "string",
                        "maxLength": 160
                    },
                    "content_base64": {
                        "type": "string",
                        "contentEncoding": "base64",
                        "maxLength": 1400000
                    }
                }
            }
        },
        "approved": {
            "type": "boolean"
        },
        "source_sha256": {
            "type": "string",
            "pattern": "^[a-f0-9]{64}$"
        },
        "reviewed": {
            "type": "object"
        }
    }
}

Spreadsheet formula assistant

spreadsheet_formula

Input and limits

The envelope contains task, columns, application, formula_language and separator. Each column has name, cell and sample. Up to 20 columns and 3,000 task characters. application is excel or sheets; function names are en, de or pl; separator is a comma or semicolon. English function names are available for compatible spreadsheet settings.

Result and review

Returns formula, explanation, assumptions and example checks. tests_executed is false: the server never runs the formula. The supported subset allows common functions and local cell references; external references, macros and network functions are rejected.

Request example

{
    "tool": "spreadsheet_formula",
    "text": "{\"task\":\"Calculate quantity times unit price.\",\"columns\":[{\"name\":\"Quantity\",\"cell\":\"A2\",\"sample\":\"3\"},{\"name\":\"Unit price\",\"cell\":\"B2\",\"sample\":\"12\"}],\"application\":\"excel\",\"formula_language\":\"en\",\"separator\":\",\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "task",
        "columns",
        "application",
        "formula_language",
        "separator"
    ],
    "properties": {
        "task": {
            "type": "string",
            "maxLength": 3000
        },
        "columns": {
            "type": "array",
            "minItems": 1,
            "maxItems": 20,
            "items": {
                "type": "object",
                "additionalProperties": false,
                "required": [
                    "name",
                    "cell",
                    "sample"
                ],
                "properties": {
                    "name": {
                        "type": "string",
                        "maxLength": 100
                    },
                    "cell": {
                        "type": "string",
                        "pattern": "^[A-Z]{1,2}[1-9][0-9]{0,5}$"
                    },
                    "sample": {
                        "type": "string",
                        "maxLength": 300
                    }
                }
            }
        },
        "application": {
            "enum": [
                "excel",
                "sheets"
            ]
        },
        "formula_language": {
            "enum": [
                "en",
                "de",
                "pl"
            ]
        },
        "separator": {
            "enum": [
                ",",
                ";"
            ]
        }
    }
}

Translation quality check

translation_check

Input and limits

The envelope contains source, translation, source_language and target_language, plus optional terms with source and target pairs. Up to 10,000 characters per text and 30 term pairs.

Result and review

Returns possible differences with exact quotations and byte offsets in both texts. Separate checks compare literal numeric occurrences and required terms. Localised formatting or inflection may explain a finding; an empty report does not certify accuracy.

Request example

{
    "tool": "translation_check",
    "text": "{\"source\":\"The battery lasts 12 hours.\",\"translation\":\"Der Akku hält 10 Stunden.\",\"source_language\":\"en\",\"target_language\":\"de\",\"terms\":[]}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "source",
        "translation",
        "source_language",
        "target_language"
    ],
    "properties": {
        "source": {
            "type": "string",
            "maxLength": 10000
        },
        "translation": {
            "type": "string",
            "maxLength": 10000
        },
        "source_language": {
            "type": "string",
            "pattern": "^[a-z]{2}(-[A-Z]{2})?$"
        },
        "target_language": {
            "type": "string",
            "pattern": "^[a-z]{2}(-[A-Z]{2})?$"
        },
        "terms": {
            "type": "array",
            "maxItems": 30,
            "items": {
                "type": "object",
                "additionalProperties": false,
                "required": [
                    "source",
                    "target"
                ],
                "properties": {
                    "source": {
                        "type": "string",
                        "maxLength": 100
                    },
                    "target": {
                        "type": "string",
                        "maxLength": 100
                    }
                }
            }
        }
    }
}

Audio and video transcription

transcribe

Input and limits

The envelope contains filename, content_base64 and language. Supported formats: WAV, MP3, M4A, MP4, WebM, OGG and FLAC. Maximum 3 MiB and 30 seconds. language is auto or a supported language code. Files are processed locally; no URL download is performed.

Result and review

Returns timestamped segments and TXT, SRT and VTT exports for review. Recognition can be wrong. A not_completed status includes a bounded error code and does not represent a successful transcript.

A real file is required. The schema below describes the decoded text envelope; content_base64 must contain the actual file bytes encoded as standard base64. Placeholder strings are not working examples.

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "filename",
        "content_base64",
        "language"
    ],
    "properties": {
        "filename": {
            "type": "string",
            "maxLength": 150
        },
        "content_base64": {
            "type": "string",
            "contentEncoding": "base64",
            "maxLength": 4194304
        },
        "language": {
            "enum": [
                "auto",
                "de",
                "en",
                "pl",
                "fr",
                "es",
                "it",
                "nl",
                "pt",
                "cs",
                "sk",
                "da",
                "sv",
                "fi",
                "et",
                "lv",
                "lt",
                "hu",
                "ro",
                "bg",
                "el",
                "hr",
                "sl",
                "uk",
                "ru",
                "tr",
                "nb"
            ]
        }
    }
}

Document template merge

document_merge

Input and limits

The envelope contains template with name and content_base64, plus data as CSV text or D4XLSX1: followed by standard base64 XLSX bytes. A DOCX template accepts up to 32 placeholders and the data up to 20 rows. options.action is review or export; options.export_format is docx or pdf. Export requires the original inputs, approved=true and source_sha256.

Result and review

The preview lists placeholders, data columns and missing values. Missing used values block export. A ZIP contains completed documents. DOCX preserves the template layout; PDF uses a simple text layout without Word styling or images.

A real file is required. The schema below describes the decoded text envelope; content_base64 must contain the actual file bytes encoded as standard base64. Placeholder strings are not working examples.

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "template",
        "data"
    ],
    "properties": {
        "template": {
            "type": "object",
            "additionalProperties": false,
            "required": [
                "name",
                "content_base64"
            ],
            "properties": {
                "name": {
                    "type": "string",
                    "maxLength": 160
                },
                "content_base64": {
                    "type": "string",
                    "contentEncoding": "base64",
                    "maxLength": 1400000
                }
            }
        },
        "data": {
            "type": "string",
            "maxLength": 1400000
        },
        "approved": {
            "type": "boolean"
        },
        "source_sha256": {
            "type": "string",
            "pattern": "^[a-f0-9]{64}$"
        }
    }
}

Import mapping assistant

import_map

Input and limits

The envelope contains source and target tables as CSV text or D4XLSX1: followed by standard base64 XLSX bytes. The target header names required columns; an optional second row declares text, number, date or email, optionally prefixed by required:. mapping contains zero-based source-column indexes or null. Up to 200 data rows and 32 columns.

Result and review

Returns suggested mappings, a reusable profile, preview rows and validation issues. Export with options.action=export requires approved=true and an explicit mapping; unresolved validation issues block export. No external system is modified.

Request example

{
    "tool": "import_map",
    "text": "{\"source\":\"name,amount\\nExample,12\",\"target\":\"name,amount\\nrequired:text,number\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "source",
        "target"
    ],
    "properties": {
        "source": {
            "type": "string",
            "maxLength": 1400000
        },
        "target": {
            "type": "string",
            "maxLength": 1400000
        },
        "approved": {
            "type": "boolean"
        },
        "mapping": {
            "type": "array",
            "maxItems": 32,
            "items": {
                "type": [
                    "integer",
                    "null"
                ],
                "minimum": 0,
                "maximum": 31
            }
        }
    }
}

Presentations from source material

presentation

Input and limits

For phase=outline, text contains title, audience, purpose and material. Material has one to eight source paragraphs, up to 1,200 characters each. For phase=pptx, send approved=true and approval_token at the top level; text must be absent or empty. Tokens are bound to the authenticated account and expire.

Result and review

The outline contains source-bound slide titles and quotations. After review, the second request exports an editable PPTX with geometric accents and speaker notes. Source text cannot modify approval, account binding or export permissions.

Request example

{
    "tool": "presentation",
    "phase": "outline",
    "text": "{\"title\":\"Workshop services\",\"audience\":\"Local businesses\",\"purpose\":\"Explain available services\",\"material\":\"The workshop repairs bicycles.\\nAppointments are required for inspections.\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "title",
        "audience",
        "purpose",
        "material"
    ],
    "properties": {
        "title": {
            "type": "string",
            "maxLength": 160
        },
        "audience": {
            "type": "string",
            "maxLength": 300
        },
        "purpose": {
            "type": "string",
            "maxLength": 500
        },
        "material": {
            "type": "string",
            "maxLength": 12000
        }
    }
}

Export an approved presentation

Replace APPROVAL_TOKEN_FROM_OWN_OUTLINE with the token from the reviewed outline result; it is not a working token.

{
    "tool": "presentation",
    "phase": "pptx",
    "approved": true,
    "approval_token": "APPROVAL_TOKEN_FROM_OWN_OUTLINE"
}

Document filing assistant

document_filing

Input and limits

The envelope contains files and an optional project name. Up to ten TXT, PDF, DOCX, PNG or JPEG files, 1 MB per file and 2 MB combined. options.action is review or export; options.export_format is zip. Export requires original inputs, approved=true, source_sha256 and a names object mapping file IDs to reviewed destinations.

Result and review

Returns proposed categories, dates and filenames. Suggestions are based on document text and require review. Export creates a ZIP of copies; originals remain unchanged. Unsafe paths, extension changes and duplicate destinations are rejected.

Request example

{
    "tool": "document_filing",
    "text": "{\"files\":[{\"name\":\"invoice.txt\",\"content_base64\":\"SW52b2ljZQoyMDI2LTExLTAyClRvdGFsIDEyMCBFVVI=\"}],\"project\":\"Project\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "files"
    ],
    "properties": {
        "files": {
            "type": "array",
            "minItems": 1,
            "maxItems": 10,
            "items": {
                "type": "object",
                "additionalProperties": false,
                "required": [
                    "name",
                    "content_base64"
                ],
                "properties": {
                    "name": {
                        "type": "string",
                        "maxLength": 160
                    },
                    "content_base64": {
                        "type": "string",
                        "contentEncoding": "base64",
                        "maxLength": 1400000
                    }
                }
            }
        },
        "project": {
            "type": "string",
            "maxLength": 80
        },
        "approved": {
            "type": "boolean"
        },
        "source_sha256": {
            "type": "string",
            "pattern": "^[a-f0-9]{64}$"
        },
        "names": {
            "type": "object",
            "additionalProperties": {
                "type": "string"
            }
        }
    }
}

Supplier offer comparison

supplier_compare

Compare supplier offer lines against required items, review AI match proposals and inspect quantity, currency, shipping and total-price differences.

Decoded input fields

requirements — Required items

CSV columns: description,quantity,unit. Use the same unit spelling as in the offers. Up to 30 rows; 40 rows across all tables.

offers — Supplier offer lines

CSV columns: supplier,description,quantity,unit,unit_price,shipping,currency,lead_days. Shipping is the charge for this line only; enter 0 if none. Lead time may be empty. Compare prices on the same tax basis.

csv_separator — CSV separator

Choose semicolon, comma or tab for the exported CSV only. Source CSV uploads still require their documented comma-separated headers.

csv_bom — UTF-8 marker in CSV

Include a UTF-8 byte-order mark for spreadsheet compatibility, or omit it for systems that require plain UTF-8.

csv_line_ending — CSV line endings

Use LF or CRLF for exported CSV records. This does not change source text or line breaks inside quoted cells.

include_json — Include JSON export

Offer an additional structured JSON report containing the complete reviewed rows, findings and summary. No action is executed by exporting it.

max_lead_days — Lead-time review limit

0 disables this additional check. Otherwise, flag offer lead times above the selected number of days and offers with no stated lead time. Prices and comparisons remain unchanged. Maximum 999 days.

row_order — Review row order

Reorder complete table rows without removing findings, changing calculations or modifying assignments. Empty sort values appear last; equal values retain their original relative order.

Maximum UTF-8 bytes in the text envelope: 60000

Result and source review

The result type is ops_review. columns contains key and label_key; rows contains the corresponding scalar cell values. findings contains localisable code values and source refs; summary contains label_key and value. review_required and ai_required are true. ai_model identifies the accepted model when supplied by the worker. Some tools also return source_sha256 or note_key. Source strings remain untrusted data and must be rendered as text, never executable HTML or instructions.

Request example

{
    "tool": "supplier_compare",
    "text": "{\"requirements\":\"description,quantity,unit\\nBlue A4 folders,10,piece\",\"offers\":\"supplier,description,quantity,unit,unit_price,shipping,currency,lead_days\\nNorth,Blue A4 document folders,10,piece,2.50,5.00,EUR,3\\nSouth,A4 folder blue,10,piece,2.30,8.00,EUR,5\",\"csv_separator\":\"semicolon\",\"csv_bom\":true,\"csv_line_ending\":\"lf\",\"include_json\":false,\"max_lead_days\":0,\"row_order\":\"source\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "requirements",
        "offers"
    ],
    "properties": {
        "requirements": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 20000
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 30,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "required": [
                            "description",
                            "quantity",
                            "unit"
                        ],
                        "properties": {
                            "description": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "quantity": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "unit": {
                                "type": "string",
                                "maxLength": 240
                            }
                        }
                    }
                }
            ]
        },
        "offers": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 20000
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 30,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "required": [
                            "supplier",
                            "description",
                            "quantity",
                            "unit",
                            "unit_price",
                            "shipping",
                            "currency",
                            "lead_days"
                        ],
                        "properties": {
                            "supplier": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "description": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "quantity": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "unit": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "unit_price": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "shipping": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "currency": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "lead_days": {
                                "type": "string",
                                "maxLength": 240
                            }
                        }
                    }
                }
            ]
        },
        "csv_separator": {
            "type": "string",
            "enum": [
                "semicolon",
                "comma",
                "tab"
            ],
            "default": "semicolon"
        },
        "csv_bom": {
            "type": "boolean",
            "default": true
        },
        "csv_line_ending": {
            "type": "string",
            "enum": [
                "lf",
                "crlf"
            ],
            "default": "lf"
        },
        "include_json": {
            "type": "boolean",
            "default": false
        },
        "max_lead_days": {
            "type": "integer",
            "minimum": 0,
            "maximum": 999,
            "default": 0
        },
        "row_order": {
            "type": "string",
            "enum": [
                "source",
                "supplier"
            ],
            "default": "source"
        }
    }
}

Order, delivery and invoice reconciliation

three_way_match

Match order, delivery and invoice lines with AI assistance, then calculate missing quantities, excess billing and unit-price differences from the source data.

Decoded input fields

orders — Order lines

CSV columns: reference,description,quantity,unit,unit_price,currency. Reference means the shared order reference, not a supplier document number. Use the same unit labels and tax basis throughout.

deliveries — Delivered lines

CSV columns: reference,description,quantity,unit. Reference means the shared order reference. Partial deliveries may be separate rows.

invoices — Invoice lines

CSV columns: reference,description,quantity,unit,unit_price,currency. Reference means the shared order reference. Tax, discounts and credit notes must not be entered as item lines.

csv_separator — CSV separator

Choose semicolon, comma or tab for the exported CSV only. Source CSV uploads still require their documented comma-separated headers.

csv_bom — UTF-8 marker in CSV

Include a UTF-8 byte-order mark for spreadsheet compatibility, or omit it for systems that require plain UTF-8.

csv_line_ending — CSV line endings

Use LF or CRLF for exported CSV records. This does not change source text or line breaks inside quoted cells.

include_json — Include JSON export

Offer an additional structured JSON report containing the complete reviewed rows, findings and summary. No action is executed by exporting it.

review_blank_references — Flag missing document references

Add a review finding for every order, delivery or invoice row with an empty reference. Existing matching decisions and warnings remain unchanged.

row_order — Review row order

Reorder complete table rows without removing findings, changing calculations or modifying assignments. Empty sort values appear last; equal values retain their original relative order.

Maximum UTF-8 bytes in the text envelope: 60000

Result and source review

The result type is ops_review. columns contains key and label_key; rows contains the corresponding scalar cell values. findings contains localisable code values and source refs; summary contains label_key and value. review_required and ai_required are true. ai_model identifies the accepted model when supplied by the worker. Some tools also return source_sha256 or note_key. Source strings remain untrusted data and must be rendered as text, never executable HTML or instructions.

Request example

{
    "tool": "three_way_match",
    "text": "{\"orders\":\"reference,description,quantity,unit,unit_price,currency\\nPO-1,Blue A4 folders,10,piece,2.50,EUR\",\"deliveries\":\"reference,description,quantity,unit\\nPO-1,A4 folders blue,8,piece\",\"invoices\":\"reference,description,quantity,unit,unit_price,currency\\nPO-1,Blue document folders A4,10,piece,2.60,EUR\",\"csv_separator\":\"semicolon\",\"csv_bom\":true,\"csv_line_ending\":\"lf\",\"include_json\":false,\"review_blank_references\":false,\"row_order\":\"source\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "orders",
        "deliveries",
        "invoices"
    ],
    "properties": {
        "orders": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 20000
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 30,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "required": [
                            "reference",
                            "description",
                            "quantity",
                            "unit",
                            "unit_price",
                            "currency"
                        ],
                        "properties": {
                            "reference": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "description": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "quantity": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "unit": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "unit_price": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "currency": {
                                "type": "string",
                                "maxLength": 240
                            }
                        }
                    }
                }
            ]
        },
        "deliveries": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 20000
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 30,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "required": [
                            "reference",
                            "description",
                            "quantity",
                            "unit"
                        ],
                        "properties": {
                            "reference": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "description": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "quantity": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "unit": {
                                "type": "string",
                                "maxLength": 240
                            }
                        }
                    }
                }
            ]
        },
        "invoices": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 20000
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 30,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "required": [
                            "reference",
                            "description",
                            "quantity",
                            "unit",
                            "unit_price",
                            "currency"
                        ],
                        "properties": {
                            "reference": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "description": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "quantity": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "unit": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "unit_price": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "currency": {
                                "type": "string",
                                "maxLength": 240
                            }
                        }
                    }
                }
            ]
        },
        "csv_separator": {
            "type": "string",
            "enum": [
                "semicolon",
                "comma",
                "tab"
            ],
            "default": "semicolon"
        },
        "csv_bom": {
            "type": "boolean",
            "default": true
        },
        "csv_line_ending": {
            "type": "string",
            "enum": [
                "lf",
                "crlf"
            ],
            "default": "lf"
        },
        "include_json": {
            "type": "boolean",
            "default": false
        },
        "review_blank_references": {
            "type": "boolean",
            "default": false
        },
        "row_order": {
            "type": "string",
            "enum": [
                "source",
                "reference"
            ],
            "default": "source"
        }
    }
}

Payment and invoice reconciliation

payment_match

Propose matches between incoming payments and open invoices, identify ambiguity and calculate instalments, remaining balances and overpayments.

Decoded input fields

payments — Incoming payments

CSV columns: reference,counterparty,date,amount,currency. Use invoice references where known, positive amounts and YYYY-MM-DD dates. One payment can be proposed for one invoice only.

invoices — Open invoices

CSV columns: reference,counterparty,date,amount,currency. Amount is the outstanding balance before the uploaded payments. References may be empty, but ambiguity can prevent allocation.

csv_separator — CSV separator

Choose semicolon, comma or tab for the exported CSV only. Source CSV uploads still require their documented comma-separated headers.

csv_bom — UTF-8 marker in CSV

Include a UTF-8 byte-order mark for spreadsheet compatibility, or omit it for systems that require plain UTF-8.

csv_line_ending — CSV line endings

Use LF or CRLF for exported CSV records. This does not change source text or line breaks inside quoted cells.

include_json — Include JSON export

Offer an additional structured JSON report containing the complete reviewed rows, findings and summary. No action is executed by exporting it.

review_date — Invoice-age reference date

Optional explicit YYYY-MM-DD date, from 1900 through 2200. Open invoice balances are checked by age from the source invoice date. Empty disables this check. Invoice age is not a payment due date or a legal overdue finding.

aging_days — Invoice-age review threshold

Flag positive open balances whose invoice date is at least this many days before the selected reference date. Range 0–3,650 days. Requires a reference date; no invoice deadline is inferred.

row_order — Review row order

Reorder complete table rows without removing findings, changing calculations or modifying assignments. Empty sort values appear last; equal values retain their original relative order.

Maximum UTF-8 bytes in the text envelope: 60000

Result and source review

The result type is ops_review. columns contains key and label_key; rows contains the corresponding scalar cell values. findings contains localisable code values and source refs; summary contains label_key and value. review_required and ai_required are true. ai_model identifies the accepted model when supplied by the worker. Some tools also return source_sha256 or note_key. Source strings remain untrusted data and must be rendered as text, never executable HTML or instructions.

Request example

{
    "tool": "payment_match",
    "text": "{\"payments\":\"reference,counterparty,date,amount,currency\\nINV-1,North Studio,2026-11-02,60.00,EUR\\nINV-1,North Studio,2026-11-03,40.00,EUR\",\"invoices\":\"reference,counterparty,date,amount,currency\\nINV-1,North Studio,2026-11-01,100.00,EUR\",\"csv_separator\":\"semicolon\",\"csv_bom\":true,\"csv_line_ending\":\"lf\",\"include_json\":false,\"review_date\":\"\",\"aging_days\":30,\"row_order\":\"source\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "payments",
        "invoices"
    ],
    "properties": {
        "payments": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 20000
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 30,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "required": [
                            "reference",
                            "counterparty",
                            "date",
                            "amount",
                            "currency"
                        ],
                        "properties": {
                            "reference": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "counterparty": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "date": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "amount": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "currency": {
                                "type": "string",
                                "maxLength": 240
                            }
                        }
                    }
                }
            ]
        },
        "invoices": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 20000
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 30,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "required": [
                            "reference",
                            "counterparty",
                            "date",
                            "amount",
                            "currency"
                        ],
                        "properties": {
                            "reference": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "counterparty": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "date": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "amount": {
                                "type": "string",
                                "maxLength": 240
                            },
                            "currency": {
                                "type": "string",
                                "maxLength": 240
                            }
                        }
                    }
                }
            ]
        },
        "csv_separator": {
            "type": "string",
            "enum": [
                "semicolon",
                "comma",
                "tab"
            ],
            "default": "semicolon"
        },
        "csv_bom": {
            "type": "boolean",
            "default": true
        },
        "csv_line_ending": {
            "type": "string",
            "enum": [
                "lf",
                "crlf"
            ],
            "default": "lf"
        },
        "include_json": {
            "type": "boolean",
            "default": false
        },
        "review_date": {
            "type": "string",
            "maxLength": 10,
            "default": ""
        },
        "aging_days": {
            "type": "integer",
            "minimum": 0,
            "maximum": 3650,
            "default": 30
        },
        "row_order": {
            "type": "string",
            "enum": [
                "source",
                "date"
            ],
            "default": "source"
        }
    }
}

Reorder planning assistant

reorder_plan

Prioritise stock items with AI, calculate replenishment quantities from supplied demand and lead times, and review minimum-order and pack-size effects.

Decoded input fields

items — Stock items

CSV columns: id,name,on_hand,daily_demand,lead_days,safety_stock,min_order,pack_size. IDs must be unique. Quantities use decimal points; lead_days and min_order are nonnegative whole numbers, and pack_size is a positive whole number.

horizon_days — Additional coverage days

Target coverage equals lead time plus these additional days. Demand is assumed constant; the result is a planning estimate.

notes — Planning notes and preferences

Optional context for the AI proposal. Availability, qualifications, dependencies and numerical limits must also be entered in the structured tables; notes never override those checks.

csv_separator — CSV separator

Choose semicolon, comma or tab for the exported CSV only. Source CSV uploads still require their documented comma-separated headers.

csv_bom — UTF-8 marker in CSV

Include a UTF-8 byte-order mark for spreadsheet compatibility, or omit it for systems that require plain UTF-8.

csv_line_ending — CSV line endings

Use LF or CRLF for exported CSV records. This does not change source text or line breaks inside quoted cells.

include_json — Include JSON export

Offer an additional structured JSON report containing the complete reviewed rows, findings and summary. No action is executed by exporting it.

large_order_threshold — Large-order review threshold

0 disables this additional check. Otherwise, flag suggested order quantities strictly above the threshold, up to 10,000,000 units. Source stock, demand, target coverage, pack rounding and order quantities are not changed.

row_order — Review row order

Reorder complete table rows without removing findings, changing calculations or modifying assignments. Empty sort values appear last; equal values retain their original relative order.

Maximum UTF-8 bytes in the text envelope: 120000

Result and source review

The result type is ops_review. columns contains key and label_key; rows contains the corresponding scalar cell values. findings contains localisable code values and source refs; summary contains label_key and value. review_required and ai_required are true. ai_model identifies the accepted model when supplied by the worker. Some tools also return source_sha256 or note_key. Source strings remain untrusted data and must be rendered as text, never executable HTML or instructions.

Request example

{
    "tool": "reorder_plan",
    "text": "{\"items\":\"id,name,on_hand,daily_demand,lead_days,safety_stock,min_order,pack_size\\nI1,Brake pads,12,3,5,4,10,5\\nI2,Chain oil,30,1,3,2,6,6\",\"horizon_days\":7,\"notes\":\"Brake pads support scheduled repairs and should be reviewed first. Chain oil is less urgent.\",\"csv_separator\":\"semicolon\",\"csv_bom\":true,\"csv_line_ending\":\"lf\",\"include_json\":false,\"large_order_threshold\":0,\"row_order\":\"source\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "items",
        "horizon_days"
    ],
    "properties": {
        "items": {
            "type": "string",
            "maxLength": 40000
        },
        "horizon_days": {
            "type": "integer",
            "minimum": 0,
            "maximum": 365
        },
        "notes": {
            "type": "string",
            "maxLength": 3000
        },
        "csv_separator": {
            "type": "string",
            "enum": [
                "semicolon",
                "comma",
                "tab"
            ],
            "default": "semicolon"
        },
        "csv_bom": {
            "type": "boolean",
            "default": true
        },
        "csv_line_ending": {
            "type": "string",
            "enum": [
                "lf",
                "crlf"
            ],
            "default": "lf"
        },
        "include_json": {
            "type": "boolean",
            "default": false
        },
        "large_order_threshold": {
            "type": "integer",
            "minimum": 0,
            "maximum": 10000000,
            "default": 0
        },
        "row_order": {
            "type": "string",
            "enum": [
                "source",
                "order_desc"
            ],
            "default": "source"
        }
    }
}

Shift planning assistant

shift_plan

Propose worker assignments with AI and check availability, required skills, overlapping shifts, supplied hour limits and remaining coverage.

Decoded input fields

workers — Workers

CSV columns: id,name,skills,max_hours. Multiple skills use |. max_hours is the maximum total planned work across this proposal, not a legal working-time limit.

availability — Availability

CSV columns: id,worker_id,start,end. Times are explicitly UTC in YYYY-MM-DDTHH:MM format. One availability interval must cover the complete shift.

shifts — Shifts

CSV columns: id,start,end,skill,needed. All times are UTC. A worker must have the exact required skill. needed is the required number of workers.

notes — Planning notes and preferences

Optional context for the AI proposal. Availability, qualifications, dependencies and numerical limits must also be entered in the structured tables; notes never override those checks.

csv_separator — CSV separator

Choose semicolon, comma or tab for the exported CSV only. Source CSV uploads still require their documented comma-separated headers.

csv_bom — UTF-8 marker in CSV

Include a UTF-8 byte-order mark for spreadsheet compatibility, or omit it for systems that require plain UTF-8.

csv_line_ending — CSV line endings

Use LF or CRLF for exported CSV records. This does not change source text or line breaks inside quoted cells.

include_json — Include JSON export

Offer an additional structured JSON report containing the complete reviewed rows, findings and summary. No action is executed by exporting it.

max_shift_hours — Shift-duration review threshold

0 disables this additional check. Otherwise, flag source shifts longer than the selected duration, up to 168 hours. Assignments and coverage warnings are retained. This is a planning preference, not a legal working-time assessment.

row_order — Review row order

Reorder complete table rows without removing findings, changing calculations or modifying assignments. Empty sort values appear last; equal values retain their original relative order.

Maximum UTF-8 bytes in the text envelope: 120000

Result and source review

The result type is ops_review. columns contains key and label_key; rows contains the corresponding scalar cell values. findings contains localisable code values and source refs; summary contains label_key and value. review_required and ai_required are true. ai_model identifies the accepted model when supplied by the worker. Some tools also return source_sha256 or note_key. Source strings remain untrusted data and must be rendered as text, never executable HTML or instructions.

Request example

{
    "tool": "shift_plan",
    "text": "{\"workers\":\"id,name,skills,max_hours\\nW1,Alex,service|repair,8\\nW2,Sam,service,8\",\"availability\":\"id,worker_id,start,end\\nA1,W1,2026-11-09T08:00,2026-11-09T16:00\\nA2,W2,2026-11-09T08:00,2026-11-09T16:00\",\"shifts\":\"id,start,end,skill,needed\\nS1,2026-11-09T08:00,2026-11-09T12:00,service,1\\nS2,2026-11-09T12:00,2026-11-09T16:00,repair,1\",\"notes\":\"Where practical, retain Alex for repair work and use Sam for the service desk.\",\"csv_separator\":\"semicolon\",\"csv_bom\":true,\"csv_line_ending\":\"lf\",\"include_json\":false,\"max_shift_hours\":0,\"row_order\":\"source\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "workers",
        "availability",
        "shifts"
    ],
    "properties": {
        "workers": {
            "type": "string",
            "maxLength": 40000
        },
        "availability": {
            "type": "string",
            "maxLength": 40000
        },
        "shifts": {
            "type": "string",
            "maxLength": 40000
        },
        "notes": {
            "type": "string",
            "maxLength": 3000
        },
        "csv_separator": {
            "type": "string",
            "enum": [
                "semicolon",
                "comma",
                "tab"
            ],
            "default": "semicolon"
        },
        "csv_bom": {
            "type": "boolean",
            "default": true
        },
        "csv_line_ending": {
            "type": "string",
            "enum": [
                "lf",
                "crlf"
            ],
            "default": "lf"
        },
        "include_json": {
            "type": "boolean",
            "default": false
        },
        "max_shift_hours": {
            "type": "number",
            "minimum": 0,
            "maximum": 168,
            "default": 0
        },
        "row_order": {
            "type": "string",
            "enum": [
                "source",
                "begin"
            ],
            "default": "source"
        }
    }
}

Project capacity planner

project_plan

Use AI to propose task priorities and resource assignments, then calculate a dependency-aware schedule against daily capacity and flag work that remains unassigned.

Decoded input fields

tasks — Project tasks

CSV columns: id,name,hours,depends,skill. Dependencies are task IDs separated by |. Empty depends means no predecessor. Task effort is allocated to one qualified resource.

resources — Project resources

CSV columns: id,name,skills,hours_per_day. Skills use |. Daily capacity is the available planning time, up to 12 hours. Calendars use Monday to Friday; holidays and leave are not modelled.

start_date — Project start date

The first working day is this date or the next Monday if it falls on a weekend. Working-day slots start at 00:00 UTC and describe capacity allocation, not actual working hours.

notes — Planning notes and preferences

Optional context for the AI proposal. Availability, qualifications, dependencies and numerical limits must also be entered in the structured tables; notes never override those checks.

csv_separator — CSV separator

Choose semicolon, comma or tab for the exported CSV only. Source CSV uploads still require their documented comma-separated headers.

csv_bom — UTF-8 marker in CSV

Include a UTF-8 byte-order mark for spreadsheet compatibility, or omit it for systems that require plain UTF-8.

csv_line_ending — CSV line endings

Use LF or CRLF for exported CSV records. This does not change source text or line breaks inside quoted cells.

include_json — Include JSON export

Offer an additional structured JSON report containing the complete reviewed rows, findings and summary. No action is executed by exporting it.

max_task_hours — Task-effort review threshold

0 disables this additional check. Otherwise, flag tasks with effort above the selected threshold, up to 10,000 hours. Dependencies, resource capacity and the calculated schedule remain unchanged.

row_order — Review row order

Reorder complete table rows without removing findings, changing calculations or modifying assignments. Empty sort values appear last; equal values retain their original relative order.

Maximum UTF-8 bytes in the text envelope: 120000

Result and source review

The result type is ops_review. columns contains key and label_key; rows contains the corresponding scalar cell values. findings contains localisable code values and source refs; summary contains label_key and value. review_required and ai_required are true. ai_model identifies the accepted model when supplied by the worker. Some tools also return source_sha256 or note_key. Source strings remain untrusted data and must be rendered as text, never executable HTML or instructions.

Request example

{
    "tool": "project_plan",
    "text": "{\"tasks\":\"id,name,hours,depends,skill\\nT1,Briefing,4,,planning\\nT2,Design,10,T1,design\\nT3,Review,3,T2,planning\",\"resources\":\"id,name,skills,hours_per_day\\nR1,Alex,planning,4\\nR2,Sam,design,5\",\"start_date\":\"2026-11-09\",\"notes\":\"The design depends on the completed briefing. Preserve time for the final review.\",\"csv_separator\":\"semicolon\",\"csv_bom\":true,\"csv_line_ending\":\"lf\",\"include_json\":false,\"max_task_hours\":0,\"row_order\":\"source\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "tasks",
        "resources",
        "start_date"
    ],
    "properties": {
        "tasks": {
            "type": "string",
            "maxLength": 40000
        },
        "resources": {
            "type": "string",
            "maxLength": 40000
        },
        "start_date": {
            "type": "string",
            "maxLength": 10
        },
        "notes": {
            "type": "string",
            "maxLength": 3000
        },
        "csv_separator": {
            "type": "string",
            "enum": [
                "semicolon",
                "comma",
                "tab"
            ],
            "default": "semicolon"
        },
        "csv_bom": {
            "type": "boolean",
            "default": true
        },
        "csv_line_ending": {
            "type": "string",
            "enum": [
                "lf",
                "crlf"
            ],
            "default": "lf"
        },
        "include_json": {
            "type": "boolean",
            "default": false
        },
        "max_task_hours": {
            "type": "number",
            "minimum": 0,
            "maximum": 10000,
            "default": 0
        },
        "row_order": {
            "type": "string",
            "enum": [
                "source",
                "finish_slot"
            ],
            "default": "source"
        }
    }
}

Barcode and QR label sheets

label_sheet

Turn article rows into printable labels with checked barcodes and QR codes.

Decoded input fields

csv — Article rows (CSV)

Required columns: id, name, code, qr. Up to 24 rows. QR payloads come only from the supplied qr column.

format — Barcode format

EAN-13 requires 13 digits and a valid check digit. Code 128 supports 1–24 printable ASCII characters.

layout — Sheet layout

A4: 2 × 7 labels measuring 99.1 × 38.1 mm, or 3 × 8 labels measuring 63.5 × 33.9 mm. Print at 100% without scaling.

csv_separator — CSV separator

Choose semicolon, comma or tab for the exported CSV only. Source CSV uploads still require their documented comma-separated headers.

csv_bom — UTF-8 marker in CSV

Include a UTF-8 byte-order mark for spreadsheet compatibility, or omit it for systems that require plain UTF-8.

csv_line_ending — CSV line endings

Use LF or CRLF for exported CSV records. This does not change source text or line breaks inside quoted cells.

include_json — Include JSON export

Offer an additional structured JSON report containing the complete reviewed rows, findings and summary. No action is executed by exporting it.

max_label_characters — Printed label-name review threshold

0 disables this additional check. Otherwise, flag printed names longer than the chosen number of Unicode characters, up to 160. The same names, labels and sequence remain in the preview, PDF and SVG; nothing is truncated by this option.

require_https_qr — Flag QR content without a plain HTTPS address

Add a finding when QR content is not an HTTPS address with a host and without embedded credentials, whitespace or backslashes. The QR payload is never changed, opened or fetched. This syntax check does not establish that a destination is safe.

Maximum UTF-8 bytes in the text envelope: 24000

Result and source review

The result type is ops_review. columns contains key and label_key; rows contains the corresponding scalar cell values. findings contains localisable code values and source refs; summary contains label_key and value. review_required and ai_required are true. ai_model identifies the accepted model when supplied by the worker. Some tools also return source_sha256 or note_key. Source strings remain untrusted data and must be rendered as text, never executable HTML or instructions.

Request example

{
    "tool": "label_sheet",
    "text": "{\"csv\":\"id,name,code,qr\\nA1,Blue storage box,BOX-001,BOX-001\\nA2,Green storage box,BOX-002,BOX-002\",\"format\":\"code128\",\"layout\":\"auto\",\"csv_separator\":\"semicolon\",\"csv_bom\":true,\"csv_line_ending\":\"lf\",\"include_json\":false,\"max_label_characters\":0,\"require_https_qr\":false}"
}

{
    "type": "object",
    "required": [
        "csv",
        "format",
        "layout"
    ],
    "additionalProperties": false,
    "properties": {
        "csv": {
            "type": "string",
            "maxLength": 18000
        },
        "format": {
            "enum": [
                "code128",
                "ean13"
            ]
        },
        "layout": {
            "enum": [
                "auto",
                "a4_2x7",
                "a4_3x8"
            ]
        },
        "csv_separator": {
            "type": "string",
            "enum": [
                "semicolon",
                "comma",
                "tab"
            ],
            "default": "semicolon"
        },
        "csv_bom": {
            "type": "boolean",
            "default": true
        },
        "csv_line_ending": {
            "type": "string",
            "enum": [
                "lf",
                "crlf"
            ],
            "default": "lf"
        },
        "include_json": {
            "type": "boolean",
            "default": false
        },
        "max_label_characters": {
            "type": "integer",
            "minimum": 0,
            "maximum": 160,
            "default": 0
        },
        "require_https_qr": {
            "type": "boolean",
            "default": false
        }
    }
}

Requirements test matrix

test_cases

Create source-linked positive and negative review cases from supplied requirements.

Decoded input fields

requirements — Requirements

Separate up to 12 requirements with blank lines. Each requirement may contain up to 1,000 UTF-8 bytes.

context — Review context

Optional scope information, up to 2,000 UTF-8 bytes. It is not an additional requirement or an instruction to execute tests.

csv_separator — CSV separator

Choose semicolon, comma or tab for the exported CSV only. Source CSV uploads still require their documented comma-separated headers.

csv_bom — UTF-8 marker in CSV

Include a UTF-8 byte-order mark for spreadsheet compatibility, or omit it for systems that require plain UTF-8.

csv_line_ending — CSV line endings

Use LF or CRLF for exported CSV records. This does not change source text or line breaks inside quoted cells.

include_json — Include JSON export

Offer an additional structured JSON report containing the complete reviewed rows, findings and summary. No action is executed by exporting it.

max_step_characters — Test-step length review threshold

0 disables this additional check. Otherwise, flag test cases containing an individual step longer than the chosen number of Unicode characters, up to 6,000. Full steps and expected results remain visible and exported.

row_order — Review row order

Reorder complete table rows without removing findings, changing calculations or modifying assignments. Empty sort values appear last; equal values retain their original relative order.

Maximum UTF-8 bytes in the text envelope: 24000

Result and source review

The result type is ops_review. columns contains key and label_key; rows contains the corresponding scalar cell values. findings contains localisable code values and source refs; summary contains label_key and value. review_required and ai_required are true. ai_model identifies the accepted model when supplied by the worker. Some tools also return source_sha256 or note_key. Source strings remain untrusted data and must be rendered as text, never executable HTML or instructions.

Request example

{
    "tool": "test_cases",
    "text": "{\"requirements\":\"The contact form requires an email address.\\n\\nThe order quantity must be between 1 and 10.\",\"context\":\"Review of the customer ordering form.\",\"csv_separator\":\"semicolon\",\"csv_bom\":true,\"csv_line_ending\":\"lf\",\"include_json\":false,\"max_step_characters\":0,\"row_order\":\"source\"}"
}

{
    "type": "object",
    "required": [
        "requirements"
    ],
    "additionalProperties": false,
    "properties": {
        "requirements": {
            "type": "string",
            "maxLength": 14000
        },
        "context": {
            "type": "string",
            "maxLength": 2000
        },
        "csv_separator": {
            "type": "string",
            "enum": [
                "semicolon",
                "comma",
                "tab"
            ],
            "default": "semicolon"
        },
        "csv_bom": {
            "type": "boolean",
            "default": true
        },
        "csv_line_ending": {
            "type": "string",
            "enum": [
                "lf",
                "crlf"
            ],
            "default": "lf"
        },
        "include_json": {
            "type": "boolean",
            "default": false
        },
        "max_step_characters": {
            "type": "integer",
            "minimum": 0,
            "maximum": 6000,
            "default": 0
        },
        "row_order": {
            "type": "string",
            "enum": [
                "source",
                "source_id"
            ],
            "default": "source"
        }
    }
}

Contract dates and deadlines

contract_deadlines

Review exact contractual date passages and unresolved notice periods, then confirm selected dates for calendar export.

Input and configuration fields

Essential settings

clauses — Contract clauses

One source clause per line. Up to 24 nonempty clauses, 1,200 characters per clause and 18,000 characters overall. Only dates written exactly as YYYY-MM-DD can become calendar events.

title — Contract title

Short title for calendar events; up to 120 characters.

Default settings: "Contract review"

Advanced settings

reference_date — Reference date

Explicit comparison date. Required for upcoming or past filters. This does not create or calculate contractual dates.

Default settings: ""

date_window — Date filter

Upcoming includes the reference date; past excludes it. Relative deadlines without an ISO date appear only with All dates.

all — All dates and unresolved deadlines · upcoming — On or after reference date · past — Before reference date

Default settings: "all"

event_type — Deadline category

Display one source-based category or all categories. Classification remains subject to review.

all — All categories · termination — Termination · payment — Payment · renewal — Renewal · delivery — Delivery · other — Other

Default settings: "all"

max_events — Maximum findings

Limit displayed findings to 1–20 after filtering and sorting. A notice identifies truncated results.

Default settings: 20

sort — Finding order

Sort by source clause and quoted position, or by date with unresolved deadlines last.

source — Source order · date — Date order

Default settings: "source"

reminder_days — Calendar reminder lead time

0 disables reminders. Otherwise a display reminder is set 1–90 days before each confirmed all-day event; calendar application behaviour can vary.

Default settings: 7

include_source — Include source quotations

Show exact quotations in the review and confirmed calendar descriptions. Source IDs and character positions remain visible in both modes.

yes — Include quotations · no — Source references only

Default settings: "yes"

calendar_name — Calendar name

Name for the confirmed ICS calendar, up to 120 characters.

Default settings: "Reviewed contract dates"

Maximum UTF-8 input bytes: 24000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "contract_deadlines",
    "text": "{\"clauses\":\"The invoice is payable on 2027-03-15.\\nNotice must be received at least 30 days before renewal.\\nThe delivery date is 2027-04-02.\",\"title\":\"Service agreement\",\"reference_date\":\"2027-01-01\",\"date_window\":\"upcoming\",\"event_type\":\"all\",\"max_events\":20,\"sort\":\"date\",\"reminder_days\":7,\"include_source\":\"yes\",\"calendar_name\":\"Reviewed agreement dates\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "clauses": {
            "type": "string",
            "maxLength": 18000,
            "minLength": 1
        },
        "title": {
            "type": "string",
            "maxLength": 120,
            "minLength": 1
        },
        "reference_date": {
            "type": "string",
            "maxLength": 10,
            "description": "Explicit ISO date YYYY-MM-DD, 2000 through 2100, or an empty string where optional."
        },
        "date_window": {
            "type": "string",
            "enum": [
                "all",
                "upcoming",
                "past"
            ]
        },
        "event_type": {
            "type": "string",
            "enum": [
                "all",
                "termination",
                "payment",
                "renewal",
                "delivery",
                "other"
            ]
        },
        "max_events": {
            "type": "integer",
            "minimum": 1,
            "maximum": 20
        },
        "sort": {
            "type": "string",
            "enum": [
                "source",
                "date"
            ]
        },
        "reminder_days": {
            "type": "integer",
            "minimum": 0,
            "maximum": 90
        },
        "include_source": {
            "type": "string",
            "enum": [
                "yes",
                "no"
            ]
        },
        "calendar_name": {
            "type": "string",
            "maxLength": 120,
            "minLength": 1
        }
    },
    "required": [
        "clauses"
    ]
}

Quotation calculator

quote_builder

Arrange supplied labor, service and material positions into a draft with deterministic markup, margin, discount and tax calculations.

Input and configuration fields

Essential settings

items — Quotation positions

CSV headers: label,kind,quantity,unit,cost. kind is labor, material or service. quantity uses a decimal point and up to three decimals; cost is the net unit cost with up to two decimals. Up to 30 rows. Labor quantities can represent hours.

currency — Currency

One currency for all positions; no exchange-rate conversion. All amounts use two decimal places.

EUR — EUR · USD — USD · GBP — GBP · PLN — PLN · CZK — CZK · DKK — DKK · SEK — SEK · NOK — NOK · CHF — CHF · HUF — HUF · RON — RON

Default settings: "EUR"

Advanced settings

pricing_mode — Pricing method

Markup adds a percentage to cost. Target margin divides cost by one minus the percentage; it is the pre-discount margin, not a markup.

markup — Markup on cost · margin — Target margin before discount

Default settings: "markup"

pricing_percent — Pricing percentage

0–500 for markup or 0–95 for target margin; at most two decimal places.

Default settings: 20

discount_percent — Discount percentage

Explicit discount of 0–100%, applied to each calculated selling price before tax. At most two decimal places.

Default settings: 0

tax_percent — Tax percentage

Explicit tax of 0–100% applied after discount. Zero means no added tax. This tool does not determine the applicable tax rate or tax treatment.

Default settings: 0

rounding — Rounding at each calculation step

Round line cost, selling price, discounted net amount and tax to two currency decimals using the selected rule. Totals sum the rounded lines.

nearest — Nearest cent; halves round up · up — Always round up · down — Always round down

Default settings: "nearest"

item_scope — Included position kinds

Include all source positions or only labor, material or service. Excluded positions do not contribute to any total.

all — All position kinds · labor — Labor only · material — Material only · service — Service only

Default settings: "all"

detail — Result detail

Show individual source positions or totals by position kind. Both views use the same line-level calculations.

full — Individual positions · totals — Totals by kind

Default settings: "full"

quote_reference — Quotation reference

Optional internal identifier up to 120 characters. It is displayed as supplied and creates no customer commitment.

Default settings: ""

valid_until — Draft validity date

Optional explicit ISO date displayed for review. It does not send an offer or create an agreement.

Default settings: ""

Maximum UTF-8 input bytes: 24000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "quote_builder",
    "text": "{\"items\":\"label,kind,quantity,unit,cost\\nWorkshop labor,labor,2.5,hour,40.00\\nReplacement kit,material,1,piece,30.00\",\"currency\":\"EUR\",\"pricing_mode\":\"markup\",\"pricing_percent\":20,\"discount_percent\":5,\"tax_percent\":19,\"rounding\":\"nearest\",\"item_scope\":\"all\",\"detail\":\"full\",\"quote_reference\":\"DRAFT-001\",\"valid_until\":\"2027-03-31\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "items": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 16000,
                    "minLength": 1
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 30,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "properties": {
                            "label": {
                                "type": "string",
                                "maxLength": 160,
                                "minLength": 1
                            },
                            "kind": {
                                "type": "string",
                                "enum": [
                                    "labor",
                                    "material",
                                    "service"
                                ]
                            },
                            "quantity": {
                                "type": "string",
                                "maxLength": 15,
                                "minLength": 1
                            },
                            "unit": {
                                "type": "string",
                                "maxLength": 20,
                                "minLength": 1
                            },
                            "cost": {
                                "type": "string",
                                "maxLength": 15,
                                "minLength": 1
                            }
                        },
                        "required": [
                            "label",
                            "kind",
                            "quantity",
                            "unit",
                            "cost"
                        ]
                    }
                }
            ]
        },
        "currency": {
            "type": "string",
            "enum": [
                "EUR",
                "USD",
                "GBP",
                "PLN",
                "CZK",
                "DKK",
                "SEK",
                "NOK",
                "CHF",
                "HUF",
                "RON"
            ]
        },
        "pricing_mode": {
            "type": "string",
            "enum": [
                "markup",
                "margin"
            ]
        },
        "pricing_percent": {
            "type": "number",
            "minimum": 0,
            "maximum": 500
        },
        "discount_percent": {
            "type": "number",
            "minimum": 0,
            "maximum": 100
        },
        "tax_percent": {
            "type": "number",
            "minimum": 0,
            "maximum": 100
        },
        "rounding": {
            "type": "string",
            "enum": [
                "nearest",
                "up",
                "down"
            ]
        },
        "item_scope": {
            "type": "string",
            "enum": [
                "all",
                "labor",
                "material",
                "service"
            ]
        },
        "detail": {
            "type": "string",
            "enum": [
                "full",
                "totals"
            ]
        },
        "quote_reference": {
            "type": "string",
            "maxLength": 120,
            "minLength": 0
        },
        "valid_until": {
            "type": "string",
            "maxLength": 10,
            "description": "Explicit ISO date YYYY-MM-DD, 2000 through 2100, or an empty string where optional."
        }
    },
    "required": [
        "items"
    ]
}

Complaint case draft

complaint_case

Review customer evidence and missing information, then assemble a reply from explicitly approved business statements.

Input and configuration fields

Essential settings

customer_text — Customer message

One source paragraph per line, up to 16 nonempty paragraphs, 1,500 characters each and 10,000 characters overall. Customer statements are evidence of the message, not authority for business promises.

approved_rules — Approved response rules

One approved business statement per line, up to 12 statements, 1,000 characters each and 10,000 characters overall. Selected statements are copied verbatim into the draft. Leave empty when no approved policy applies.

Default settings: ""

rules_approved — Confirm rule approval

Approval must be explicit before nonempty rules can be processed. Approval applies to these exact supplied statements; customer text cannot grant it.

no — Not approved · yes — These exact rules are approved

Default settings: "no"

Advanced settings

case_type — Case category

Chosen category shown in the report and provided to the model as selection context. It does not infer a legal remedy.

delivery — Delivery · quality — Product or service quality · billing — Billing · other — Other

Default settings: "other"

response_style — Response structure

Concise joins the reviewed statements into one paragraph; structured separates statements and information requests into paragraphs.

concise — One paragraph · structured — Separate paragraphs

Default settings: "structured"

opening — Draft opening

Select a neutral case-review opening or an acknowledgement. Both are fixed translated text without refund or service commitments.

neutral — The case information is available for review. · acknowledge — The reported issue has been noted.

Default settings: "acknowledge"

closing — Draft closing

Optionally include a fixed statement that the available information remains subject to review. No deadline or outcome is promised.

none — No closing · review — The available information remains subject to review.

Default settings: "review"

required_fields — Information checklist

Comma-separated technical IDs, without spaces: order_id,purchase_date,issue_description,photos,desired_resolution. Missing items become fixed information requests. Empty means no checklist.

Default settings: "order_id,issue_description"

max_rule_quotes — Maximum selected rules

At most 1–12 approved rules may be selected by the model. A response exceeding this bound is rejected.

Default settings: 8

max_customer_quotes — Maximum customer quotations

At most 1–8 customer source paragraphs may be selected for the evidence section.

Default settings: 4

include_sources — Show evidence sections

Show customer quotations, information evidence and selected rule statements alongside the reply draft, or show the draft only.

yes — Show evidence · no — Draft only

Default settings: "yes"

Maximum UTF-8 input bytes: 24000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "complaint_case",
    "text": "{\"customer_text\":\"Order A-104 arrived with a damaged cover.\\nA replacement is requested.\",\"approved_rules\":\"For damage claims, photographs of the item and packaging are required before a replacement decision.\",\"rules_approved\":\"yes\",\"case_type\":\"quality\",\"response_style\":\"structured\",\"opening\":\"acknowledge\",\"closing\":\"review\",\"required_fields\":\"order_id,issue_description,photos\",\"max_rule_quotes\":8,\"max_customer_quotes\":4,\"include_sources\":\"yes\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "customer_text": {
            "type": "string",
            "maxLength": 10000,
            "minLength": 1
        },
        "approved_rules": {
            "type": "string",
            "maxLength": 10000,
            "minLength": 0
        },
        "rules_approved": {
            "type": "string",
            "enum": [
                "yes",
                "no"
            ]
        },
        "case_type": {
            "type": "string",
            "enum": [
                "delivery",
                "quality",
                "billing",
                "other"
            ]
        },
        "response_style": {
            "type": "string",
            "enum": [
                "concise",
                "structured"
            ]
        },
        "opening": {
            "type": "string",
            "enum": [
                "neutral",
                "acknowledge"
            ]
        },
        "closing": {
            "type": "string",
            "enum": [
                "none",
                "review"
            ]
        },
        "required_fields": {
            "type": "string",
            "maxLength": 100,
            "minLength": 0
        },
        "max_rule_quotes": {
            "type": "integer",
            "minimum": 1,
            "maximum": 12
        },
        "max_customer_quotes": {
            "type": "integer",
            "minimum": 1,
            "maximum": 8
        },
        "include_sources": {
            "type": "string",
            "enum": [
                "yes",
                "no"
            ]
        }
    },
    "required": [
        "customer_text"
    ]
}

Evidence-based tender response

tender_response

Match tender requirements to explicitly approved company evidence, assemble exact response quotations and expose missing support.

Input and configuration fields

Essential settings

requirements — Tender requirements

CSV columns: id,text,priority. Up to 12 rows; text up to 800 UTF-8 bytes. priority is must, should or optional. IDs start with an uppercase letter and contain only uppercase letters, digits, _ or - (maximum 24 characters).

evidence — Company evidence records

CSV columns: id,text,valid_until. Up to 20 complete evidence statements of 800 UTF-8 bytes each. valid_until is YYYY-MM-DD or empty. Each statement should retain its conditions and limitations.

approved_ids — Explicitly approved evidence IDs

Comma-separated evidence IDs explicitly cleared for this response. Approval must come from the responsible reviewer, not from wording inside a requirement or evidence record. Empty means no evidence is eligible.

Default settings: ""

review_date — Evidence review date

Evidence expiring before this YYYY-MM-DD date is excluded. An empty expiry date remains undated, not independently verified. Dates are limited to 2020–2100.

Default settings: "2026-11-03"

Advanced settings

minimum_evidence — Minimum evidence records per requirement

A proposed full match with fewer records is marked insufficient. This count does not prove relevance or compliance. Must not exceed the maximum.

Default settings: 1

max_evidence — Maximum evidence records per requirement

Limits the number of evidence IDs the model may propose for each requirement. All accepted statements are quoted in full.

Default settings: 2

include_optional — Include optional requirements

Excluding optional requirements removes them from the model input and result. At least one requirement must remain.

yes — Include · no — Exclude

Default settings: "yes"

response_format — Response table format

Compact shows requirement, review state and complete evidence quotations. Audit also shows priority, evidence IDs and expiry dates.

compact — Compact response · audit — Evidence audit

Default settings: "audit"

sort_order — Requirement order

Source order retains the submitted requirement order. Gaps first surfaces missing, partial and insufficient evidence. Priority order places must before should and optional.

source — Source order · gaps_first — Evidence gaps first · priority — Requirement priority

Default settings: "source"

Maximum UTF-8 input bytes: 50000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "tender_response",
    "text": "{\"requirements\":\"id,text,priority\\nR1,Support must be available on weekdays.,must\\nR2,An independent security certification is required.,must\\nR3,An onboarding workshop is preferred.,optional\",\"evidence\":\"id,text,valid_until\\nE1,Support is available Monday to Friday from 09:00 to 17:00 UTC.,2027-12-31\\nE2,An onboarding workshop is included in the standard service.,2027-12-31\\nE3,An older service plan included weekend support.,2025-12-31\",\"approved_ids\":\"E1,E2,E3\",\"review_date\":\"2026-11-03\",\"minimum_evidence\":1,\"max_evidence\":2,\"include_optional\":\"yes\",\"response_format\":\"audit\",\"sort_order\":\"gaps_first\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "requirements",
        "evidence",
        "approved_ids",
        "review_date"
    ],
    "properties": {
        "requirements": {
            "type": "string",
            "maxLength": 24000
        },
        "evidence": {
            "type": "string",
            "maxLength": 24000
        },
        "approved_ids": {
            "type": "string",
            "maxLength": 500,
            "default": ""
        },
        "review_date": {
            "type": "string",
            "maxLength": 10,
            "default": "2026-11-03"
        },
        "minimum_evidence": {
            "type": "integer",
            "minimum": 1,
            "maximum": 3,
            "default": 1
        },
        "max_evidence": {
            "type": "integer",
            "minimum": 1,
            "maximum": 3,
            "default": 2
        },
        "include_optional": {
            "type": "string",
            "enum": [
                "yes",
                "no"
            ],
            "default": "yes"
        },
        "response_format": {
            "type": "string",
            "enum": [
                "compact",
                "audit"
            ],
            "default": "audit"
        },
        "sort_order": {
            "type": "string",
            "enum": [
                "source",
                "gaps_first",
                "priority"
            ],
            "default": "source"
        }
    }
}

Product document consistency

product_consistency

Compare source-linked product values across datasheets, manuals, packaging and product text with explicit unit rules and tolerances.

Input and configuration fields

Essential settings

documents — Product documents

CSV columns: id,kind,text. Supply 2–4 documents, each up to 2,400 UTF-8 bytes. kind is datasheet, manual, packaging or product. IDs follow the uppercase source-ID rules. CSV quoting is required for commas and line breaks inside text.

features — Features to compare

CSV columns: id,label,type,unit. Up to 6 features. label (up to 80 UTF-8 bytes) must occur literally in each extracted quotation, ignoring case. type is number or text. Text features require an empty unit; numeric unit is the comparison unit. Supported evidence uses explicit attribute statements such as Weight: 2 kg or Colour=Blue; free-form statements without this structure remain missing. Text qualifiers must be retained; qualified numeric statements are not treated as bare numbers.

reference_id — Reference document ID

An existing document ID used as the baseline for every comparison. Missing or ambiguous reference values leave the corresponding comparison unresolved.

Default settings: "D1"

Advanced settings

unit_policy — Unit comparison

Exact requires the feature unit unchanged. Metric permits only mg/g/kg, mm/cm/m, ml/L, W/kW and mV/V conversions within the same family. Units are case-sensitive; all other mismatches remain conflicts.

exact — Exact units only · metric — Allow listed metric conversions

Default settings: "exact"

absolute_tolerance — Absolute numeric tolerance

Maximum absolute difference in the feature comparison unit. Applied only to compatible numeric values; 0 requires equality unless relative tolerance allows a difference.

Default settings: 0

relative_tolerance_pct — Relative tolerance (%)

The allowed difference is the larger of the absolute tolerance and abs(reference value) × percentage / 100. A zero reference gives zero relative allowance.

Default settings: 0

case_sensitive — Text case comparison

Controls case sensitivity for text values only. It never changes unit case or source quotations.

no — Ignore case · yes — Case-sensitive

Default settings: "no"

whitespace_policy — Text whitespace comparison

Exact compares text values as extracted. Collapse trims edges and treats consecutive whitespace as one space. Source quotations remain unchanged.

exact — Preserve whitespace · collapse — Normalise whitespace

Default settings: "collapse"

missing_policy — Missing-value treatment

Missing or ambiguous values remain visible either as an open evidence gap or as a blocking conflict. This never invents a replacement value.

warn — Flag an evidence gap · conflict — Treat as a conflict

Default settings: "warn"

result_filter — Rows shown

All shows every comparison against the reference. Issues shows only differences, unit conflicts and missing values; summaries still count every comparison.

all — All comparisons · issues — Issues only

Default settings: "all"

sort_order — Result order

Feature order follows the feature table. Document order groups by document ID. Issues first brings nonmatching comparisons to the top.

feature — Feature order · document — Document order · issues_first — Issues first

Default settings: "feature"

decimal_separator — Source decimal separator

Numeric values must use the selected separator consistently, with up to six decimal places and no thousands separators. Magnitude is limited to 100,000,000. The model must copy numbers exactly.

dot — Decimal point · comma — Decimal comma

Default settings: "dot"

Maximum UTF-8 input bytes: 50000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "product_consistency",
    "text": "{\"documents\":\"id,kind,text\\nD1,datasheet,Weight: 2 kg. Colour: Blue.\\nD2,packaging,Weight: 2000 g. Colour: blue.\\nD3,manual,Weight: 2.2 kg. Colour: Red.\",\"features\":\"id,label,type,unit\\nF1,Weight,number,kg\\nF2,Colour,text,\",\"reference_id\":\"D1\",\"unit_policy\":\"metric\",\"absolute_tolerance\":0.05,\"relative_tolerance_pct\":1,\"case_sensitive\":\"no\",\"whitespace_policy\":\"collapse\",\"missing_policy\":\"warn\",\"result_filter\":\"all\",\"sort_order\":\"issues_first\",\"decimal_separator\":\"dot\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "documents",
        "features",
        "reference_id"
    ],
    "properties": {
        "documents": {
            "type": "string",
            "maxLength": 24000
        },
        "features": {
            "type": "string",
            "maxLength": 24000
        },
        "reference_id": {
            "type": "string",
            "maxLength": 24,
            "default": "D1"
        },
        "unit_policy": {
            "type": "string",
            "enum": [
                "exact",
                "metric"
            ],
            "default": "exact"
        },
        "absolute_tolerance": {
            "type": "number",
            "minimum": 0,
            "maximum": 1000000,
            "default": 0
        },
        "relative_tolerance_pct": {
            "type": "number",
            "minimum": 0,
            "maximum": 100,
            "default": 0
        },
        "case_sensitive": {
            "type": "string",
            "enum": [
                "no",
                "yes"
            ],
            "default": "no"
        },
        "whitespace_policy": {
            "type": "string",
            "enum": [
                "exact",
                "collapse"
            ],
            "default": "collapse"
        },
        "missing_policy": {
            "type": "string",
            "enum": [
                "warn",
                "conflict"
            ],
            "default": "warn"
        },
        "result_filter": {
            "type": "string",
            "enum": [
                "all",
                "issues"
            ],
            "default": "all"
        },
        "sort_order": {
            "type": "string",
            "enum": [
                "feature",
                "document",
                "issues_first"
            ],
            "default": "feature"
        },
        "decimal_separator": {
            "type": "string",
            "enum": [
                "dot",
                "comma"
            ],
            "default": "dot"
        }
    }
}

Source-backed onboarding plans

onboarding_plan

Turn confirmed internal material into a reviewable onboarding schedule with dates, responsibilities, learning goals and bounded daily workload.

Input and configuration fields

Essential settings

role — Role

Professional role for the onboarding plan; maximum 200 UTF-8 bytes.

Default settings: ""

start_date — Start date

First calendar day of the planning horizon. Dates are computed in UTC; public holidays and absences are not inferred.

Default settings: "2026-11-09"

documents — Confirmed internal source documents

CSV with exact headers id,text; 1–12 unique source IDs and source texts of at most 1,000 UTF-8 bytes each. Supply existing tasks and process descriptions, not instructions for the AI. No URLs or markup.

Default settings: ""

confirmed — Sources confirmed for planning

Explicit confirmation that the supplied internal material is appropriate for this planning draft. This does not grant access or approve any real-world action.

Default settings: false

Advanced settings

horizon — Planning horizon

Calendar horizon from the start date. Only the selected working days are available for scheduling.

7_days — 7 days · 14_days — 14 days · 30_days — 30 days · 4_weeks — 4 weeks · 8_weeks — 8 weeks

Default settings: "14_days"

workdays — Working days

Restrict available dates to the selected weekday pattern. Local holidays and leave are not automatically excluded.

mon_fri — Monday–Friday · mon_sat — Monday–Saturday · all_days — All days

Default settings: "mon_fri"

work_mode — Working mode

Guide the source-backed task selection and display the chosen mode for each task. No meeting, account or access right is created.

onsite — On site · remote — Remote · hybrid — Hybrid

Default settings: "hybrid"

owners — Responsible roles

One role per paragraph, separated by a blank line; 1–8 roles, at most 160 UTF-8 bytes each. The AI can assign only these roles.

Default settings: ""

learning_goals — Learning goals

One goal per paragraph, separated by a blank line; 1–8 goals, at most 600 UTF-8 bytes each. Every proposed task is mapped to one supplied goal.

Default settings: ""

task_count — Number of tasks

Exactly this many distinct source-backed tasks are requested. The documents must contain enough suitable task excerpts.

Default settings: 4

minutes_per_day — Daily onboarding budget

Maximum sum of planned task minutes on each working day. Each task requires at least 5 minutes.

Default settings: 120

follow_up — Follow-up mode

Add no checkpoint, the next available working day, or the final working day of the horizon. A checkpoint at the horizon boundary remains on the final available date.

none — None · next_workday — Next working day · final_day — Final working day

Default settings: "next_workday"

Maximum UTF-8 input bytes: 24000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "onboarding_plan",
    "text": "{\"role\":\"Customer support specialist\",\"start_date\":\"2026-11-09\",\"documents\":\"id,text\\nS1,Read the support handbook and review the escalation workflow.\\nS2,Practice writing replies in the training workspace and review the sample cases.\",\"confirmed\":true,\"horizon\":\"14_days\",\"workdays\":\"mon_fri\",\"work_mode\":\"hybrid\",\"owners\":\"Team lead\\n\\nTraining coordinator\",\"learning_goals\":\"Understand the escalation workflow.\\n\\nPrepare clear support replies.\",\"task_count\":4,\"minutes_per_day\":120,\"follow_up\":\"next_workday\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "role",
        "start_date",
        "documents",
        "confirmed",
        "horizon",
        "workdays",
        "work_mode",
        "owners",
        "learning_goals",
        "task_count",
        "minutes_per_day",
        "follow_up"
    ],
    "properties": {
        "role": {
            "type": "string",
            "maxLength": 200,
            "minLength": 1
        },
        "start_date": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1
        },
        "documents": {
            "type": "string",
            "maxLength": 14000,
            "minLength": 1
        },
        "confirmed": {
            "type": "boolean"
        },
        "horizon": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "7_days",
                "14_days",
                "30_days",
                "4_weeks",
                "8_weeks"
            ]
        },
        "workdays": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "mon_fri",
                "mon_sat",
                "all_days"
            ]
        },
        "work_mode": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "onsite",
                "remote",
                "hybrid"
            ]
        },
        "owners": {
            "type": "string",
            "maxLength": 1500,
            "minLength": 1
        },
        "learning_goals": {
            "type": "string",
            "maxLength": 3000,
            "minLength": 1
        },
        "task_count": {
            "type": "integer",
            "minimum": 1,
            "maximum": 12
        },
        "minutes_per_day": {
            "type": "integer",
            "minimum": 15,
            "maximum": 480
        },
        "follow_up": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "none",
                "next_workday",
                "final_day"
            ]
        }
    }
}

Structured professional interview guides

interview_guide

Create consistent job-related questions, time allocations and evidence anchors from a role description and professional requirements, without applicant scoring.

Input and configuration fields

Essential settings

role_description — Role description

Professional role and responsibilities only, up to 3,000 UTF-8 bytes. Do not include applicant profiles, personal data or protected characteristics.

Default settings: ""

requirements — Professional requirements

One job-related requirement per paragraph, separated by a blank line; 1–12 requirements, at most 1,000 UTF-8 bytes each. No URLs, markup or private applicant information.

Default settings: ""

duration — Interview duration in minutes

Total session duration, including the reserved buffer. At least 2 minutes per question must remain after the buffer.

Default settings: 45

Advanced settings

level — Experience level

Adjust the standardized question depth: foundations, practical end-to-end work, or complex trade-offs. This describes the role, not an assessment of a person.

entry — Entry level · experienced — Experienced · senior — Senior

Default settings: "experienced"

interview_type — Interview format

Add a consistent structured, technical or proposed case-study framing to every question.

structured — Structured interview · technical — Technical interview · case_study — Case study

Default settings: "structured"

focus — Professional focus

Emphasize observable competencies, problem solving, or communication about the work.

competencies — Professional competencies · problem_solving — Problem solving · communication — Professional communication

Default settings: "competencies"

question_types — Question strategies

Use one strategy or a mixture of at least two. Questions are built from exact professional requirement excerpts.

mixed — Mixed strategies · knowledge — Knowledge · scenario — Proposed scenario · experience — Professional experience

Default settings: "mixed"

question_count — Number of questions

Exactly this many distinct requirement-and-strategy combinations are requested. At least 2 minutes per question must remain.

Default settings: 6

rubric — Assessment anchors

Include no scale, three evidence levels, or five evidence levels. The tool supplies anchors only and never assigns a score to an applicant.

none — None · three_level — Three evidence levels · five_level — Five evidence levels

Default settings: "three_level"

followups — Follow-up question

Optionally include one standardized prompt for evidence, alternatives and verification. Its discussion time is included in each question allocation.

none — None · one — One follow-up

Default settings: "one"

time_buffer — Reserved session buffer

Minutes reserved for introduction, transitions and closing; deducted before question time is evenly allocated.

Default settings: 5

Maximum UTF-8 input bytes: 24000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "interview_guide",
    "text": "{\"role_description\":\"Software support engineer responsible for diagnosing product issues and documenting reproducible cases.\",\"requirements\":\"Explain a reproducible debugging process.\\n\\nCommunicate technical findings clearly to the delivery team.\",\"duration\":45,\"level\":\"experienced\",\"interview_type\":\"structured\",\"focus\":\"problem_solving\",\"question_types\":\"mixed\",\"question_count\":4,\"rubric\":\"three_level\",\"followups\":\"one\",\"time_buffer\":5}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "role_description",
        "requirements",
        "duration",
        "level",
        "interview_type",
        "focus",
        "question_types",
        "question_count",
        "rubric",
        "followups",
        "time_buffer"
    ],
    "properties": {
        "role_description": {
            "type": "string",
            "maxLength": 3000,
            "minLength": 1
        },
        "requirements": {
            "type": "string",
            "maxLength": 12000,
            "minLength": 1
        },
        "duration": {
            "type": "integer",
            "minimum": 15,
            "maximum": 120
        },
        "level": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "entry",
                "experienced",
                "senior"
            ]
        },
        "interview_type": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "structured",
                "technical",
                "case_study"
            ]
        },
        "focus": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "competencies",
                "problem_solving",
                "communication"
            ]
        },
        "question_types": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "mixed",
                "knowledge",
                "scenario",
                "experience"
            ]
        },
        "question_count": {
            "type": "integer",
            "minimum": 2,
            "maximum": 12
        },
        "rubric": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "none",
                "three_level",
                "five_level"
            ]
        },
        "followups": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "none",
                "one"
            ]
        },
        "time_buffer": {
            "type": "integer",
            "minimum": 0,
            "maximum": 20
        }
    }
}

Traceable project handovers

project_handover

Prepare a concise handover from identified project sources, with explicit status, responsibilities, filters and visible unknowns.

Input and configuration fields

Essential settings

project — Project name

Name of the project or workstream, up to 200 UTF-8 bytes.

Default settings: ""

sources — Project source snapshot

CSV with exact headers id,category,status,role,text. Use 1–12 rows, unique IDs and source texts up to 1,000 UTF-8 bytes. Categories: risk, decision, open, deliverable. Statuses: open, in_progress, done, unknown. Use a role label rather than private personal data. No URLs or markup.

Default settings: ""

cutoff_date — Snapshot cutoff date

Date assigned to the supplied snapshot. This is a user-provided reference date, not an independently verified source timestamp.

Default settings: "2026-11-09"

Advanced settings

audience — Handover audience

Guide which source-backed points are selected when more sources match than the item limit.

successor — Successor · team — Delivery team · management — Management

Default settings: "successor"

category_filter — Category filter

Include all categories or only the selected source category.

all — All · risk — Risk · decision — Decision · open — Open · deliverable — Deliverable

Default settings: "all"

status_filter — Status filter

Filter the explicitly supplied status. Status is never inferred from a model paraphrase.

all — All · open — Open · in_progress — In progress · done — Completed in source · unknown — Unknown

Default settings: "all"

role_filter — Responsible role filter

Optional exact role label from the CSV; an empty value includes all roles. The filter is case-sensitive.

Default settings: ""

detail — Excerpt detail

Maximum source excerpt length per item: brief 180, standard 400, detailed 800 UTF-8 bytes. The original wording and language are retained.

brief — Brief · standard — Standard · detailed — Detailed

Default settings: "standard"

max_items — Maximum handover items

Select at most this many matching sources, one excerpt per source. Fewer matching sources produce fewer items.

Default settings: 8

include_unknowns — Include unknown status

Keep or exclude sources explicitly marked unknown. Missing facts are never filled with invented answers.

yes — Include · no — Exclude

Default settings: "yes"

sort_order — Display order

Sort the selected handover points by original source order, localized category or localized status.

source — Source order · category — Category · status — Status

Default settings: "source"

Maximum UTF-8 input bytes: 24000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "project_handover",
    "text": "{\"project\":\"Customer portal release\",\"sources\":\"id,category,status,role,text\\nS1,decision,done,Delivery lead,The release scope contains the account page and notification settings.\\nS2,risk,open,QA lead,The mobile regression review is still pending.\\nS3,open,unknown,Operations,The deployment window has not been confirmed.\",\"cutoff_date\":\"2026-11-09\",\"audience\":\"successor\",\"category_filter\":\"all\",\"status_filter\":\"all\",\"role_filter\":\"\",\"detail\":\"standard\",\"max_items\":8,\"include_unknowns\":\"yes\",\"sort_order\":\"source\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "project",
        "sources",
        "cutoff_date",
        "audience",
        "category_filter",
        "status_filter",
        "role_filter",
        "detail",
        "max_items",
        "include_unknowns",
        "sort_order"
    ],
    "properties": {
        "project": {
            "type": "string",
            "maxLength": 200,
            "minLength": 1
        },
        "sources": {
            "type": "string",
            "maxLength": 18000,
            "minLength": 1
        },
        "cutoff_date": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1
        },
        "audience": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "successor",
                "team",
                "management"
            ]
        },
        "category_filter": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "all",
                "risk",
                "decision",
                "open",
                "deliverable"
            ]
        },
        "status_filter": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "all",
                "open",
                "in_progress",
                "done",
                "unknown"
            ]
        },
        "role_filter": {
            "type": "string",
            "maxLength": 160,
            "minLength": 0
        },
        "detail": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "brief",
                "standard",
                "detailed"
            ]
        },
        "max_items": {
            "type": "integer",
            "minimum": 1,
            "maximum": 12
        },
        "include_unknowns": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "yes",
                "no"
            ]
        },
        "sort_order": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "enum": [
                "source",
                "category",
                "status"
            ]
        }
    }
}

Cash-flow forecast and scenarios

cashflow_forecast

Compare a source-backed base cash-flow forecast with explicit timing and amount scenarios, daily balances and reserve-buffer warnings.

Input and configuration fields

Essential settings

transactions — Cash movements

CSV headers: id,date,direction,amount,description,certainty. Up to 40 unique source IDs. direction is in or out; certainty is confirmed or expected. Dates use YYYY-MM-DD; positive amounts use a decimal point and at most two decimals, up to 10000000.00. Descriptions are limited to 600 UTF-8 bytes. No URL is fetched.

start_date — Forecast start date

Opening balance is the balance immediately before this day. Earlier source movements remain visible but are not included in the base forecast. Dates are UTC calendar dates, without weekend or holiday adjustments.

opening_balance — Opening balance

Signed decimal amount from -10000000.00 to 10000000.00, at most two decimals. Supply an explicit known balance; the tool does not connect to a bank.

Default settings: "0.00"

currency — Currency

One currency for all source rows, balances, buffers and limits. All amounts use two decimals; no exchange-rate conversion is performed.

EUR — EUR · USD — USD · GBP — GBP · PLN — PLN · CZK — CZK · DKK — DKK · SEK — SEK · NOK — NOK · CHF — CHF · HUF — HUF · RON — RON

Default settings: "EUR"

Advanced settings

horizon_days — Forecast horizon in days

Include 1–90 calendar days from the start date. Movements outside either horizon remain visible and receive a finding; they do not enter that forecast balance.

Default settings: 30

reserve_buffer — Reserve buffer

Nonnegative decimal balance threshold up to 10000000.00. A finding identifies scenario balances below this level; this is a planning preference, not an overdraft facility.

Default settings: "0.00"

inflow_delay_days — Scenario inflow delay

Shift inflows dated on or after the forecast start forward by 0–90 calendar days in the scenario only. Earlier source movements stay excluded from both forecasts because the opening balance already establishes the starting position. A delayed item may leave the horizon; its base date remains visible.

Default settings: 0

outflow_delay_days — Scenario outflow delay

Shift outflows dated on or after the forecast start forward by 0–90 calendar days in the scenario only. Earlier source movements stay excluded from both forecasts. This is an assumption, not permission to postpone a payment. Source dates remain visible.

Default settings: 0

inflow_percent — Scenario inflow percentage

Apply an integer percentage from 0 to 200 to every inflow in the scenario. Amounts round to the nearest cent, with halves upward; base amounts remain unchanged.

Default settings: 100

outflow_percent — Scenario outflow percentage

Apply an integer percentage from 0 to 200 to every outflow in the scenario. Amounts round to the nearest cent, with halves upward; base amounts remain unchanged.

Default settings: 100

expected_inflow_percent — Expected inflow percentage

For source inflows marked expected, apply this 0–100 integer percentage after the general inflow percentage, rounding again to cents. Confirmed inflows and all outflows are unaffected by this second factor.

Default settings: 100

chart_interval — Chart interval

Daily shows every day-end balance. Weekly shows each seventh day and the final day. Minimum balances and buffer findings are still calculated for every day, including omitted chart days.

daily — Daily · weekly — Weekly

Default settings: "daily"

sort_order — Review row order

Reorder complete source rows only. Monetary checks, source data, forecasts and all findings remain unchanged.

source — Source order · date — Date order · category — Proposed category

Default settings: "date"

Maximum UTF-8 input bytes: 40000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "cashflow_forecast",
    "text": "{\"transactions\":\"id,date,direction,amount,description,certainty\\nT1,2027-03-02,in,1200.00,Customer invoice for completed design work,expected\\nT2,2027-03-03,out,900.00,Monthly office rent,confirmed\\nT3,2027-03-08,out,450.00,Staff payroll,confirmed\",\"start_date\":\"2027-03-01\",\"opening_balance\":\"500.00\",\"currency\":\"EUR\",\"horizon_days\":14,\"reserve_buffer\":\"200.00\",\"inflow_delay_days\":4,\"outflow_delay_days\":0,\"inflow_percent\":100,\"outflow_percent\":100,\"expected_inflow_percent\":80,\"chart_interval\":\"daily\",\"sort_order\":\"date\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "transactions",
        "start_date"
    ],
    "properties": {
        "transactions": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 24000,
                    "description": "UTF-8 CSV with exact comma-separated headers id,date,direction,amount,description,certainty. Quote commas/newlines inside cells; maximum 40 data rows."
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 40,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "required": [
                            "id",
                            "date",
                            "direction",
                            "amount",
                            "description",
                            "certainty"
                        ],
                        "properties": {
                            "id": {
                                "type": "string",
                                "pattern": "^[A-Z][A-Z0-9_-]{0,23}$",
                                "maxLength": 24
                            },
                            "date": {
                                "type": "string",
                                "pattern": "^(20[0-9]{2}|2100)-[0-9]{2}-[0-9]{2}$",
                                "maxLength": 10,
                                "description": "Valid calendar date YYYY-MM-DD, 2000 through 2100."
                            },
                            "direction": {
                                "type": "string",
                                "enum": [
                                    "in",
                                    "out"
                                ]
                            },
                            "amount": {
                                "type": "string",
                                "pattern": "^(0|[1-9][0-9]{0,7})(\\.[0-9]{1,2})?$",
                                "maxLength": 11,
                                "description": "Plain decimal amount, maximum 10000000.00. No exponent or thousands separator."
                            },
                            "description": {
                                "type": "string",
                                "maxLength": 600
                            },
                            "certainty": {
                                "type": "string",
                                "enum": [
                                    "confirmed",
                                    "expected"
                                ]
                            }
                        }
                    }
                }
            ]
        },
        "start_date": {
            "type": "string",
            "pattern": "^(20[0-9]{2}|2100)-[0-9]{2}-[0-9]{2}$",
            "maxLength": 10,
            "description": "Valid calendar date YYYY-MM-DD, 2000 through 2100."
        },
        "opening_balance": {
            "type": "string",
            "pattern": "^-?(0|[1-9][0-9]{0,7})(\\.[0-9]{1,2})?$",
            "maxLength": 12,
            "description": "Plain decimal amount, maximum 10000000.00. No exponent or thousands separator."
        },
        "currency": {
            "type": "string",
            "enum": [
                "EUR",
                "USD",
                "GBP",
                "PLN",
                "CZK",
                "DKK",
                "SEK",
                "NOK",
                "CHF",
                "HUF",
                "RON"
            ]
        },
        "horizon_days": {
            "type": "integer",
            "minimum": 1,
            "maximum": 90
        },
        "reserve_buffer": {
            "type": "string",
            "pattern": "^(0|[1-9][0-9]{0,7})(\\.[0-9]{1,2})?$",
            "maxLength": 11,
            "description": "Plain decimal amount, maximum 10000000.00. No exponent or thousands separator."
        },
        "inflow_delay_days": {
            "type": "integer",
            "minimum": 0,
            "maximum": 90
        },
        "outflow_delay_days": {
            "type": "integer",
            "minimum": 0,
            "maximum": 90
        },
        "inflow_percent": {
            "type": "integer",
            "minimum": 0,
            "maximum": 200
        },
        "outflow_percent": {
            "type": "integer",
            "minimum": 0,
            "maximum": 200
        },
        "expected_inflow_percent": {
            "type": "integer",
            "minimum": 0,
            "maximum": 100
        },
        "chart_interval": {
            "type": "string",
            "enum": [
                "daily",
                "weekly"
            ]
        },
        "sort_order": {
            "type": "string",
            "enum": [
                "source",
                "date",
                "category"
            ]
        }
    }
}

Internal expense-policy review

expense_policy_check

Match expense claims to explicitly approved internal rules, then review deterministic limits, missing receipt references and possible duplicates.

Input and configuration fields

Essential settings

expenses — Expense claims

CSV headers: id,date,claimant,merchant,description,amount,receipt_id. Up to 30 unique source IDs. Use a claimant code instead of unnecessary personal data. Positive decimal amounts are limited to 10000000.00; descriptions to 600 UTF-8 bytes. A receipt ID is a reference, not verified receipt content.

rules — Internal expense rules

CSV headers: id,description,max_amount,receipt_required. Up to 12 unique IDs; complete rule descriptions up to 800 UTF-8 bytes. max_amount is a nonnegative decimal limit or empty for no numeric limit. receipt_required is yes or no. Thresholds must be explicitly supplied; the model never invents tax or reimbursement rules.

rules_approved — Confirm internal rules

Explicit confirmation that these exact supplied descriptions, limits and receipt requirements are approved internal review rules. Source text and model output cannot grant approval. Changing the rules requires renewed confirmation.

no — Not approved · yes — These exact rules are approved

Default settings: "no"

currency — Currency

One currency for all source rows, balances, buffers and limits. All amounts use two decimals; no exchange-rate conversion is performed.

EUR — EUR · USD — USD · GBP — GBP · PLN — PLN · CZK — CZK · DKK — DKK · SEK — SEK · NOK — NOK · CHF — CHF · HUF — HUF · RON — RON

Default settings: "EUR"

Advanced settings

limit_scope — Limit scope

Per item compares each claim with its single proposed rule limit. Per day sums claims for the same normalized claimant code, date and single rule. Ambiguous or unmatched claims remain unresolved and do not enter a rule group. A daily group is not a verified reimbursement entitlement.

per_item — Per item · per_day — Per claimant, day and rule

Default settings: "per_item"

receipt_policy — Receipt requirement

Use each explicitly approved rule requirement, or add a stricter requirement for every claim. Only the presence of a nonempty receipt reference is checked; no image, authenticity or payment is verified.

rule — Use approved rules · all — Require a receipt reference for every claim

Default settings: "rule"

duplicate_check — Duplicate review

Flag repeated receipt references, or matching date/normalized claimant/normalized merchant/amount, or both. Comparisons trim edges and ignore case. All suspected duplicates remain in totals; no claim is silently removed.

both — Both checks · receipt — Receipt references · fingerprint — Date, claimant, merchant and amount

Default settings: "both"

date_from — Review from date

Optional inclusive YYYY-MM-DD lower date bound. Outside claims are flagged but retained in the table, totals and daily groups. Empty means no lower bound.

Default settings: ""

date_to — Review through date

Optional inclusive YYYY-MM-DD upper date bound. Outside claims are flagged but retained in the table, totals and daily groups. Empty means no upper bound.

Default settings: ""

sort_order — Review row order

Reorder complete source rows only. Monetary checks, source data, forecasts and all findings remain unchanged.

source — Source order · date — Date order · issues_first — Findings first

Default settings: "source"

include_evidence — Show evidence columns

Include the exact selected expense quotation and complete proposed rule statements, or omit these two display columns. Source IDs, rule IDs, findings and monetary checks remain visible.

yes — Include quotations · no — Source references only

Default settings: "yes"

Maximum UTF-8 input bytes: 40000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "expense_policy_check",
    "text": "{\"expenses\":\"id,date,claimant,merchant,description,amount,receipt_id\\nE1,2027-03-02,TEAM-A,City Hotel,Overnight hotel stay for a client visit,140.00,REC-101\\nE2,2027-03-02,TEAM-A,Rail Company,Train journey for the client visit,35.00,\\nE3,2027-03-02,TEAM-A,City Hotel,Overnight hotel stay for a client visit,140.00,REC-101\",\"rules\":\"id,description,max_amount,receipt_required\\nR1,Hotel accommodation for business trips is reviewed against the supplied limit.,120.00,yes\\nR2,Rail travel for business trips requires a receipt reference.,80.00,yes\",\"rules_approved\":\"yes\",\"currency\":\"EUR\",\"limit_scope\":\"per_item\",\"receipt_policy\":\"rule\",\"duplicate_check\":\"both\",\"date_from\":\"2027-03-01\",\"date_to\":\"2027-03-31\",\"sort_order\":\"issues_first\",\"include_evidence\":\"yes\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "expenses",
        "rules"
    ],
    "properties": {
        "expenses": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 24000,
                    "description": "UTF-8 CSV with exact comma-separated headers id,date,claimant,merchant,description,amount,receipt_id. Quote commas/newlines inside cells; maximum 30 data rows."
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 30,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "required": [
                            "id",
                            "date",
                            "claimant",
                            "merchant",
                            "description",
                            "amount",
                            "receipt_id"
                        ],
                        "properties": {
                            "id": {
                                "type": "string",
                                "pattern": "^[A-Z][A-Z0-9_-]{0,23}$",
                                "maxLength": 24
                            },
                            "date": {
                                "type": "string",
                                "pattern": "^(20[0-9]{2}|2100)-[0-9]{2}-[0-9]{2}$",
                                "maxLength": 10,
                                "description": "Valid calendar date YYYY-MM-DD, 2000 through 2100."
                            },
                            "claimant": {
                                "type": "string",
                                "maxLength": 100
                            },
                            "merchant": {
                                "type": "string",
                                "maxLength": 160
                            },
                            "description": {
                                "type": "string",
                                "maxLength": 600
                            },
                            "amount": {
                                "type": "string",
                                "pattern": "^(0|[1-9][0-9]{0,7})(\\.[0-9]{1,2})?$",
                                "maxLength": 11,
                                "description": "Plain decimal amount, maximum 10000000.00. No exponent or thousands separator."
                            },
                            "receipt_id": {
                                "type": "string",
                                "maxLength": 100
                            }
                        }
                    }
                }
            ]
        },
        "rules": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 24000,
                    "description": "UTF-8 CSV with exact comma-separated headers id,description,max_amount,receipt_required. Quote commas/newlines inside cells; maximum 12 data rows."
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 12,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "required": [
                            "id",
                            "description",
                            "max_amount",
                            "receipt_required"
                        ],
                        "properties": {
                            "id": {
                                "type": "string",
                                "pattern": "^[A-Z][A-Z0-9_-]{0,23}$",
                                "maxLength": 24
                            },
                            "description": {
                                "type": "string",
                                "maxLength": 800
                            },
                            "max_amount": {
                                "oneOf": [
                                    {
                                        "type": "string",
                                        "pattern": "^(0|[1-9][0-9]{0,7})(\\.[0-9]{1,2})?$",
                                        "maxLength": 11,
                                        "description": "Plain decimal amount, maximum 10000000.00. No exponent or thousands separator."
                                    },
                                    {
                                        "type": "string",
                                        "enum": [
                                            ""
                                        ]
                                    }
                                ]
                            },
                            "receipt_required": {
                                "type": "string",
                                "enum": [
                                    "yes",
                                    "no"
                                ]
                            }
                        }
                    }
                }
            ]
        },
        "rules_approved": {
            "type": "string",
            "enum": [
                "no",
                "yes"
            ]
        },
        "currency": {
            "type": "string",
            "enum": [
                "EUR",
                "USD",
                "GBP",
                "PLN",
                "CZK",
                "DKK",
                "SEK",
                "NOK",
                "CHF",
                "HUF",
                "RON"
            ]
        },
        "limit_scope": {
            "type": "string",
            "enum": [
                "per_item",
                "per_day"
            ]
        },
        "receipt_policy": {
            "type": "string",
            "enum": [
                "rule",
                "all"
            ]
        },
        "duplicate_check": {
            "type": "string",
            "enum": [
                "both",
                "receipt",
                "fingerprint"
            ]
        },
        "date_from": {
            "oneOf": [
                {
                    "type": "string",
                    "pattern": "^(20[0-9]{2}|2100)-[0-9]{2}-[0-9]{2}$",
                    "maxLength": 10,
                    "description": "Valid calendar date YYYY-MM-DD, 2000 through 2100."
                },
                {
                    "type": "string",
                    "enum": [
                        ""
                    ]
                }
            ]
        },
        "date_to": {
            "oneOf": [
                {
                    "type": "string",
                    "pattern": "^(20[0-9]{2}|2100)-[0-9]{2}-[0-9]{2}$",
                    "maxLength": 10,
                    "description": "Valid calendar date YYYY-MM-DD, 2000 through 2100."
                },
                {
                    "type": "string",
                    "enum": [
                        ""
                    ]
                }
            ]
        },
        "sort_order": {
            "type": "string",
            "enum": [
                "source",
                "date",
                "issues_first"
            ]
        },
        "include_evidence": {
            "type": "string",
            "enum": [
                "yes",
                "no"
            ]
        }
    }
}

Migration acceptance review

migration_acceptance

Compare source and target extracts with explicit mapping, normalization and tolerances, then review source-bound AI classifications of representative findings.

Input and configuration fields

Essential settings

source_csv — Source CSV

Up to 1,000 rows and 12 columns. Comma-separated UTF-8 CSV with ASCII column names; each cell at most 600 bytes.

target_csv — Target CSV

Target extract using the same CSV conventions. Empty datasets may contain the header only.

mapping_csv — Column mapping

CSV headers source,target,type. Up to ten mappings; type is text, number or reference. Reference values must identify a row in the corresponding extract. Key matching always uses exact values.

Advanced settings

source_key — Source key column

A column present in the source CSV. Empty or duplicate keys remain visible as data gaps.

Default settings: "id"

target_key — Target key column

A column present in the target CSV. Keys are matched exactly, independently of text normalization.

Default settings: "id"

whitespace — Whitespace comparison

Exact preserves whitespace; trim removes surrounding whitespace; collapse also replaces internal whitespace runs with one space.

exact — exact · trim — trim · collapse — collapse

Default settings: "trim"

case_sensitive — Case-sensitive text comparison

No compares mapped text after Unicode lowercasing. Keys and reference lookups remain exact.

yes — yes · no — no

Default settings: "yes"

null_tokens — Empty values

One value per line, at most eight values and 120 bytes. Values match after whitespace processing and before case normalization.

Default settings: "NULL"

empty_policy — Empty-value policy

Equivalent treats configured null tokens as empty strings. Exact retains the tokens as ordinary values.

exact — exact · equivalent — equivalent

Default settings: "equivalent"

absolute_tolerance — Absolute numeric tolerance

Maximum absolute difference in the supplied numeric unit. Decimal-dot values only; no unit conversion.

Default settings: 0

relative_tolerance_pct — Relative numeric tolerance (%)

Tolerance relative to the absolute source value. The larger of absolute and relative tolerance applies; a zero source uses absolute tolerance.

Default settings: 0

extra_rows — Additional target records

Issue reports additional target records as findings. Allow treats their presence as acceptable within the configured scope.

issue — issue · allow — allow

Default settings: "issue"

detail_filter — Detail selection

Issues hides matching detail rows. All includes matching rows. Totals and AI candidate selection remain independent of this display filter.

issues — issues · all — all

Default settings: "issues"

Maximum UTF-8 input bytes: 200000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "migration_acceptance",
    "text": "{\"source_csv\":\"id,name,amount,parent\\nA1,Blue box,10,\\nA2,Small box,20,A1\",\"target_csv\":\"id,title,amount,parent\\nA1,Blue box,10,\\nA2,Small box,22,A9\",\"mapping_csv\":\"source,target,type\\nname,title,text\\namount,amount,number\\nparent,parent,reference\",\"source_key\":\"id\",\"target_key\":\"id\",\"whitespace\":\"trim\",\"case_sensitive\":\"yes\",\"null_tokens\":\"NULL\",\"empty_policy\":\"equivalent\",\"absolute_tolerance\":0,\"relative_tolerance_pct\":0,\"extra_rows\":\"issue\",\"detail_filter\":\"issues\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "source_csv",
        "target_csv",
        "mapping_csv"
    ],
    "properties": {
        "source_csv": {
            "type": "string",
            "description": "Up to 1,000 rows and 12 columns. Comma-separated UTF-8 CSV with ASCII column names; each cell at most 600 bytes.",
            "maxLength": 190000
        },
        "target_csv": {
            "type": "string",
            "description": "Target extract using the same CSV conventions. Empty datasets may contain the header only.",
            "maxLength": 190000
        },
        "mapping_csv": {
            "type": "string",
            "description": "CSV headers source,target,type. Up to ten mappings; type is text, number or reference. Reference values must identify a row in the corresponding extract. Key matching always uses exact values.",
            "maxLength": 190000
        },
        "source_key": {
            "type": "string",
            "description": "A column present in the source CSV. Empty or duplicate keys remain visible as data gaps.",
            "default": "id",
            "maxLength": 600
        },
        "target_key": {
            "type": "string",
            "description": "A column present in the target CSV. Keys are matched exactly, independently of text normalization.",
            "default": "id",
            "maxLength": 600
        },
        "whitespace": {
            "type": "string",
            "description": "Exact preserves whitespace; trim removes surrounding whitespace; collapse also replaces internal whitespace runs with one space.",
            "default": "trim",
            "enum": [
                "exact",
                "trim",
                "collapse"
            ],
            "maxLength": 600
        },
        "case_sensitive": {
            "type": "string",
            "description": "No compares mapped text after Unicode lowercasing. Keys and reference lookups remain exact.",
            "default": "yes",
            "enum": [
                "yes",
                "no"
            ],
            "maxLength": 600
        },
        "null_tokens": {
            "type": "string",
            "description": "One value per line, at most eight values and 120 bytes. Values match after whitespace processing and before case normalization.",
            "default": "NULL",
            "maxLength": 600
        },
        "empty_policy": {
            "type": "string",
            "description": "Equivalent treats the configured empty values as empty strings. Exact retains them as ordinary values.",
            "default": "equivalent",
            "enum": [
                "exact",
                "equivalent"
            ],
            "maxLength": 600
        },
        "absolute_tolerance": {
            "type": "number",
            "description": "Maximum absolute difference in the supplied numeric unit. Decimal-dot values only; no unit conversion.",
            "default": 0,
            "minimum": 0,
            "maximum": 1000000
        },
        "relative_tolerance_pct": {
            "type": "number",
            "description": "Tolerance relative to the absolute source value. The larger of absolute and relative tolerance applies; a zero source uses absolute tolerance.",
            "default": 0,
            "minimum": 0,
            "maximum": 100
        },
        "extra_rows": {
            "type": "string",
            "description": "Issue reports additional target records as findings. Allow treats their presence as acceptable within the configured scope.",
            "default": "issue",
            "enum": [
                "issue",
                "allow"
            ],
            "maxLength": 600
        },
        "detail_filter": {
            "type": "string",
            "description": "Issues hides matching detail rows. All includes matching rows. Totals and AI candidate selection remain independent of this display filter.",
            "default": "issues",
            "enum": [
                "issues",
                "all"
            ],
            "maxLength": 600
        }
    }
}

Batch and serial trace review

batch_traceability

Trace supplied batch relationships and movement ledgers, calculate unit-bound balances and identify recipients and missing evidence within an explicit scope.

Input and configuration fields

Essential settings

batches_csv — Batch register CSV

Headers id,product,serial,unit,opening,notes. Up to 100 unique batch IDs. Opening is the starting ledger balance, not an independently verified inventory.

links_csv — Batch relationships CSV

Headers parent,child; up to 200 relationships. Relationships describe trace links and never deduct stock. A header-only file is valid.

movements_csv — Movement ledger CSV

Headers id,batch,kind,quantity,unit,party,reference,date. Up to 300 unique movements. Kinds: receive,consume,ship,adjust_in,adjust_out. Quantities are non-negative decimal-dot values; dates use YYYY-MM-DD.

start_batch — Starting batch

Exactly one starting batch or serial number is required. Batch IDs use letters, digits, underscores or hyphens.

Default settings: ""

Advanced settings

start_serial — Starting serial number

Alternative to the starting batch. The serial must identify exactly one supplied batch.

Default settings: ""

direction — Trace direction

Forward follows parent to child; backward follows child to parent; both explores both directions. Cycles are visited once.

forward — forward · backward — backward · both — both

Default settings: "both"

max_depth — Maximum relationship depth

Between 1 and 20. Relationships beyond the boundary are counted explicitly; their batches are not included.

Default settings: 10

as_of — Movement cutoff date

Optional inclusive cutoff in YYYY-MM-DD. Only movement dates are filtered; undated batch relationships are not historical snapshots.

Default settings: ""

include_zero — Display zero stock

No hides zero-balance stock detail rows. Trace reach, movements and AI note review are unaffected.

yes — yes · no — no

Default settings: "yes"

movement_view — Movement details

All displays every selected movement. Shipments displays shipping movements only; stock displays other movements. Every selected movement still contributes to the stock balance.

all — all · shipments — shipments · stock — stock

Default settings: "all"

minimum_stock — Stock warning threshold

Balances below this threshold are highlighted. Units must match each batch exactly. Incompatible movements are excluded from arithmetic and reported as gaps.

Default settings: 0

detail_order — Detail order

Batch sorts by batch identifier. Issues first places deterministic findings before ordinary details.

batch — batch · issues_first — issues first

Default settings: "batch"

Maximum UTF-8 input bytes: 200000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "batch_traceability",
    "text": "{\"batches_csv\":\"id,product,serial,unit,opening,notes\\nB1,Base material,S1,kg,100,Keep dry\\nB2,Finished product,S2,kg,20,Inspect packaging\",\"links_csv\":\"parent,child\\nB1,B2\",\"movements_csv\":\"id,batch,kind,quantity,unit,party,reference,date\\nM1,B2,ship,5,kg,Warehouse A,DEL-1,2026-11-01\",\"start_batch\":\"B1\",\"start_serial\":\"\",\"direction\":\"forward\",\"max_depth\":10,\"as_of\":\"\",\"include_zero\":\"yes\",\"movement_view\":\"all\",\"minimum_stock\":0,\"detail_order\":\"batch\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "batches_csv",
        "links_csv",
        "movements_csv"
    ],
    "properties": {
        "batches_csv": {
            "type": "string",
            "description": "Headers id,product,serial,unit,opening,notes. Up to 100 unique batch IDs. Opening is the starting ledger balance, not an independently verified inventory.",
            "maxLength": 190000
        },
        "links_csv": {
            "type": "string",
            "description": "Headers parent,child; up to 200 relationships. Relationships describe trace links and never deduct stock. A header-only file is valid.",
            "maxLength": 190000
        },
        "movements_csv": {
            "type": "string",
            "description": "Headers id,batch,kind,quantity,unit,party,reference,date. Up to 300 unique movements. Kinds: receive,consume,ship,adjust_in,adjust_out. Quantities are non-negative decimal-dot values; dates use YYYY-MM-DD.",
            "maxLength": 190000
        },
        "start_batch": {
            "type": "string",
            "description": "Exactly one starting batch or serial number is required. Batch IDs use letters, digits, underscores or hyphens.",
            "default": "",
            "maxLength": 600
        },
        "start_serial": {
            "type": "string",
            "description": "Alternative to the starting batch. The serial must identify exactly one supplied batch.",
            "default": "",
            "maxLength": 600
        },
        "direction": {
            "type": "string",
            "description": "Forward follows parent to child; backward follows child to parent; both explores both directions. Cycles are visited once.",
            "default": "both",
            "enum": [
                "forward",
                "backward",
                "both"
            ],
            "maxLength": 600
        },
        "max_depth": {
            "type": "integer",
            "description": "Between 1 and 20. Relationships beyond the boundary are counted explicitly; their batches are not included.",
            "default": 10,
            "minimum": 1,
            "maximum": 20
        },
        "as_of": {
            "type": "string",
            "description": "Optional inclusive cutoff in YYYY-MM-DD. Only movement dates are filtered; undated batch relationships are not historical snapshots.",
            "default": "",
            "maxLength": 600
        },
        "include_zero": {
            "type": "string",
            "description": "No hides zero-balance stock detail rows. Trace reach, movements and AI note review are unaffected.",
            "default": "yes",
            "enum": [
                "yes",
                "no"
            ],
            "maxLength": 600
        },
        "movement_view": {
            "type": "string",
            "description": "All displays every selected movement. Shipments displays shipping movements only; stock displays other movements. Every selected movement still contributes to the stock balance.",
            "default": "all",
            "enum": [
                "all",
                "shipments",
                "stock"
            ],
            "maxLength": 600
        },
        "minimum_stock": {
            "type": "number",
            "description": "Balances below this threshold are highlighted. Units must match each batch exactly. Incompatible movements are excluded from arithmetic and reported as gaps.",
            "default": 0,
            "minimum": -1000000,
            "maximum": 1000000
        },
        "detail_order": {
            "type": "string",
            "description": "Batch sorts by batch identifier. Issues first places deterministic findings before ordinary details.",
            "default": "batch",
            "enum": [
                "batch",
                "issues_first"
            ],
            "maxLength": 600
        }
    }
}

Source-backed maintenance planning

maintenance_planner

Plan date-based and operating-hour maintenance from explicit source intervals, with overdue review, responsible roles and separately confirmed calendar export.

Input and configuration fields

Essential settings

assets — Maintenance source records

CSV headers: id,asset,instruction,interval_days,interval_hours,last_date,last_hours,current_hours,priority,owner. CSV text is limited to 40,000 UTF-8 bytes. Up to 20 unique uppercase IDs; asset and owner up to 160 UTF-8 bytes, manufacturer instruction up to 1,200 bytes. At least one explicit positive interval is required: days 1–3,650 or hours 1–100,000. Day intervals require a last_date in YYYY-MM-DD, no later than the reference date. Hour intervals require integer last_hours and current_hours from 0–10,000,000, with current_hours not below last_hours. Unused interval and meter cells stay empty. priority is high, medium or low. owner may be empty. Intervals must be transcribed from the actual manufacturer or approved maintenance source; the tool does not verify their authority.

Default settings: ""

reference_date — Reference date

Explicit YYYY-MM-DD date used to compare due dates and action deadlines; supported years 2000–2100. No current date is guessed.

Default settings: "2026-11-03"

Advanced settings

horizon_days — Planning horizon in days

Calendar due dates on or before the reference date plus this many days are due within the planning horizon. The next date is last_date plus the supplied interval; overdue work is not silently rolled forward.

Default settings: 30

priority_filter — Priority filter

Filter the explicitly supplied priority before model selection. Priority is not inferred from the asset name or generated by the model.

all — All · high — High · high_medium — High and medium

Default settings: "all"

due_filter — Due-state filter

All includes later work. Due includes overdue tasks, dates within the horizon and hour thresholds within the configured warning. Overdue requires a date strictly before the reference date or a negative remaining-hour balance. Equality is due, not overdue.

all — All · due — Due within selected bounds · overdue — Overdue

Default settings: "due"

hours_warning — Operating-hour warning threshold

Include hour-based tasks when remaining operating hours are at or below this threshold. This is not an operating-rate estimate and never creates a calendar date from hours.

Default settings: 20

max_items — Maximum selected records

The AI selects this many eligible records, or all eligible records when fewer remain. Selection is a review proposal; excluded or unselected records are counted rather than presented as completed.

Default settings: 12

detail — Excerpt detail

Brief excerpts contain at most 180 UTF-8 bytes; detailed excerpts at most 600. Excerpts remain exact source wording in the original language.

brief — Brief excerpts · detailed — Detailed excerpts

Default settings: "detailed"

sort_order — Display order

Sort complete selected rows by supplied priority, calculated due date with hour-only tasks last, or original source order.

priority — Supplied priority · date — Source date · source — Source order

Default settings: "priority"

calendar_name — Calendar name

Name of the separately confirmed calendar export, at most 120 UTF-8 bytes. Only validated date-based due dates can become events. No calendar service is contacted.

Default settings: "Maintenance review"

reminder_days — Reminder lead time

0 disables reminders. Otherwise the confirmed calendar uses a display reminder 1–90 days before each selected all-day event. Separate human confirmation is mandatory.

Default settings: 7

Maximum UTF-8 input bytes: 50000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "maintenance_planner",
    "text": "{\"assets\":\"id,asset,instruction,interval_days,interval_hours,last_date,last_hours,current_hours,priority,owner\\nA1,Air filter,Inspect the filter and record visible contamination.,30,,2026-10-10,,,high,Maintenance team\\nA2,Workshop compressor,Check the belt condition and record the result.,,200,,1000,1190,medium,Workshop lead\",\"reference_date\":\"2026-11-03\",\"horizon_days\":30,\"priority_filter\":\"all\",\"due_filter\":\"due\",\"hours_warning\":20,\"max_items\":12,\"detail\":\"detailed\",\"sort_order\":\"priority\",\"calendar_name\":\"Reviewed maintenance dates\",\"reminder_days\":7}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "assets",
        "reference_date",
        "horizon_days",
        "priority_filter",
        "due_filter",
        "hours_warning",
        "max_items",
        "detail",
        "sort_order",
        "calendar_name",
        "reminder_days"
    ],
    "properties": {
        "assets": {
            "type": "string",
            "maxLength": 40000,
            "minLength": 1
        },
        "reference_date": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "format": "date"
        },
        "horizon_days": {
            "type": "integer",
            "minimum": 1,
            "maximum": 365
        },
        "priority_filter": {
            "type": "string",
            "enum": [
                "all",
                "high",
                "high_medium"
            ],
            "maxLength": 100,
            "minLength": 1
        },
        "due_filter": {
            "type": "string",
            "enum": [
                "all",
                "due",
                "overdue"
            ],
            "maxLength": 100,
            "minLength": 1
        },
        "hours_warning": {
            "type": "integer",
            "minimum": 0,
            "maximum": 10000
        },
        "max_items": {
            "type": "integer",
            "minimum": 1,
            "maximum": 20
        },
        "detail": {
            "type": "string",
            "enum": [
                "brief",
                "detailed"
            ],
            "maxLength": 100,
            "minLength": 1
        },
        "sort_order": {
            "type": "string",
            "enum": [
                "priority",
                "date",
                "source"
            ],
            "maxLength": 100,
            "minLength": 1
        },
        "calendar_name": {
            "type": "string",
            "maxLength": 120,
            "minLength": 1
        },
        "reminder_days": {
            "type": "integer",
            "minimum": 0,
            "maximum": 90
        }
    }
}

Evidence-linked incident review

incident_review

Organise incident chronology, impact, source-confirmed causes, hypotheses and action follow-up with exact evidence and visible unanswered questions.

Input and configuration fields

Essential settings

incident — Incident title

Short incident name, up to 200 UTF-8 bytes. No secrets, personal identifiers or credentials.

Default settings: ""

evidence — Incident evidence records

CSV headers: id,date,kind,status,impact,owner,deadline,text,criterion. CSV text is limited to 40,000 UTF-8 bytes. Up to 24 unique uppercase IDs; text up to 1,200 UTF-8 bytes, criterion up to 500 and owner up to 160. date and deadline are YYYY-MM-DD or empty. kind is event, impact, cause, hypothesis or action. status is recorded, confirmed, open, in_progress, done, verified or unknown. A cause requires confirmed; a hypothesis requires open or unknown; an action requires open, in_progress, done, verified or unknown. impact is high, medium, low or unknown. Confirmation and verification are claims in the supplied record, not independent findings. Missing owner, deadline and effectiveness criterion stay empty.

Default settings: ""

reference_date — Reference date

Explicit YYYY-MM-DD date used to compare due dates and action deadlines; supported years 2000–2100. No current date is guessed.

Default settings: "2026-11-03"

Advanced settings

focus — Review focus

Select all evidence, event/impact chronology, causes/hypotheses, or actions before model selection. Related event and impact evidence can still be referenced from the full supplied snapshot.

all — All · chronology — Chronology · causes — Causes and hypotheses · actions — Actions

Default settings: "all"

cause_filter — Cause classification filter

Keep both source classifications, only source-confirmed causes, or only hypotheses. The model cannot promote a hypothesis to a confirmed cause. Confirmation is copied from the input and must be independently reviewed.

all — All · confirmed — Confirmed in supplied source · hypotheses — Hypotheses only

Default settings: "all"

action_filter — Action status filter

All retains every supplied action status. Open includes open, in_progress and unknown; verified includes only source-reported verified actions. A completed action is not automatically effective or verified.

all — All · open — Open · verified — Verified in supplied source

Default settings: "all"

impact_filter — Impact filter

Filter the explicit impact level across all record types. Unknown impact is included only with All. The model never invents severity.

all — All · high — High · high_medium — High and medium

Default settings: "all"

max_items — Maximum selected records

The AI selects this many eligible records, or all eligible records when fewer remain. Selection is a review proposal; excluded or unselected records are counted rather than presented as completed.

Default settings: 16

detail — Excerpt detail

Brief excerpts contain at most 180 UTF-8 bytes; detailed excerpts at most 600. Excerpts remain exact source wording in the original language.

brief — Brief excerpts · detailed — Detailed excerpts

Default settings: "detailed"

sort_order — Display order

Sort complete selected rows by source date with undated items last, localized record type, or original source order. Missing dates are never inferred.

chronology — Chronology · type — Record type · source — Source order

Default settings: "chronology"

include_questions — Include evidence-gap questions

Add fixed review questions for missing cause support, responsible roles, action deadlines and effectiveness criteria. Unverified actions remain questions, not automatically completed checks. Overdue open action deadlines are compared with the explicit reference date.

Default settings: true

Maximum UTF-8 input bytes: 50000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "incident_review",
    "text": "{\"incident\":\"Packaging line interruption\",\"evidence\":\"id,date,kind,status,impact,owner,deadline,text,criterion\\nE1,2026-11-02,event,recorded,high,,,The line stopped after the temperature alarm.,\\nE2,2026-11-02,impact,recorded,high,,,Production was paused for 25 minutes.,\\nH1,2026-11-02,hypothesis,open,high,,,A blocked cooling path is suspected but has not been confirmed.,\\nA1,2026-11-03,action,open,high,Maintenance lead,2026-11-05,Inspect the cooling path and record the findings.,No temperature alarm during the documented validation run.\",\"reference_date\":\"2026-11-03\",\"focus\":\"all\",\"cause_filter\":\"all\",\"action_filter\":\"all\",\"impact_filter\":\"all\",\"max_items\":16,\"detail\":\"detailed\",\"sort_order\":\"chronology\",\"include_questions\":true}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "incident",
        "evidence",
        "reference_date",
        "focus",
        "cause_filter",
        "action_filter",
        "impact_filter",
        "max_items",
        "detail",
        "sort_order",
        "include_questions"
    ],
    "properties": {
        "incident": {
            "type": "string",
            "maxLength": 200,
            "minLength": 1
        },
        "evidence": {
            "type": "string",
            "maxLength": 40000,
            "minLength": 1
        },
        "reference_date": {
            "type": "string",
            "maxLength": 100,
            "minLength": 1,
            "format": "date"
        },
        "focus": {
            "type": "string",
            "enum": [
                "all",
                "chronology",
                "causes",
                "actions"
            ],
            "maxLength": 100,
            "minLength": 1
        },
        "cause_filter": {
            "type": "string",
            "enum": [
                "all",
                "confirmed",
                "hypotheses"
            ],
            "maxLength": 100,
            "minLength": 1
        },
        "action_filter": {
            "type": "string",
            "enum": [
                "all",
                "open",
                "verified"
            ],
            "maxLength": 100,
            "minLength": 1
        },
        "impact_filter": {
            "type": "string",
            "enum": [
                "all",
                "high",
                "high_medium"
            ],
            "maxLength": 100,
            "minLength": 1
        },
        "max_items": {
            "type": "integer",
            "minimum": 1,
            "maximum": 24
        },
        "detail": {
            "type": "string",
            "enum": [
                "brief",
                "detailed"
            ],
            "maxLength": 100,
            "minLength": 1
        },
        "sort_order": {
            "type": "string",
            "enum": [
                "chronology",
                "type",
                "source"
            ],
            "maxLength": 100,
            "minLength": 1
        },
        "include_questions": {
            "type": "boolean"
        }
    }
}

Job cost and margin review

job_cost_review

Compare net plan costs with source-backed actual cost assignments, exact line calculations, variances and margin warnings.

Input and configuration fields

Essential settings

plans — Planned positions

CSV columns id,description,planned_net,revenue_net. Up to 15 positions. All amounts are net, in one currency.

actuals — Actual costs

CSV columns id,date,kind,description,quantity,unit_net,plan_id. Up to 30 rows. Labor quantity means hours; material and external quantities mean units. An empty plan_id requests a model proposal.

currency — Currency

A single two-decimal currency applies to all figures.

EUR — EUR · USD — USD · GBP — GBP · PLN — PLN · CZK — CZK · DKK — DKK · SEK — SEK · NOK — NOK · CHF — CHF · HUF — HUF · RON — RON

Default settings: "EUR"

net_basis — Net amounts confirmed

Only net inputs are supported; no tax calculation or gross-to-net conversion is performed.

yes — Yes · no — No

Default settings: "no"

Advanced settings

rounding — Line rounding

Round each quantity multiplied by unit rate to cents before summing. Half-even rounds exact ties to an even cent.

half_up — Round ties upward · half_even — Round ties to even cents

Default settings: "half_up"

group_by — Detail ordering

Order detail rows by assigned plan, cost kind or source ID. Plan totals always remain separate.

plan — Plan position · kind — Cost kind · source — Source ID

Default settings: "plan"

variance_filter — Plan-row visibility

Optionally show only plan rows whose positive cost variance exceeds the threshold. Summary totals and all findings remain complete.

all — All · over_threshold — Above variance threshold

Default settings: "all"

variance_threshold — Cost variance threshold

Nonnegative net amount used for plan overspend findings and the optional plan-row filter.

Default settings: "0.00"

margin_warning_percent — Margin warning percentage

Flag allocated plan margins below this whole percentage, from 0 to 100. Zero revenue produces an unknown margin.

Default settings: 20

date_from — Period start

Optional inclusive date. Only actual costs are filtered; supplied full plan costs and revenues remain unchanged.

Default settings: ""

date_to — Period end

Optional inclusive date. Excluded actual rows remain visible and do not enter selected cost totals.

Default settings: ""

mapping_mode — Assignment policy

Assisted uses a single validated model proposal. Explicit-only allocates only supplied plan_id values; ambiguous and unmatched costs stay unallocated.

assisted — Assisted assignment · explicit_only — Explicit assignments only

Default settings: "assisted"

show_details — Show ordinary detail rows

Unassigned, ambiguous and excluded rows remain visible even when ordinary details are hidden.

yes — Yes · no — No

Default settings: "yes"

include_quotes — Show source quotations

Evidence remains validated even when quotations are hidden from the report.

yes — Yes · no — No

Default settings: "yes"

Maximum UTF-8 input bytes: 50000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "job_cost_review",
    "text": "{\"plans\":\"id,description,planned_net,revenue_net\\nP1,Website launch,600.00,1000.00\",\"actuals\":\"id,date,kind,description,quantity,unit_net,plan_id\\nA1,2027-03-01,labor,Website launch implementation,8.000,80.00,P1\\nA2,2027-03-02,material,Website launch graphics,1.000,100.00,\",\"currency\":\"EUR\",\"net_basis\":\"yes\",\"rounding\":\"half_up\",\"group_by\":\"plan\",\"variance_filter\":\"all\",\"variance_threshold\":\"0.00\",\"margin_warning_percent\":30,\"date_from\":\"\",\"date_to\":\"\",\"mapping_mode\":\"assisted\",\"show_details\":\"yes\",\"include_quotes\":\"yes\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "plans",
        "actuals"
    ],
    "properties": {
        "plans": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 24000
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 15,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "required": [
                            "id",
                            "description",
                            "planned_net",
                            "revenue_net"
                        ],
                        "properties": {
                            "id": {
                                "type": "string",
                                "maxLength": 24,
                                "pattern": "^[A-Z][A-Z0-9_-]{0,23}$"
                            },
                            "description": {
                                "type": "string",
                                "maxLength": 300
                            },
                            "planned_net": {
                                "type": "string",
                                "maxLength": 11,
                                "pattern": "^(0|[1-9][0-9]{0,7})(?:\\.[0-9]{1,2})?$"
                            },
                            "revenue_net": {
                                "type": "string",
                                "maxLength": 11,
                                "pattern": "^(0|[1-9][0-9]{0,7})(?:\\.[0-9]{1,2})?$"
                            }
                        }
                    }
                }
            ]
        },
        "actuals": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 24000
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 30,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "required": [
                            "id",
                            "date",
                            "kind",
                            "description",
                            "quantity",
                            "unit_net",
                            "plan_id"
                        ],
                        "properties": {
                            "id": {
                                "type": "string",
                                "maxLength": 24,
                                "pattern": "^[A-Z][A-Z0-9_-]{0,23}$"
                            },
                            "date": {
                                "type": "string",
                                "maxLength": 10,
                                "pattern": "^(20[0-9]{2}|2100)-[0-9]{2}-[0-9]{2}$"
                            },
                            "kind": {
                                "type": "string",
                                "maxLength": 100,
                                "enum": [
                                    "labor",
                                    "material",
                                    "external"
                                ]
                            },
                            "description": {
                                "type": "string",
                                "maxLength": 300
                            },
                            "quantity": {
                                "type": "string",
                                "maxLength": 12,
                                "pattern": "^(0|[1-9][0-9]{0,7})(?:\\.[0-9]{1,3})?$"
                            },
                            "unit_net": {
                                "type": "string",
                                "maxLength": 11,
                                "pattern": "^(0|[1-9][0-9]{0,7})(?:\\.[0-9]{1,2})?$"
                            },
                            "plan_id": {
                                "type": "string",
                                "maxLength": 24,
                                "pattern": "^(?:|[A-Z][A-Z0-9_-]{0,23})$"
                            }
                        }
                    }
                }
            ]
        },
        "currency": {
            "type": "string",
            "enum": [
                "EUR",
                "USD",
                "GBP",
                "PLN",
                "CZK",
                "DKK",
                "SEK",
                "NOK",
                "CHF",
                "HUF",
                "RON"
            ]
        },
        "net_basis": {
            "type": "string",
            "enum": [
                "yes",
                "no"
            ]
        },
        "rounding": {
            "type": "string",
            "enum": [
                "half_up",
                "half_even"
            ]
        },
        "group_by": {
            "type": "string",
            "enum": [
                "plan",
                "kind",
                "source"
            ]
        },
        "variance_filter": {
            "type": "string",
            "enum": [
                "all",
                "over_threshold"
            ]
        },
        "variance_threshold": {
            "type": "string",
            "pattern": "^(0|[1-9][0-9]{0,7})(?:\\.[0-9]{1,2})?$",
            "maxLength": 11
        },
        "margin_warning_percent": {
            "type": "integer",
            "minimum": 0,
            "maximum": 100
        },
        "date_from": {
            "type": "string",
            "pattern": "^(?:|(20[0-9]{2}|2100)-[0-9]{2}-[0-9]{2})$",
            "maxLength": 10,
            "description": "Optional valid calendar date YYYY-MM-DD, 2000 through 2100."
        },
        "date_to": {
            "type": "string",
            "pattern": "^(?:|(20[0-9]{2}|2100)-[0-9]{2}-[0-9]{2})$",
            "maxLength": 10,
            "description": "Optional valid calendar date YYYY-MM-DD, 2000 through 2100."
        },
        "mapping_mode": {
            "type": "string",
            "enum": [
                "assisted",
                "explicit_only"
            ]
        },
        "show_details": {
            "type": "string",
            "enum": [
                "yes",
                "no"
            ]
        },
        "include_quotes": {
            "type": "string",
            "enum": [
                "yes",
                "no"
            ]
        }
    }
}

SLA event and response-time review

sla_review

Check explicit ticket events against confirmed rules with working-time, pause and daylight-saving calculations.

Input and configuration fields

Essential settings

events — Ticket events

CSV columns id,ticket,at,event,priority,description. Up to 60 events across 12 tickets within 31 elapsed days. Timestamps require seconds and an explicit UTC offset.

rules — Approved SLA rules

CSV columns id,priority,description,response_minutes,resolution_minutes. Up to 12 rules; each threshold is an explicit positive integer in minutes.

rules_approved — Rules explicitly confirmed

The supplied descriptions and numeric thresholds require explicit confirmation before processing.

no — No · yes — Yes

Default settings: "no"

Advanced settings

timezone — Business timezone

Local calendar dates, weekdays, business hours and supplied holidays use this timezone. Source timestamp offsets determine chronological order.

UTC — UTC · Europe/Berlin — Europe/Berlin · Europe/Warsaw — Europe/Warsaw · Europe/London — Europe/London · Europe/Paris — Europe/Paris · Europe/Rome — Europe/Rome · Europe/Madrid — Europe/Madrid · Europe/Lisbon — Europe/Lisbon · Europe/Helsinki — Europe/Helsinki · Europe/Athens — Europe/Athens · Europe/Bucharest — Europe/Bucharest · Europe/Kyiv — Europe/Kyiv · Europe/Dublin — Europe/Dublin · Europe/Vilnius — Europe/Vilnius · Europe/Riga — Europe/Riga · Europe/Tallinn — Europe/Tallinn · America/New_York — America/New_York

Default settings: "Europe/Berlin"

clock — SLA clock

Business time counts only configured local working intervals. Elapsed time counts all seconds. Selected pauses are excluded in both modes.

business — Business time · elapsed — Elapsed time

Default settings: "business"

business_hours — Daily business interval

One same-day HH:MM-HH:MM interval, with start before end and both times before 24:00. Applies in business-time mode.

Default settings: "09:00-17:00"

business_days — Working weekdays

Unique comma-separated weekday numbers, Monday 1 through Sunday 7. Applies in business-time mode.

Default settings: "1,2,3,4,5"

holidays — Excluded local dates

Up to 31 unique YYYY-MM-DD dates, separated by commas or newlines. No automatic holiday calendar is assumed; applies in business-time mode.

Default settings: ""

pause_states — Paused clock states

Comma-separated pause_customer and/or pause_vendor, or empty for no clock exclusions. resume closes an active pause; resolve also closes it.

Default settings: "pause_customer"

cycle_mode — Cycle selection

Review all explicit open/reopen cycles or only the first. Other cycles remain visible as excluded; sequence warnings are retained.

all — All · first — First cycle only

Default settings: "all"

date_from — Period start

Optional inclusive local opening date for cycle selection. Events are retained for sequence validation.

Default settings: ""

date_to — Period end

Optional inclusive local opening date for cycle selection. Excluded cycles remain visible.

Default settings: ""

sort_order — Result ordering

Order by ticket or put findings and incomplete cycles first.

ticket — Ticket · issues_first — Findings first

Default settings: "ticket"

include_quotes — Show source quotations

Evidence remains validated even when quotations are hidden from the report.

yes — Yes · no — No

Default settings: "yes"

Maximum UTF-8 input bytes: 50000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "sla_review",
    "text": "{\"events\":\"id,ticket,at,event,priority,description\\nE1,T1,2027-03-01T09:00:00+01:00,open,high,Hosted service interruption\\nE2,T1,2027-03-01T09:30:00+01:00,response,high,Initial support response\\nE3,T1,2027-03-01T10:00:00+01:00,pause_customer,high,Waiting for customer information\\nE4,T1,2027-03-01T11:00:00+01:00,resume,high,Customer information received\\nE5,T1,2027-03-01T12:00:00+01:00,resolve,high,Service restored\",\"rules\":\"id,priority,description,response_minutes,resolution_minutes\\nR1,high,High-priority hosted service incidents,60,180\",\"rules_approved\":\"yes\",\"timezone\":\"Europe/Berlin\",\"clock\":\"business\",\"business_hours\":\"09:00-17:00\",\"business_days\":\"1,2,3,4,5\",\"holidays\":\"\",\"pause_states\":\"pause_customer\",\"cycle_mode\":\"all\",\"date_from\":\"\",\"date_to\":\"\",\"sort_order\":\"ticket\",\"include_quotes\":\"yes\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "events",
        "rules"
    ],
    "properties": {
        "events": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 24000
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 60,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "required": [
                            "id",
                            "ticket",
                            "at",
                            "event",
                            "priority",
                            "description"
                        ],
                        "properties": {
                            "id": {
                                "type": "string",
                                "maxLength": 24,
                                "pattern": "^[A-Z][A-Z0-9_-]{0,23}$"
                            },
                            "ticket": {
                                "type": "string",
                                "maxLength": 24,
                                "pattern": "^[A-Z][A-Z0-9_-]{0,23}$"
                            },
                            "at": {
                                "type": "string",
                                "maxLength": 25,
                                "pattern": "^(20[0-9]{2}|2100)-[0-9]{2}-[0-9]{2}T[0-9]{2}:[0-9]{2}:[0-9]{2}(?:Z|[+-][0-9]{2}:[0-9]{2})$"
                            },
                            "event": {
                                "type": "string",
                                "maxLength": 100,
                                "enum": [
                                    "open",
                                    "response",
                                    "resolve",
                                    "reopen",
                                    "pause_customer",
                                    "pause_vendor",
                                    "resume"
                                ]
                            },
                            "priority": {
                                "type": "string",
                                "maxLength": 100,
                                "enum": [
                                    "critical",
                                    "high",
                                    "normal",
                                    "low"
                                ]
                            },
                            "description": {
                                "type": "string",
                                "maxLength": 300
                            }
                        }
                    }
                }
            ]
        },
        "rules": {
            "oneOf": [
                {
                    "type": "string",
                    "maxLength": 24000
                },
                {
                    "type": "array",
                    "minItems": 1,
                    "maxItems": 12,
                    "items": {
                        "type": "object",
                        "additionalProperties": false,
                        "required": [
                            "id",
                            "priority",
                            "description",
                            "response_minutes",
                            "resolution_minutes"
                        ],
                        "properties": {
                            "id": {
                                "type": "string",
                                "maxLength": 24,
                                "pattern": "^[A-Z][A-Z0-9_-]{0,23}$"
                            },
                            "priority": {
                                "type": "string",
                                "maxLength": 100,
                                "enum": [
                                    "critical",
                                    "high",
                                    "normal",
                                    "low"
                                ]
                            },
                            "description": {
                                "type": "string",
                                "maxLength": 400
                            },
                            "response_minutes": {
                                "type": "string",
                                "maxLength": 5,
                                "pattern": "^[1-9][0-9]{0,4}$"
                            },
                            "resolution_minutes": {
                                "type": "string",
                                "maxLength": 5,
                                "pattern": "^[1-9][0-9]{0,4}$"
                            }
                        }
                    }
                }
            ]
        },
        "rules_approved": {
            "type": "string",
            "enum": [
                "no",
                "yes"
            ]
        },
        "timezone": {
            "type": "string",
            "enum": [
                "UTC",
                "Europe/Berlin",
                "Europe/Warsaw",
                "Europe/London",
                "Europe/Paris",
                "Europe/Rome",
                "Europe/Madrid",
                "Europe/Lisbon",
                "Europe/Helsinki",
                "Europe/Athens",
                "Europe/Bucharest",
                "Europe/Kyiv",
                "Europe/Dublin",
                "Europe/Vilnius",
                "Europe/Riga",
                "Europe/Tallinn",
                "America/New_York"
            ]
        },
        "clock": {
            "type": "string",
            "enum": [
                "business",
                "elapsed"
            ]
        },
        "business_hours": {
            "type": "string",
            "maxLength": 11,
            "pattern": "^[0-2][0-9]:[0-5][0-9]-[0-2][0-9]:[0-5][0-9]$"
        },
        "business_days": {
            "type": "string",
            "maxLength": 32
        },
        "holidays": {
            "type": "string",
            "maxLength": 500
        },
        "pause_states": {
            "type": "string",
            "maxLength": 100
        },
        "cycle_mode": {
            "type": "string",
            "enum": [
                "all",
                "first"
            ]
        },
        "date_from": {
            "type": "string",
            "pattern": "^(?:|(20[0-9]{2}|2100)-[0-9]{2}-[0-9]{2})$",
            "maxLength": 10,
            "description": "Optional valid calendar date YYYY-MM-DD, 2000 through 2100."
        },
        "date_to": {
            "type": "string",
            "pattern": "^(?:|(20[0-9]{2}|2100)-[0-9]{2}-[0-9]{2})$",
            "maxLength": 10,
            "description": "Optional valid calendar date YYYY-MM-DD, 2000 through 2100."
        },
        "sort_order": {
            "type": "string",
            "enum": [
                "ticket",
                "issues_first"
            ]
        },
        "include_quotes": {
            "type": "string",
            "enum": [
                "yes",
                "no"
            ]
        }
    }
}

One-dimensional cutting plan

cutting_plan

Allocate requested lengths to supplied bars or rolls with explicit kerf, end trims, material groups, remnants and cost preferences, then review every allocation and remaining piece.

Input and configuration fields

Essential settings

demand — Required parts CSV

Headers id,material,unit,length,quantity,note. Up to 20 unique uppercase IDs and 200 total part instances. Positive lengths use a decimal point with at most three decimals; positive integer quantity. Notes are at most 300 UTF-8 bytes. All rows in both files must use one common unit: mm, cm, m or in.

stock — Available stock CSV

Headers id,material,unit,length,quantity,cost,kind,note. Up to 30 unique uppercase IDs and 100 total stock instances. kind is new or remnant; cost is a nonnegative per-item decimal amount with at most two decimals in the selected currency. Existing remnants are explicit stock items; no stock is purchased.

Advanced settings

kerf — Kerf per part

Nonnegative decimal string in the common length unit, at most three decimals. Exactly one kerf is charged per allocated part, including the final part.

Default settings: "0"

trim_start — Leading trim

Nonnegative decimal string deducted once from the start of every used stock item, in the common unit.

Default settings: "0"

trim_end — Trailing trim

Nonnegative decimal string deducted once from the end of every used stock item, in the common unit.

Default settings: "0"

minimum_remnant — Minimum reusable remainder

Nonnegative decimal string. Positive remaining lengths at or above this threshold are reported as reusable; shorter remnants count as waste. This classification does not change placement.

Default settings: "0"

strategy — Placement strategy

Best fit chooses the smallest remaining capacity after the higher-priority stock preferences. First fit omits that remaining-capacity ranking. Both are bounded greedy heuristics and never prove optimality.

best_fit — Best fit · first_fit — First fit

Default settings: "best_fit"

objective — Stock selection objective

Material prefers shorter fitting stock after remnant and opened-stock preferences. Cost ranks unopened stock by the supplied per-item cost before remaining capacity. Already opened stock is preferred. No global cost minimum is claimed.

material — Material usage · cost — Supplied cost

Default settings: "material"

currency — Cost currency

A single label for all supplied stock costs. No exchange-rate conversion or independent price verification is performed.

EUR — EUR · USD — USD · GBP — GBP · PLN — PLN · CHF — CHF

Default settings: "EUR"

prefer_remnants — Prioritize existing remnants

Yes ranks compatible remnant stock before new stock. This preference comes before opened-stock, cost and remaining-capacity ranking.

yes — Yes · no — No

Default settings: "yes"

allow_remnants — Allow existing remnants

No excludes stock rows marked remnant from placement. Their exclusion never creates new stock or changes demand quantities.

yes — Yes · no — No

Default settings: "yes"

material_case — Case-sensitive material groups

Material names are trimmed. No additionally compares Unicode lowercase names; different material groups are never mixed.

yes — Yes · no — No

Default settings: "yes"

cut_order — Part placement order

Longest first orders by length and uses the AI-proposed source order to break ties. AI priority follows the source-bound order proposed from demand notes. Every demand instance remains accounted for.

longest_first — Longest first · ai_priority — AI-proposed priority

Default settings: "longest_first"

display_order — Result order

Stock orders used stock instances by identifier. Waste first orders by combined kerf, trim and remaining length. Display order never changes the computed plan.

stock — Stock identifier · waste_first — Largest nonproduct length first

Default settings: "stock"

Maximum UTF-8 input bytes: 180000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "cutting_plan",
    "text": "{\"demand\":\"id,material,unit,length,quantity,note\\nD1,Aluminium,mm,400,2,Handle edges carefully\\nD2,Aluminium,mm,180,2,Standard workshop order\",\"stock\":\"id,material,unit,length,quantity,cost,kind,note\\nS1,Aluminium,mm,1000,2,12.50,new,Available standard bar\\nS2,Aluminium,mm,400,1,2.00,remnant,Existing remnant\",\"kerf\":\"2\",\"trim_start\":\"5\",\"trim_end\":\"5\",\"minimum_remnant\":\"100\",\"strategy\":\"best_fit\",\"objective\":\"material\",\"currency\":\"EUR\",\"prefer_remnants\":\"yes\",\"allow_remnants\":\"yes\",\"material_case\":\"yes\",\"cut_order\":\"longest_first\",\"display_order\":\"stock\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "demand",
        "stock"
    ],
    "properties": {
        "demand": {
            "type": "string",
            "description": "Headers id,material,unit,length,quantity,note. Up to 20 unique uppercase IDs and 200 total part instances. Positive lengths use a decimal point with at most three decimals; positive integer quantity. Notes are at most 300 UTF-8 bytes. All rows in both files must use one common unit: mm, cm, m or in.",
            "maxLength": 160000
        },
        "stock": {
            "type": "string",
            "description": "Headers id,material,unit,length,quantity,cost,kind,note. Up to 30 unique uppercase IDs and 100 total stock instances. kind is new or remnant; cost is a nonnegative per-item decimal amount with at most two decimals in the selected currency. Existing remnants are explicit stock items; no stock is purchased.",
            "maxLength": 160000
        },
        "kerf": {
            "type": "string",
            "description": "Nonnegative decimal string in the common length unit, at most three decimals. Exactly one kerf is charged per allocated part, including the final part.",
            "default": "0",
            "maxLength": 120,
            "pattern": "^(0|[1-9][0-9]{0,5})(\\.[0-9]{1,3})?$"
        },
        "trim_start": {
            "type": "string",
            "description": "Nonnegative decimal string deducted once from the start of every used stock item, in the common unit.",
            "default": "0",
            "maxLength": 120,
            "pattern": "^(0|[1-9][0-9]{0,5})(\\.[0-9]{1,3})?$"
        },
        "trim_end": {
            "type": "string",
            "description": "Nonnegative decimal string deducted once from the end of every used stock item, in the common unit.",
            "default": "0",
            "maxLength": 120,
            "pattern": "^(0|[1-9][0-9]{0,5})(\\.[0-9]{1,3})?$"
        },
        "minimum_remnant": {
            "type": "string",
            "description": "Nonnegative decimal string. Positive remaining lengths at or above this threshold are reported as reusable; shorter remnants count as waste. This classification does not change placement.",
            "default": "0",
            "maxLength": 120,
            "pattern": "^(0|[1-9][0-9]{0,5})(\\.[0-9]{1,3})?$"
        },
        "strategy": {
            "type": "string",
            "description": "Best fit chooses the smallest remaining capacity after the higher-priority stock preferences. First fit omits that remaining-capacity ranking. Both are bounded greedy heuristics and never prove optimality.",
            "default": "best_fit",
            "enum": [
                "best_fit",
                "first_fit"
            ],
            "maxLength": 120
        },
        "objective": {
            "type": "string",
            "description": "Material prefers shorter fitting stock after remnant and opened-stock preferences. Cost ranks unopened stock by the supplied per-item cost before remaining capacity. Already opened stock is preferred. No global cost minimum is claimed.",
            "default": "material",
            "enum": [
                "material",
                "cost"
            ],
            "maxLength": 120
        },
        "currency": {
            "type": "string",
            "description": "A single label for all supplied stock costs. No exchange-rate conversion or independent price verification is performed.",
            "default": "EUR",
            "enum": [
                "EUR",
                "USD",
                "GBP",
                "PLN",
                "CHF"
            ],
            "maxLength": 120
        },
        "prefer_remnants": {
            "type": "string",
            "description": "Yes ranks compatible remnant stock before new stock. This preference comes before opened-stock, cost and remaining-capacity ranking.",
            "default": "yes",
            "enum": [
                "yes",
                "no"
            ],
            "maxLength": 120
        },
        "allow_remnants": {
            "type": "string",
            "description": "No excludes stock rows marked remnant from placement. Their exclusion never creates new stock or changes demand quantities.",
            "default": "yes",
            "enum": [
                "yes",
                "no"
            ],
            "maxLength": 120
        },
        "material_case": {
            "type": "string",
            "description": "Material names are trimmed. No additionally compares Unicode lowercase names; different material groups are never mixed.",
            "default": "yes",
            "enum": [
                "yes",
                "no"
            ],
            "maxLength": 120
        },
        "cut_order": {
            "type": "string",
            "description": "Longest first orders by length and uses the AI-proposed source order to break ties. AI priority follows the source-bound order proposed from demand notes. Every demand instance remains accounted for.",
            "default": "longest_first",
            "enum": [
                "longest_first",
                "ai_priority"
            ],
            "maxLength": 120
        },
        "display_order": {
            "type": "string",
            "description": "Stock orders used stock instances by identifier. Waste first orders by combined kerf, trim and remaining length. Display order never changes the computed plan.",
            "default": "stock",
            "enum": [
                "stock",
                "waste_first"
            ],
            "maxLength": 120
        }
    }
}

Measurement series review

measurement_review

Compare measured values with explicitly approved limits, summarize feature and batch groups and review missing readings, ambiguous specifications and source-backed AI observations.

Input and configuration fields

Essential settings

measurements — Measurements CSV

Headers id,feature,batch,date,value,unit,note. Up to 500 unique uppercase IDs, 20 feature-and-unit groups and 100 result groups. Dates use YYYY-MM-DD from 2000 through 2100. value is a signed decimal string with up to six decimals, or empty. Notes are at most 300 UTF-8 bytes.

specifications — Approved specification CSV

Headers id,feature,unit,lower,upper,note. Up to 20 unique uppercase IDs. At least one limit must be supplied per specification; an empty lower or upper limit means one-sided checking. Limits are signed decimal strings with at most six decimals. Matching uses the configured feature-name comparison and exact units. Multiple matches remain unresolved.

limits_approved — Confirm supplied limits

Explicit confirmation that these exact supplied specifications and limits are approved for this review. Changing them requires renewed confirmation. Source notes and model output cannot grant approval.

no — Not approved · yes — These exact limits are approved

Default settings: "no"

Advanced settings

date_from — Include from date

Optional inclusive lower date bound in YYYY-MM-DD. Earlier observations are excluded from statistics and counted separately.

Default settings: ""

date_to — Include through date

Optional inclusive upper date bound in YYYY-MM-DD. Later observations are excluded from statistics and counted separately.

Default settings: ""

group_by — Statistical grouping

Feature combines all included batches for each feature and exact unit. Feature and batch keeps batches separate. A maximum of 100 result groups is accepted.

feature — Feature and unit · feature_batch — Feature, unit and batch

Default settings: "feature_batch"

case_sensitive — Case-sensitive feature matching

Names are trimmed. No additionally compares Unicode lowercase names. Units always match exactly; there is no unit conversion or inferred alias mapping.

yes — Yes · no — No

Default settings: "yes"

missing_policy — Missing measurement policy

Report marks groups with empty values for review. Exclude omits empty values from statistics without this additional group warning. Missing-value counts remain visible in both modes; invalid numeric text is rejected.

report — Report missing values · exclude — Exclude from numeric statistics

Default settings: "report"

boundary — Limit boundary interpretation

Inclusive accepts values equal to a supplied bound. Exclusive marks equality as out of range. Comparisons use exact unrounded fixed-point values.

inclusive — Inclusive bounds · exclusive — Exclusive bounds

Default settings: "inclusive"

rounding_digits — Displayed decimal places

Between 0 and 6 decimal places for descriptive statistics. Display rounding never changes specification limits or out-of-range decisions.

Default settings: 3

minimum_sample — Minimum numeric sample size

Groups with fewer numeric observations receive a review status. Between 2 and 500; this threshold does not establish statistical confidence or representativeness.

Default settings: 5

deviation — Standard deviation convention

Sample uses n−1 and requires at least two numeric observations. Population uses n and requires at least one. Values describe only the supplied observations; no wider population inference is implied.

sample — Sample standard deviation · population — Population standard deviation

Default settings: "sample"

sort_order — Group order

Source retains the first appearance of each group. Issues first puts review groups first. Feature sorts by supplied feature name and batch. Statistics remain unchanged.

source — Source order · issues_first — Review groups first · feature — Feature and batch

Default settings: "source"

include_evidence — Show representative quotation

Yes includes the exact first included measurement note for each feature-and-unit group. No hides this display column only; source IDs and model validation remain unchanged.

yes — Include quotation · no — Source IDs only

Default settings: "yes"

detail_filter — Displayed groups

All shows every included group. Issues shows groups with missing or ambiguous specifications, out-of-range values, insufficient numeric sample size or report-mode missing values. Complete summary counts remain unchanged.

all — All groups · issues — Review groups only

Default settings: "all"

Maximum UTF-8 input bytes: 180000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "measurement_review",
    "text": "{\"measurements\":\"id,feature,batch,date,value,unit,note\\nM1,Diameter,B1,2026-11-01,10.010,mm,Instrument zero checked\\nM2,Diameter,B1,2026-11-01,10.020,mm,Same measurement procedure\\nM3,Diameter,B1,2026-11-01,10.060,mm,Repeat measurement recommended\\nM4,Diameter,B1,2026-11-01,,mm,Reading unavailable\",\"specifications\":\"id,feature,unit,lower,upper,note\\nS1,Diameter,mm,9.950,10.050,Approved drawing revision A\",\"limits_approved\":\"yes\",\"date_from\":\"\",\"date_to\":\"\",\"group_by\":\"feature_batch\",\"case_sensitive\":\"yes\",\"missing_policy\":\"report\",\"boundary\":\"inclusive\",\"rounding_digits\":3,\"minimum_sample\":3,\"deviation\":\"sample\",\"sort_order\":\"source\",\"include_evidence\":\"yes\",\"detail_filter\":\"all\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "measurements",
        "specifications",
        "limits_approved"
    ],
    "properties": {
        "measurements": {
            "type": "string",
            "description": "Headers id,feature,batch,date,value,unit,note. Up to 500 unique uppercase IDs, 20 feature-and-unit groups and 100 result groups. Dates use YYYY-MM-DD from 2000 through 2100. value is a signed decimal string with up to six decimals, or empty. Notes are at most 300 UTF-8 bytes.",
            "maxLength": 160000
        },
        "specifications": {
            "type": "string",
            "description": "Headers id,feature,unit,lower,upper,note. Up to 20 unique uppercase IDs. At least one limit must be supplied per specification; an empty lower or upper limit means one-sided checking. Limits are signed decimal strings with at most six decimals. Matching uses the configured feature-name comparison and exact units. Multiple matches remain unresolved.",
            "maxLength": 160000
        },
        "limits_approved": {
            "type": "string",
            "description": "Explicit confirmation that these exact supplied specifications and limits are approved for this review. Changing them requires renewed confirmation. Source notes and model output cannot grant approval.",
            "default": "no",
            "enum": [
                "no",
                "yes"
            ],
            "maxLength": 120
        },
        "date_from": {
            "type": "string",
            "description": "Optional inclusive lower date bound in YYYY-MM-DD. Earlier observations are excluded from statistics and counted separately.",
            "default": "",
            "maxLength": 120
        },
        "date_to": {
            "type": "string",
            "description": "Optional inclusive upper date bound in YYYY-MM-DD. Later observations are excluded from statistics and counted separately.",
            "default": "",
            "maxLength": 120
        },
        "group_by": {
            "type": "string",
            "description": "Feature combines all included batches for each feature and exact unit. Feature and batch keeps batches separate. A maximum of 100 result groups is accepted.",
            "default": "feature_batch",
            "enum": [
                "feature",
                "feature_batch"
            ],
            "maxLength": 120
        },
        "case_sensitive": {
            "type": "string",
            "description": "Names are trimmed. No additionally compares Unicode lowercase names. Units always match exactly; there is no unit conversion or inferred alias mapping.",
            "default": "yes",
            "enum": [
                "yes",
                "no"
            ],
            "maxLength": 120
        },
        "missing_policy": {
            "type": "string",
            "description": "Report marks groups with empty values for review. Exclude omits empty values from statistics without this additional group warning. Missing-value counts remain visible in both modes; invalid numeric text is rejected.",
            "default": "report",
            "enum": [
                "report",
                "exclude"
            ],
            "maxLength": 120
        },
        "boundary": {
            "type": "string",
            "description": "Inclusive accepts values equal to a supplied bound. Exclusive marks equality as out of range. Comparisons use exact unrounded fixed-point values.",
            "default": "inclusive",
            "enum": [
                "inclusive",
                "exclusive"
            ],
            "maxLength": 120
        },
        "rounding_digits": {
            "type": "integer",
            "description": "Between 0 and 6 decimal places for descriptive statistics. Display rounding never changes specification limits or out-of-range decisions.",
            "default": 3,
            "minimum": 0,
            "maximum": 6
        },
        "minimum_sample": {
            "type": "integer",
            "description": "Groups with fewer numeric observations receive a review status. Between 2 and 500; this threshold does not establish statistical confidence or representativeness.",
            "default": 5,
            "minimum": 2,
            "maximum": 500
        },
        "deviation": {
            "type": "string",
            "description": "Sample uses n−1 and requires at least two numeric observations. Population uses n and requires at least one. Values describe only the supplied observations; no wider population inference is implied.",
            "default": "sample",
            "enum": [
                "sample",
                "population"
            ],
            "maxLength": 120
        },
        "sort_order": {
            "type": "string",
            "description": "Source retains the first appearance of each group. Issues first puts review groups first. Feature sorts by supplied feature name and batch. Statistics remain unchanged.",
            "default": "source",
            "enum": [
                "source",
                "issues_first",
                "feature"
            ],
            "maxLength": 120
        },
        "include_evidence": {
            "type": "string",
            "description": "Yes includes the exact first included measurement note for each feature-and-unit group. No hides this display column only; source IDs and model validation remain unchanged.",
            "default": "yes",
            "enum": [
                "yes",
                "no"
            ],
            "maxLength": 120
        },
        "detail_filter": {
            "type": "string",
            "description": "All shows every included group. Issues shows groups with missing or ambiguous specifications, out-of-range values, insufficient numeric sample size or report-mode missing values. Complete summary counts remain unchanged.",
            "default": "all",
            "enum": [
                "all",
                "issues"
            ],
            "maxLength": 120
        }
    }
}

Maintained change-impact review

change_impact

Trace concrete changes through maintained dependencies and review affected documents, tests, processes and responsible roles without inventing impact links.

Input and configuration fields

Essential settings

elements — Maintained elements

CSV headers id,name,kind,criticality,approval,owner,description. Up to 40 unique uppercase IDs. name and owner: at most 160 UTF-8 bytes; description: 1,200 bytes. kind: document,test,process,system. criticality: high,medium,low. approval: approved,draft,unknown. Empty owner is allowed and remains a gap. Approval is a source claim, not an approval granted by this tool. CSV limit: 40,000 UTF-8 bytes.

Default settings: ""

dependencies — Maintained dependencies

CSV headers id,from,to,evidence. from is the prerequisite or upstream element; to is the dependent or downstream element. Up to 100 unique edge IDs and pairs for change review, 50 for recovery. Both endpoints must exist. Evidence is required and limited to 600 UTF-8 bytes. A header-only file is valid; missing external relationships cannot be inferred. CSV limit: 40,000 UTF-8 bytes.

Default settings: ""

changes — Concrete proposed changes

CSV headers id,element,description. Up to 12 unique uppercase change IDs, each linked to an existing element. Description: 1,200 UTF-8 bytes. Each selected starting element must have at least one supplied change. Only changes on selected starting elements enter the review. CSV limit: 40,000 UTF-8 bytes.

Default settings: ""

start_ids — Starting element IDs

One to eight distinct existing element IDs separated by whitespace or commas. Every starting element requires a concrete change record. No model-generated starting ID is accepted.

Default settings: ""

baseline — Baseline version

Required baseline identifier, up to 120 UTF-8 bytes. It labels this supplied snapshot; the tool does not retrieve or compare an external version.

Default settings: "BASE-1"

Advanced settings

direction — Traversal direction

Forward follows prerequisite to dependent; backward follows dependent to prerequisite; both explores both directions. Reverse traversal is a dependency review, not proof of downstream impact. Every displayed path consists only of supplied edge IDs.

forward — Forward · backward — Backward · both — Both directions

Default settings: "forward"

max_depth — Maximum path depth

One to twenty edges from each changed starting element. The shortest discovered path is retained per change and element. Boundary edges are reported; undiscovered elements beyond the limit are not considered unaffected.

Default settings: 10

kind_filter — Element type filter

Filters review candidates after full graph traversal. Hidden intermediate elements still participate in dependency paths.

all — All · document — Document · test — Test · process — Process · system — System

Default settings: "all"

criticality_filter — Criticality filter

Filters review candidates by the explicitly supplied criticality. Recovery scheduling still includes all affected processes before this display selection; low-priority dependencies cannot disappear from the plan.

all — All · high — High · high_medium — High and medium

Default settings: "all"

approval_filter — Source approval filter

Filters the supplied approval field only. approved is a source label, not an independently verified release authorisation.

all — All · approved — Approved in source · draft — Draft in source · unknown — Unknown

Default settings: "all"

max_items — Maximum review selections

The AI selects exactly this many eligible records, or all eligible records when fewer remain. The complete deterministic analysis still covers every supplied dependency and affected process. Unselected records are not certified unaffected or restored.

Default settings: 10

detail — Quotation detail

Brief excerpts are limited to 120 characters, detailed excerpts to 240. The model is asked to prefer passages of at most 80 characters. Each quotation must be an exact contiguous passage in its original source language.

brief — Brief · detailed — Detailed

Default settings: "detailed"

sort_order — Display order

Reorders selected rows by original source order, shortest selected path depth, or supplied criticality. Reachability and selection do not change.

source — Source order · depth — Path depth · criticality — Supplied criticality

Default settings: "depth"

include_gaps — Include ownership and approval gaps

When enabled, missing owners, non-approved source labels and absent reachable document or test records produce explicit findings. Cycle and depth-boundary warnings and the review notice remain visible regardless of this setting.

Default settings: true

Maximum UTF-8 input bytes: 100000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "change_impact",
    "text": "{\"elements\":\"id,name,kind,criticality,approval,owner,description\\nS1,Order service,system,high,approved,Platform team,The service validates incoming order fields.\\nD1,Order handbook,document,medium,draft,Documentation team,The handbook describes required order fields.\\nT1,Order validation test,test,high,approved,Quality team,The test checks missing required order fields.\",\"dependencies\":\"id,from,to,evidence\\nL1,S1,D1,The handbook documents the order service interface.\\nL2,S1,T1,The test exercises the order service validation.\",\"changes\":\"id,element,description\\nC1,S1,Add a mandatory delivery reference to each order.\",\"start_ids\":\"S1\",\"direction\":\"forward\",\"max_depth\":10,\"kind_filter\":\"all\",\"criticality_filter\":\"all\",\"approval_filter\":\"all\",\"baseline\":\"BASE-1\",\"max_items\":10,\"detail\":\"detailed\",\"sort_order\":\"depth\",\"include_gaps\":true}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "elements",
        "dependencies",
        "changes",
        "start_ids",
        "direction",
        "max_depth",
        "kind_filter",
        "criticality_filter",
        "approval_filter",
        "baseline",
        "max_items",
        "detail",
        "sort_order",
        "include_gaps"
    ],
    "properties": {
        "elements": {
            "type": "string",
            "maxLength": 40000
        },
        "dependencies": {
            "type": "string",
            "maxLength": 40000
        },
        "changes": {
            "type": "string",
            "maxLength": 40000
        },
        "start_ids": {
            "type": "string",
            "maxLength": 1200
        },
        "direction": {
            "type": "string",
            "maxLength": 100,
            "enum": [
                "forward",
                "backward",
                "both"
            ]
        },
        "max_depth": {
            "type": "integer",
            "minimum": 1,
            "maximum": 20
        },
        "kind_filter": {
            "type": "string",
            "maxLength": 100,
            "enum": [
                "all",
                "document",
                "test",
                "process",
                "system"
            ]
        },
        "criticality_filter": {
            "type": "string",
            "maxLength": 100,
            "enum": [
                "all",
                "high",
                "high_medium"
            ]
        },
        "approval_filter": {
            "type": "string",
            "maxLength": 100,
            "enum": [
                "all",
                "approved",
                "draft",
                "unknown"
            ]
        },
        "baseline": {
            "type": "string",
            "maxLength": 120
        },
        "max_items": {
            "type": "integer",
            "minimum": 1,
            "maximum": 20
        },
        "detail": {
            "type": "string",
            "maxLength": 100,
            "enum": [
                "brief",
                "detailed"
            ]
        },
        "sort_order": {
            "type": "string",
            "maxLength": 100,
            "enum": [
                "source",
                "depth",
                "criticality"
            ]
        },
        "include_gaps": {
            "type": "boolean"
        }
    }
}

Operational recovery planning

recovery_plan

Build a reviewable recovery sequence from maintained process dependencies and tasks, explicit resources, outage simulations and RTO/RPO targets.

Input and configuration fields

Essential settings

processes — Critical operational and IT processes

CSV headers id,name,criticality,rto_minutes,rpo_minutes,backup_at,owner,description. Up to 20 unique uppercase IDs. name and owner: 160 UTF-8 bytes; description: 1,200 bytes. criticality: high,medium,low. rto_minutes: integer 1–10,080; rpo_minutes: integer 0–10,080. backup_at is YYYY-MM-DDTHH:MM:00Z or empty and cannot be later than reference_time. Missing backup times remain unknown. Years 2000–2100. CSV limit: 40,000 UTF-8 bytes.

Default settings: ""

dependencies — Maintained dependencies

CSV headers id,from,to,evidence. from is the prerequisite or upstream element; to is the dependent or downstream element. Up to 100 unique edge IDs and pairs for change review, 50 for recovery. Both endpoints must exist. Evidence is required and limited to 600 UTF-8 bytes. A header-only file is valid; missing external relationships cannot be inferred. CSV limit: 40,000 UTF-8 bytes.

Default settings: ""

tasks — Maintained recovery tasks

CSV headers id,process,sequence,duration_minutes,resource,units,instruction. Up to 40 unique task IDs. Process and resource IDs must exist. sequence: integer 1–100, unique within each process; duration_minutes: integer 1–10,080; units: integer 1–100. Instructions: 1,200 UTF-8 bytes. A header-only file is valid and produces missing-task blockers. All tasks of each affected process are planned in sequence; none are invented or executed. CSV limit: 40,000 UTF-8 bytes.

Default settings: ""

resources — Recovery resources

CSV headers id,name,capacity. Up to 20 unique uppercase resource IDs. name: 160 UTF-8 bytes; capacity: integer 1–100 simultaneous units. Capacity is supplied, not verified. CSV limit: 40,000 UTF-8 bytes.

Default settings: ""

failed_processes — Simulated failed process IDs

One to twenty distinct existing process IDs separated by whitespace or commas. All downstream dependent processes are included transitively. Unaffected prerequisites are assumed available; missing relationships remain an evidence limitation.

Default settings: ""

reference_time — Explicit UTC reference time

YYYY-MM-DDTHH:MM:00Z, years 2000–2100. This is minute zero of the simulated outage and the reference for backup age. Seconds must be 00. No current timestamp is inferred.

Default settings: "2026-11-03T09:00:00Z"

Advanced settings

unavailable_resources — Simulated unavailable resource IDs

Zero to twenty distinct existing resource IDs separated by whitespace or commas. Listed resources have zero effective capacity. Empty means no additional simulated resource outage.

Default settings: ""

horizon_minutes — Planning horizon in minutes

One to 10,080 minutes from the explicit reference time. A process whose tasks cannot all finish within the horizon is blocked; partial reservations are rolled back and are not presented as a completed recovery.

Default settings: 480

priority_order — Recovery scheduling priority

Ready processes are selected by supplied criticality, shortest RTO, or source order. Each process waits for all affected prerequisites. Resource reservations use a deterministic greedy earliest-fit schedule; the result is not a globally optimal schedule.

criticality — Supplied criticality · rto — Shortest supplied RTO · source — Source order

Default settings: "criticality"

capacity_percent — Available resource capacity (%)

An integer from 1 to 100 scales each supplied resource capacity downward, rounded down to whole units. Simulated unavailable resources remain at zero. Tasks needing more units are blocked; capacity is never silently exceeded.

Default settings: 100

rto_margin_minutes — RTO safety margin in minutes

Zero to 1,440 minutes added to each calculated process finish when comparing it with its RTO target. This comparison margin does not reserve resources or alter the task schedule.

Default settings: 0

rpo_policy — Backup-age policy

Warn keeps the supplied recovery tasks schedulable while flagging missing or too-old backup evidence. Block prevents that process from being scheduled when backup age is unknown or exceeds RPO; downstream dependants are blocked too. Backup timestamps do not prove backup integrity or actual data loss.

warn — Warn about backup gaps · block — Block on backup gaps

Default settings: "warn"

criticality_filter — Criticality filter

Filters review candidates by the explicitly supplied criticality. Recovery scheduling still includes all affected processes before this display selection; low-priority dependencies cannot disappear from the plan.

all — All · high — High · high_medium — High and medium

Default settings: "all"

max_items — Maximum review selections

The AI selects exactly this many eligible records, or all eligible records when fewer remain. The complete deterministic analysis still covers every supplied dependency and affected process. Unselected records are not certified unaffected or restored.

Default settings: 10

detail — Quotation detail

Brief excerpts are limited to 120 characters, detailed excerpts to 240. The model is asked to prefer passages of at most 80 characters. Each quotation must be an exact contiguous passage in its original source language.

brief — Brief · detailed — Detailed

Default settings: "detailed"

sort_order — Display order

Sorts selected process rows by calculated start time with blocked rows last, or original source order. It does not change the recovery schedule.

schedule — Scheduled start · source — Source order

Default settings: "schedule"

Maximum UTF-8 input bytes: 100000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "recovery_plan",
    "text": "{\"processes\":\"id,name,criticality,rto_minutes,rpo_minutes,backup_at,owner,description\\nP1,Order database,high,60,60,2026-11-03T08:30:00Z,Database team,Restore the maintained order database before restarting order intake.\\nP2,Order intake,high,120,60,2026-11-03T08:30:00Z,Service team,Order intake requires the order database to be available.\",\"dependencies\":\"id,from,to,evidence\\nL1,P1,P2,Order intake reads and writes the order database.\",\"tasks\":\"id,process,sequence,duration_minutes,resource,units,instruction\\nT1,P1,1,30,R1,1,Restore the approved database backup and record the validation result.\\nT2,P2,1,20,R1,1,Restart the maintained order intake service after database verification.\",\"resources\":\"id,name,capacity\\nR1,Recovery operator,1\",\"failed_processes\":\"P1\",\"unavailable_resources\":\"\",\"reference_time\":\"2026-11-03T09:00:00Z\",\"horizon_minutes\":480,\"priority_order\":\"criticality\",\"capacity_percent\":100,\"rto_margin_minutes\":0,\"rpo_policy\":\"warn\",\"criticality_filter\":\"all\",\"max_items\":10,\"detail\":\"detailed\",\"sort_order\":\"schedule\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "processes",
        "dependencies",
        "tasks",
        "resources",
        "failed_processes",
        "unavailable_resources",
        "reference_time",
        "horizon_minutes",
        "priority_order",
        "capacity_percent",
        "rto_margin_minutes",
        "rpo_policy",
        "criticality_filter",
        "max_items",
        "detail",
        "sort_order"
    ],
    "properties": {
        "processes": {
            "type": "string",
            "maxLength": 40000
        },
        "dependencies": {
            "type": "string",
            "maxLength": 40000
        },
        "tasks": {
            "type": "string",
            "maxLength": 40000
        },
        "resources": {
            "type": "string",
            "maxLength": 40000
        },
        "failed_processes": {
            "type": "string",
            "maxLength": 1200
        },
        "unavailable_resources": {
            "type": "string",
            "maxLength": 1200,
            "minLength": 0
        },
        "reference_time": {
            "type": "string",
            "maxLength": 20
        },
        "horizon_minutes": {
            "type": "integer",
            "minimum": 1,
            "maximum": 10080
        },
        "priority_order": {
            "type": "string",
            "maxLength": 100,
            "enum": [
                "criticality",
                "rto",
                "source"
            ]
        },
        "capacity_percent": {
            "type": "integer",
            "minimum": 1,
            "maximum": 100
        },
        "rto_margin_minutes": {
            "type": "integer",
            "minimum": 0,
            "maximum": 1440
        },
        "rpo_policy": {
            "type": "string",
            "maxLength": 100,
            "enum": [
                "warn",
                "block"
            ]
        },
        "criticality_filter": {
            "type": "string",
            "maxLength": 100,
            "enum": [
                "all",
                "high",
                "high_medium"
            ]
        },
        "max_items": {
            "type": "integer",
            "minimum": 1,
            "maximum": 20
        },
        "detail": {
            "type": "string",
            "maxLength": 100,
            "enum": [
                "brief",
                "detailed"
            ]
        },
        "sort_order": {
            "type": "string",
            "maxLength": 100,
            "enum": [
                "schedule",
                "source"
            ]
        }
    }
}

Evidence-based assessment rubric

assessment_rubrics

Compare work artifacts with explicit criteria and performance levels, inspect exact evidence quotations and calculate scores from supplied human ratings.

Input and configuration fields

Essential settings

criteria — Rubric criteria CSV

Headers id,name,weight,definition. Up to 8 unique criteria. IDs use uppercase letters, digits, underscores or hyphens and begin with a letter, at most 24 characters. name: 120 UTF-8 bytes; definition: 300 bytes. weight is an integer from 1 to 1000. At most 60 evidence-and-criterion pairs are accepted. Each CSV is limited to 24,000 UTF-8 bytes; exact comma-separated headers and valid CSV quoting are required. Links, encoded URLs, markup and control characters are rejected.

Default settings: ""

levels — Performance levels CSV

Headers id,criterion,points,descriptor. Between 2 and 4 levels per criterion and at most 32 total. Each criterion ID must exist. points is an integer from 0 to 1000, unique within that criterion; the maximum must be positive. descriptor: 300 UTF-8 bytes. IDs are unique across the whole level table. Each CSV is limited to 24,000 UTF-8 bytes; exact comma-separated headers and valid CSV quoting are required. Links, encoded URLs, markup and control characters are rejected.

Default settings: ""

evidence — Artifact evidence CSV

Headers id,name,text. Up to 10 uniquely identified work artifacts, with a name of at most 120 UTF-8 bytes and evidence text of at most 1,200 bytes. This tool supports reviewing supplied work artifacts, not making consequential decisions about people. Each CSV is limited to 24,000 UTF-8 bytes; exact comma-separated headers and valid CSV quoting are required. Links, encoded URLs, markup and control characters are rejected.

Default settings: ""

ratings — Supplied human ratings CSV

Headers id,evidence,criterion,level. Optional rows, up to 80; a header-only file is valid. Every reference must exist and the level must belong to that criterion. At most one supplied human rating per evidence-and-criterion pair. These are input records, not identity-verified approvals. Each CSV is limited to 24,000 UTF-8 bytes; exact comma-separated headers and valid CSV quoting are required. Links, encoded URLs, markup and control characters are rejected.

Default settings: ""

definitions_approved — Confirm supplied definitions

Explicit confirmation applies to these exact criterion definitions and performance levels or this exact codebook. Changing any definition requires renewed confirmation. It does not approve an AI result.

no — Not confirmed · yes — Confirmed for this review

Default settings: "no"

version — Confirmed definition version

A required source version label of at most 120 UTF-8 bytes. The tool does not retrieve or verify an external version.

Default settings: "VERSION-1"

Advanced settings

detail — Quotation detail

Brief quotes are limited to 60 characters, detailed quotes to 120. Every quote must be an exact contiguous source passage. Sources are never rewritten or translated.

brief — Brief · detailed — Detailed

Default settings: "detailed"

sort_order — Result order

Source order follows the first supplied source appearance; identifier order sorts by source ID. Within a source, the original proposed criterion or code order remains stable. Sorting does not change calculations or warnings.

source — Source order · id — Source identifier

Default settings: "source"

weight_mode — Criterion weighting

Supplied weights use each positive integer weight. Equal uses weight 1 for every criterion. Human points are divided by that criterion's maximum supplied level points, then combined using these weights.

supplied — Supplied criterion weights · equal — Equal criterion weights

Default settings: "supplied"

missing_rating — Missing human-rating policy

Incomplete withholds the aggregate score until all criteria have a supplied human rating. Zero explicitly counts a missing human rating as zero while retaining the missing-rating warning. AI suggestions never enter the aggregate.

incomplete — Show incomplete score · zero — Treat missing human ratings as zero

Default settings: "incomplete"

score_scale — Aggregate score scale

Integer from 1 to 1000. The normalized weighted human score is multiplied by this scale. This is a numerical display scale, not a grade or pass threshold.

Default settings: 100

rounding_digits — Score decimal places

From 0 to 4 decimal places. Exact rational arithmetic is rounded half up only for final display; no binary floating-point scores are used.

Default settings: 2

level_order — Displayed performance-level order

Sort each criterion's supplied level descriptors by ascending or descending points. The scoring definition and AI suggestion remain unchanged.

ascending — Ascending points · descending — Descending points

Default settings: "ascending"

Maximum UTF-8 input bytes: 80000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "assessment_rubrics",
    "text": "{\"criteria\":\"id,name,weight,definition\\nC1,Traceability,2,Claims are linked to identifiable source records.\\nC2,Clarity,1,The method and its limitations are stated explicitly.\",\"levels\":\"id,criterion,points,descriptor\\nL1,C1,0,No identifiable supporting source is provided.\\nL2,C1,2,The main claim cites an identifiable source record.\\nL3,C2,0,The method or its limitations are absent.\\nL4,C2,2,The method and at least one limitation are explicit.\",\"evidence\":\"id,name,text\\nE1,Workshop report,The conclusion uses record DATA-7. The method compares two samples and cannot establish causality.\",\"ratings\":\"id,evidence,criterion,level\\nH1,E1,C1,L2\",\"definitions_approved\":\"yes\",\"version\":\"RUBRIC-1\",\"detail\":\"detailed\",\"sort_order\":\"source\",\"weight_mode\":\"supplied\",\"missing_rating\":\"incomplete\",\"score_scale\":100,\"rounding_digits\":2,\"level_order\":\"ascending\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "criteria",
        "levels",
        "evidence",
        "ratings",
        "definitions_approved",
        "version",
        "detail",
        "sort_order",
        "weight_mode",
        "missing_rating",
        "score_scale",
        "rounding_digits",
        "level_order"
    ],
    "properties": {
        "criteria": {
            "type": "string",
            "maxLength": 24000
        },
        "levels": {
            "type": "string",
            "maxLength": 24000
        },
        "evidence": {
            "type": "string",
            "maxLength": 24000
        },
        "ratings": {
            "type": "string",
            "maxLength": 24000
        },
        "definitions_approved": {
            "type": "string",
            "enum": [
                "no",
                "yes"
            ]
        },
        "version": {
            "type": "string",
            "maxLength": 120
        },
        "detail": {
            "type": "string",
            "enum": [
                "brief",
                "detailed"
            ]
        },
        "sort_order": {
            "type": "string",
            "enum": [
                "source",
                "id"
            ]
        },
        "weight_mode": {
            "type": "string",
            "enum": [
                "supplied",
                "equal"
            ]
        },
        "missing_rating": {
            "type": "string",
            "enum": [
                "incomplete",
                "zero"
            ]
        },
        "score_scale": {
            "type": "integer",
            "minimum": 1,
            "maximum": 1000
        },
        "rounding_digits": {
            "type": "integer",
            "minimum": 0,
            "maximum": 4
        },
        "level_order": {
            "type": "string",
            "enum": [
                "ascending",
                "descending"
            ]
        }
    }
}

Literature screening review

literature_screening

Review supplied abstracts against explicit inclusion and exclusion criteria, keep evidence quotations and show reviewer disagreements before selection.

Input and configuration fields

Essential settings

records — Supplied publication records CSV

Headers id,title,year,language,text. Up to 15 unique records; title: 200 UTF-8 bytes; text: 1,200 bytes. year is 1900–2100. language is an exact two- or three-letter lowercase tag with an optional uppercase region. Supply the abstract or a bounded full-text excerpt; no publication is retrieved. Each CSV is limited to 24,000 UTF-8 bytes; exact comma-separated headers and valid CSV quoting are required. Links, encoded URLs, markup and control characters are rejected.

Default settings: ""

criteria — Screening criteria CSV

Headers id,kind,definition. Up to 8 unique criteria. kind is include or exclude; definition: 300 UTF-8 bytes. At most 60 record-and-criterion pairs are accepted. Every criterion is reviewed for every supplied record. Each CSV is limited to 24,000 UTF-8 bytes; exact comma-separated headers and valid CSV quoting are required. Links, encoded URLs, markup and control characters are rejected.

Default settings: ""

reviews — Supplied human reviews CSV

Headers id,record,reviewer,decision,note. Optional rows, up to 30; a header-only file is valid. record must exist; reviewer is an uppercase ID. decision is include, exclude or unclear. note: 300 UTF-8 bytes, optionally empty. At most two distinct reviewers per record and one row per reviewer-and-record pair. Reviewer identities are labels, not verified accounts. Each CSV is limited to 24,000 UTF-8 bytes; exact comma-separated headers and valid CSV quoting are required. Links, encoded URLs, markup and control characters are rejected.

Default settings: ""

definitions_approved — Confirm supplied definitions

Explicit confirmation applies to these exact criterion definitions and performance levels or this exact codebook. Changing any definition requires renewed confirmation. It does not approve an AI result.

no — Not confirmed · yes — Confirmed for this review

Default settings: "no"

version — Confirmed definition version

A required source version label of at most 120 UTF-8 bytes. The tool does not retrieve or verify an external version.

Default settings: "VERSION-1"

Advanced settings

detail — Quotation detail

Brief quotes are limited to 60 characters, detailed quotes to 120. Every quote must be an exact contiguous source passage. Sources are never rewritten or translated.

brief — Brief · detailed — Detailed

Default settings: "detailed"

sort_order — Result order

Source order follows the first supplied source appearance; identifier order sorts by source ID. Within a source, the original proposed criterion or code order remains stable. Sorting does not change calculations or warnings.

source — Source order · id — Source identifier

Default settings: "source"

year_from — First publication year

Inclusive lower bound from 1900 to 2100. Out-of-range records remain in the result with an explicit flag.

Default settings: 2000

year_to — Last publication year

Inclusive upper bound from 1900 to 2100; must not precede the lower bound. No year is inferred from source prose.

Default settings: 2100

languages — Allowed supplied language tags

Optional comma-separated list of up to 12 exact tags such as en,de,pl. Tags use two or three lowercase letters with an optional uppercase region, for example en-GB. Empty disables this metadata check. No translation or language detection is performed.

Default settings: ""

reviewer_policy — Required human reviewers

Require one or two distinct supplied reviewers per record. Missing reviewers, disagreement and unclear decisions are flagged independently of AI criterion suggestions.

one — One human reviewer · two — Two human reviewers

Default settings: "two"

duplicate_policy — Duplicate-title review

Flag compares titles after trimming, whitespace normalization and Unicode lowercase conversion. Ignore disables only this duplicate-title heuristic. No record is removed or merged.

flag — Flag for review · ignore — Skip duplicate-title flags

Default settings: "flag"

screening_stage — Declared supplied evidence scope

Distinguishes supplied title/abstract material from supplied full-text excerpts in the model request and report. Neither setting fetches a publication or establishes that an excerpt is complete.

title_abstract — Supplied title and abstract · supplied_full_text — Supplied full-text excerpt

Default settings: "title_abstract"

Maximum UTF-8 input bytes: 80000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "literature_screening",
    "text": "{\"records\":\"id,title,year,language,text\\nR1,Workshop training comparison,2024,en,The study compares two training groups of adult participants. Allocation was not randomized.\\nR2,Workshop training comparison,2023,en,The report describes a single training group without a comparison group.\",\"criteria\":\"id,kind,definition\\nC1,include,A comparison group is described.\\nC2,exclude,Participants are children.\",\"reviews\":\"id,record,reviewer,decision,note\\nV1,R1,REVIEWER1,include,Comparison group is described.\\nV2,R1,REVIEWER2,unclear,Check whether the comparison meets the protocol.\",\"definitions_approved\":\"yes\",\"version\":\"PROTOCOL-1\",\"detail\":\"detailed\",\"sort_order\":\"source\",\"year_from\":2020,\"year_to\":2026,\"languages\":\"en\",\"reviewer_policy\":\"two\",\"duplicate_policy\":\"flag\",\"screening_stage\":\"title_abstract\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "records",
        "criteria",
        "reviews",
        "definitions_approved",
        "version",
        "detail",
        "sort_order",
        "year_from",
        "year_to",
        "languages",
        "reviewer_policy",
        "duplicate_policy",
        "screening_stage"
    ],
    "properties": {
        "records": {
            "type": "string",
            "maxLength": 24000
        },
        "criteria": {
            "type": "string",
            "maxLength": 24000
        },
        "reviews": {
            "type": "string",
            "maxLength": 24000
        },
        "definitions_approved": {
            "type": "string",
            "enum": [
                "no",
                "yes"
            ]
        },
        "version": {
            "type": "string",
            "maxLength": 120
        },
        "detail": {
            "type": "string",
            "enum": [
                "brief",
                "detailed"
            ]
        },
        "sort_order": {
            "type": "string",
            "enum": [
                "source",
                "id"
            ]
        },
        "year_from": {
            "type": "integer",
            "minimum": 1900,
            "maximum": 2100
        },
        "year_to": {
            "type": "integer",
            "minimum": 1900,
            "maximum": 2100
        },
        "languages": {
            "type": "string",
            "maxLength": 120
        },
        "reviewer_policy": {
            "type": "string",
            "enum": [
                "one",
                "two"
            ]
        },
        "duplicate_policy": {
            "type": "string",
            "enum": [
                "flag",
                "ignore"
            ]
        },
        "screening_stage": {
            "type": "string",
            "enum": [
                "title_abstract",
                "supplied_full_text"
            ]
        }
    }
}

Interview coding with a maintained codebook

interview_coding

Assign supplied codes to interview segments with exact quotations, review uncoded passages and conflicting proposals, and count segments or interviews.

Input and configuration fields

Essential settings

segments — Interview segments CSV

Headers id,interview,speaker,text. Up to 20 unique segments. interview is an uppercase ID; speaker is a supplied label of at most 80 UTF-8 bytes; text: 1,200 bytes. All IDs begin with a letter and use uppercase letters, digits, underscores or hyphens, at most 24 characters. Use pseudonymous labels where appropriate. Each CSV is limited to 24,000 UTF-8 bytes; exact comma-separated headers and valid CSV quoting are required. Links, encoded URLs, markup and control characters are rejected.

Default settings: ""

codebook — Confirmed codebook CSV

Headers id,name,definition,exclusion,group. Up to 12 unique code IDs. name: 120 UTF-8 bytes; definition and exclusion: 300 bytes each. exclusion may be empty. A nonempty uppercase group ID marks mutually exclusive codes. Empty group means no exclusivity group. The codebook defines interpretation boundaries; no new code is invented. Each CSV is limited to 24,000 UTF-8 bytes; exact comma-separated headers and valid CSV quoting are required. Links, encoded URLs, markup and control characters are rejected.

Default settings: ""

definitions_approved — Confirm supplied definitions

Explicit confirmation applies to these exact criterion definitions and performance levels or this exact codebook. Changing any definition requires renewed confirmation. It does not approve an AI result.

no — Not confirmed · yes — Confirmed for this review

Default settings: "no"

version — Confirmed definition version

A required source version label of at most 120 UTF-8 bytes. The tool does not retrieve or verify an external version.

Default settings: "VERSION-1"

Advanced settings

detail — Quotation detail

Brief quotes are limited to 60 characters, detailed quotes to 120. Every quote must be an exact contiguous source passage. Sources are never rewritten or translated.

brief — Brief · detailed — Detailed

Default settings: "detailed"

sort_order — Result order

Source order follows the first supplied source appearance; identifier order sorts by source ID. Within a source, the original proposed criterion or code order remains stable. Sorting does not change calculations or warnings.

source — Source order · id — Source identifier

Default settings: "source"

coding_mode — Coding mode

Single permits at most one proposed code per segment and requires maximum codes to be 1. Multiple allows several distinct supplied codes, bounded by the configured maximum.

single — One code per segment · multiple — Multiple codes per segment

Default settings: "multiple"

max_codes — Maximum codes per segment

Integer from 1 to 3. Every proposed code requires its own exact quotation from that segment. Repeated codes are rejected; empty code arrays remain explicitly uncoded.

Default settings: 3

speaker_filter — Exact speaker filter

Optional exact supplied speaker label, up to 80 UTF-8 bytes. No identity inference is performed. Excluded segment counts remain visible.

Default settings: ""

interview_filter — Exact interview filter

Optional existing interview ID. The two filters are combined; an empty resulting selection is rejected. No excluded segment is sent to the model.

Default settings: ""

minimum_count — Low-frequency review threshold

Integer from 1 to 20 applied to the selected counting unit. Codes below the threshold are flagged, including unused codes; no rows are suppressed.

Default settings: 2

count_mode — Code frequency unit

Segments counts each segment at most once per code. Interviews counts each supplied interview ID at most once per code, even when several segments match. Neither count is a population estimate.

segments — Distinct segments · interviews — Distinct interviews

Default settings: "segments"

uncoded_policy — Uncoded segment handling

Flag adds a review finding for uncoded segments. Report retains the uncoded result rows without that extra finding. It never fabricates a supporting quotation or code.

flag — Flag for review · report — Report uncoded rows without extra flag

Default settings: "flag"

Maximum UTF-8 input bytes: 80000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "interview_coding",
    "text": "{\"segments\":\"id,interview,speaker,text\\nS1,I1,Participant A,The handover checklist helped me locate the missing files.\\nS2,I1,Participant A,I still needed a colleague to explain the approval step.\\nS3,I2,Participant B,The process takes place every Monday.\",\"codebook\":\"id,name,definition,exclusion,group\\nC1,Helpful guidance,An explicit description of guidance that helped complete work.,A neutral mention without reported help.,\\nC2,Unclear approval,An explicit need for clarification of the approval procedure.,A statement that approval was already clear.,APPROVAL\\nC3,Clear approval,An explicit statement that the approval procedure was clear.,A request for help understanding approval.,APPROVAL\",\"definitions_approved\":\"yes\",\"version\":\"CODEBOOK-1\",\"detail\":\"detailed\",\"sort_order\":\"source\",\"coding_mode\":\"multiple\",\"max_codes\":3,\"speaker_filter\":\"\",\"interview_filter\":\"\",\"minimum_count\":2,\"count_mode\":\"segments\",\"uncoded_policy\":\"flag\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "required": [
        "segments",
        "codebook",
        "definitions_approved",
        "version",
        "detail",
        "sort_order",
        "coding_mode",
        "max_codes",
        "speaker_filter",
        "interview_filter",
        "minimum_count",
        "count_mode",
        "uncoded_policy"
    ],
    "properties": {
        "segments": {
            "type": "string",
            "maxLength": 24000
        },
        "codebook": {
            "type": "string",
            "maxLength": 24000
        },
        "definitions_approved": {
            "type": "string",
            "enum": [
                "no",
                "yes"
            ]
        },
        "version": {
            "type": "string",
            "maxLength": 120
        },
        "detail": {
            "type": "string",
            "enum": [
                "brief",
                "detailed"
            ]
        },
        "sort_order": {
            "type": "string",
            "enum": [
                "source",
                "id"
            ]
        },
        "coding_mode": {
            "type": "string",
            "enum": [
                "single",
                "multiple"
            ]
        },
        "max_codes": {
            "type": "integer",
            "minimum": 1,
            "maximum": 3
        },
        "speaker_filter": {
            "type": "string",
            "maxLength": 80
        },
        "interview_filter": {
            "type": "string",
            "maxLength": 24
        },
        "minimum_count": {
            "type": "integer",
            "minimum": 1,
            "maximum": 20
        },
        "count_mode": {
            "type": "string",
            "enum": [
                "segments",
                "interviews"
            ]
        },
        "uncoded_policy": {
            "type": "string",
            "enum": [
                "flag",
                "report"
            ]
        }
    }
}

Recipe quantities and purchasing

recipe_scaling

Scale recipe quantities, apply explicit yields, combine equivalent ingredients and calculate whole packages with an auditable source trail.

Input and configuration fields

Essential settings

ingredients — Recipe ingredients

Up to 30 rows. ingredient_key identifies an explicitly equivalent ingredient; names do not create equivalence. quantity and pack_size use the row unit. yield_percent is the retained percentage after the supplied loss assumption. Empty pack_size leaves purchasing unresolved. Notes are limited to 160 UTF-8 bytes.

base_portions — Base portions

Positive decimal up to 10000; the portion count represented by the original recipe.

Default settings: "4"

target_portions — Target portions

Positive decimal up to 10000; used only in portion-ratio mode.

Default settings: "8"

Advanced settings

scaling_mode — Scaling method

Use target portions divided by base portions, or the explicit factor. Both input fields remain strictly validated.

portions — Portion ratio · factor — Explicit factor

Default settings: "portions"

scale_factor — Explicit scaling factor

Positive decimal up to 100; used only in explicit-factor mode.

Default settings: "2"

apply_yield — Apply supplied yields

Divide the scaled net amount by each row’s supplied yield fraction. Disabled mode treats the supplied quantity as the required amount before reserve.

yes — Enabled · no — Disabled

Default settings: "yes"

reserve_percent — Reserve percentage

Nonnegative decimal from 0 to 100, applied to the calculated material requirement before package rounding.

Default settings: "0"

aggregation — Ingredient aggregation

Combine only equal ingredient_key values with compatible dimensions, or keep each source row separate. Different mass and volume dimensions under one combined key are rejected.

ingredient — Combine matching ingredient keys · source — Source order

Default settings: "ingredient"

unit_display — Displayed quantity units

Use base mass and volume units, or convert them to kg and L. Count quantities remain pcs.

base — g, ml and pcs · large — kg, L and pcs

Default settings: "base"

purchase_packs — Calculate whole packages

Round the exact requirement upward to whole supplied packages. Missing or conflicting package sizes remain visible; no purchase is placed.

yes — Enabled · no — Disabled

Default settings: "yes"

rounding_digits — Displayed decimal places

Display zero to six decimal places. Exact comparisons and whole-pack rounding use unrounded values.

Default settings: 3

rounding_mode — Display rounding

Round halfway values upward, or always round a positive remainder upward. This affects display only.

half_up — Round half upward · up — Round upward

Default settings: "half_up"

sort_order — Result order

Preserve source order or place rows with calculation gaps and AI review concerns first.

source — Source order · issues_first — Review concerns first

Default settings: "source"

include_evidence — Show exact source notes

Include exact source notes in the table and exports, or retain only record identifiers. Source validation always remains active.

yes — Enabled · no — Disabled

Default settings: "yes"

Maximum UTF-8 input bytes: 160000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "recipe_scaling",
    "text": "{\"ingredients\":\"id,ingredient_key,name,quantity,unit,yield_percent,pack_size,note\\nI1,FLOUR,Flour,0.5,kg,100,1,Keep the flour type unchanged.\\nI2,CARROT,Carrots,300,g,80,500,Yield includes the supplied trimming allowance.\\nI3,FLOUR,Flour for dusting,100,g,100,1000,Same flour stock as I1.\",\"base_portions\":\"4\",\"target_portions\":\"10\",\"scaling_mode\":\"portions\",\"scale_factor\":\"2\",\"apply_yield\":\"yes\",\"reserve_percent\":\"10\",\"aggregation\":\"ingredient\",\"unit_display\":\"base\",\"purchase_packs\":\"yes\",\"rounding_digits\":3,\"rounding_mode\":\"half_up\",\"sort_order\":\"source\",\"include_evidence\":\"yes\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "ingredients": {
            "type": "string",
            "maxLength": 100000
        },
        "base_portions": {
            "type": "string",
            "default": "4",
            "maxLength": 32
        },
        "target_portions": {
            "type": "string",
            "default": "8",
            "maxLength": 32
        },
        "scaling_mode": {
            "type": "string",
            "default": "portions",
            "enum": [
                "portions",
                "factor"
            ]
        },
        "scale_factor": {
            "type": "string",
            "default": "2",
            "maxLength": 32
        },
        "apply_yield": {
            "type": "string",
            "default": "yes",
            "enum": [
                "yes",
                "no"
            ]
        },
        "reserve_percent": {
            "type": "string",
            "default": "0",
            "maxLength": 32
        },
        "aggregation": {
            "type": "string",
            "default": "ingredient",
            "enum": [
                "ingredient",
                "source"
            ]
        },
        "unit_display": {
            "type": "string",
            "default": "base",
            "enum": [
                "base",
                "large"
            ]
        },
        "purchase_packs": {
            "type": "string",
            "default": "yes",
            "enum": [
                "yes",
                "no"
            ]
        },
        "rounding_digits": {
            "type": "integer",
            "default": 3,
            "minimum": 0,
            "maximum": 6
        },
        "rounding_mode": {
            "type": "string",
            "default": "half_up",
            "enum": [
                "half_up",
                "up"
            ]
        },
        "sort_order": {
            "type": "string",
            "default": "source",
            "enum": [
                "source",
                "issues_first"
            ]
        },
        "include_evidence": {
            "type": "string",
            "default": "yes",
            "enum": [
                "yes",
                "no"
            ]
        }
    },
    "required": [
        "ingredients"
    ]
}

Room measurements and material quantities

room_quantities

Convert rectangular room measurements into surface quantities and material requirements using supplied coverage, coats, reserve and package sizes.

Input and configuration fields

Essential settings

rooms — Room measurements

Up to 20 rectangular rooms. Each row assigns one supplied product to its target surface. All dimensions use the selected common unit. A separate row is required for a different surface or product.

openings — Opening measurements

Up to 40 openings linked to room_id. Width × height × quantity is deducted only from walls when subtraction is enabled. A header-only CSV represents no openings. Opening geometry and overlap are not inferred.

products — Product coverage definitions

Up to 20 products. target is floor, ceiling or walls; coverage is m² per L or kg per coat. unit is L or kg. Empty pack_size leaves purchasing unresolved. All rates are supplied assumptions, not retrieved specifications.

dimension_unit — Measurement unit

Apply m, cm or mm consistently to every room and opening dimension. Outputs use m and m².

m — m · cm — cm · mm — mm

Default settings: "m"

Advanced settings

opening_policy — Opening deductions

Subtract linked opening rectangles from wall area, or retain gross area. Floor and ceiling calculations do not deduct openings.

subtract — Subtract from walls · ignore — Keep gross wall area

Default settings: "subtract"

reserve_percent — Reserve percentage

Nonnegative decimal from 0 to 100, applied to the calculated material requirement before package rounding.

Default settings: "10"

coats_mode — Coat-count source

Use the supplied coat count of each product, or one explicit global coat count.

product — Aggregate by product · global — Global coat count

Default settings: "product"

coats — Global coat count

Integer from 1 to 10, used only when the global coat-count source is selected.

Default settings: 2

purchase_packs — Calculate whole packages

Round the exact requirement upward to whole supplied packages. Missing or conflicting package sizes remain visible; no purchase is placed.

yes — Enabled · no — Disabled

Default settings: "yes"

pack_scope — Package rounding scope

Aggregate requirements for each product before whole-package rounding, or round each room separately. No stock is deducted automatically.

product — Aggregate by product · room — Round each room separately

Default settings: "product"

rounding_digits — Displayed decimal places

Display zero to six decimal places. Exact comparisons and whole-pack rounding use unrounded values.

Default settings: 3

rounding_mode — Display rounding

Round halfway values upward, or always round a positive remainder upward. This affects display only.

half_up — Round half upward · up — Round upward

Default settings: "half_up"

sort_order — Result order

Preserve source order or place rows with calculation gaps and AI review concerns first.

source — Source order · issues_first — Review concerns first

Default settings: "source"

include_evidence — Show exact source notes

Include exact source notes in the table and exports, or retain only record identifiers. Source validation always remains active.

yes — Enabled · no — Disabled

Default settings: "yes"

Maximum UTF-8 input bytes: 160000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "room_quantities",
    "text": "{\"rooms\":\"id,name,length,width,height,product_id,note\\nR1,Workshop,5,4,2.5,P1,Verify the wall measurements before ordering.\\nR2,Store room,3,2,2.5,P1,Use the same supplied wall coating.\",\"openings\":\"id,room_id,width,height,quantity,note\\nO1,R1,1,2,1,Door opening.\\nO2,R1,1.5,1,2,Two windows.\",\"products\":\"id,name,target,coverage,unit,pack_size,coats,note\\nP1,Wall coating,walls,10,L,5,2,Coverage is a supplied planning value per coat.\",\"dimension_unit\":\"m\",\"opening_policy\":\"subtract\",\"reserve_percent\":\"10\",\"coats_mode\":\"product\",\"coats\":2,\"purchase_packs\":\"yes\",\"pack_scope\":\"product\",\"rounding_digits\":3,\"rounding_mode\":\"half_up\",\"sort_order\":\"source\",\"include_evidence\":\"yes\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "rooms": {
            "type": "string",
            "maxLength": 100000
        },
        "openings": {
            "type": "string",
            "maxLength": 100000
        },
        "products": {
            "type": "string",
            "maxLength": 100000
        },
        "dimension_unit": {
            "type": "string",
            "default": "m",
            "enum": [
                "m",
                "cm",
                "mm"
            ]
        },
        "opening_policy": {
            "type": "string",
            "default": "subtract",
            "enum": [
                "subtract",
                "ignore"
            ]
        },
        "reserve_percent": {
            "type": "string",
            "default": "10",
            "maxLength": 32
        },
        "coats_mode": {
            "type": "string",
            "default": "product",
            "enum": [
                "product",
                "global"
            ]
        },
        "coats": {
            "type": "integer",
            "default": 2,
            "minimum": 1,
            "maximum": 10
        },
        "purchase_packs": {
            "type": "string",
            "default": "yes",
            "enum": [
                "yes",
                "no"
            ]
        },
        "pack_scope": {
            "type": "string",
            "default": "product",
            "enum": [
                "product",
                "room"
            ]
        },
        "rounding_digits": {
            "type": "integer",
            "default": 3,
            "minimum": 0,
            "maximum": 6
        },
        "rounding_mode": {
            "type": "string",
            "default": "half_up",
            "enum": [
                "half_up",
                "up"
            ]
        },
        "sort_order": {
            "type": "string",
            "default": "source",
            "enum": [
                "source",
                "issues_first"
            ]
        },
        "include_evidence": {
            "type": "string",
            "default": "yes",
            "enum": [
                "yes",
                "no"
            ]
        }
    },
    "required": [
        "rooms",
        "openings",
        "products"
    ]
}

Crop spacing and rotation review

crop_planning

Calculate rectangular planting capacities and review occupancy conflicts, family intervals and missing history without inventing growing rules.

Input and configuration fields

Essential settings

beds — Measured beds

Up to 20 rectangular beds. Every row has its own m, cm or mm unit. A planting occupies the whole referenced bed for conflict checks.

plantings — Planned crops

Up to 30 plantings with inclusive start and end dates. family is an explicit uppercase key, not a botanical classification inferred from the crop name. Spacing uses the selected common spacing unit.

history — Recorded previous crops

Up to 40 historical records with inclusive dates. A header-only CSV is allowed. Records are checked for overlap and same-family intervals; missing external history cannot be reconstructed.

rules — Supplied family intervals

Up to 20 unique family rules. break_days is an integer from 0 to 3650, counting complete clear days. A header-only CSV is allowed; missing rules remain visible in explicit-rule mode.

Advanced settings

spacing_unit — Spacing and edge-margin unit

Apply m, cm or mm to both plant and row spacing and the common edge margin. Bed dimensions retain their own explicitly supplied units.

m — m · cm — cm · mm — mm

Default settings: "cm"

edge_margin — Edge margin on each side

Nonnegative decimal up to 10000 in the spacing unit. Twice this margin is removed from both bed dimensions before counting positions.

Default settings: "0"

path_percent — Planning allowance for paths

Nonnegative decimal from 0 to 100. Reduce the rectangular grid count by this percentage and round down. This is an aggregate allowance, not a geometric path layout.

Default settings: "0"

rotation_mode — Rotation-rule source

Use explicit rules for each family, or the supplied global interval. No agronomic interval is inferred automatically.

rules — Explicit family rules · default — Global supplied interval

Default settings: "rules"

default_break_days — Global clear-day interval

Integer from 0 to 3650, used only in global-interval mode. The interval applies to the same supplied family key.

Default settings: 365

overlap_policy — Counts during occupancy conflicts

Retain draft capacities with explicit overlap warnings, or set the affected planting counts to zero. Overlaps with supplied history also count.

flag — Flag conflicts and retain draft counts · block_count — Suppress counts for overlapping plantings

Default settings: "flag"

history_status — Historical coverage declaration

Declare whether the supplied historical records are complete for the intended review. This declaration does not verify external records.

complete — Declared complete · incomplete — Declared incomplete

Default settings: "incomplete"

date_from — Display interval start

Optional YYYY-MM-DD date from 2000 through 2100. Include plantings that overlap the display interval; all supplied records still participate in conflict checks.

Default settings: ""

date_to — Display interval end

Optional inclusive YYYY-MM-DD end date from 2000 through 2100. Empty means no upper display bound.

Default settings: ""

rounding_digits — Displayed decimal places

Display zero to six decimal places. Exact comparisons and whole-pack rounding use unrounded values.

Default settings: 3

sort_order — Result order

Preserve source order or place rows with calculation gaps and AI review concerns first.

source — Source order · issues_first — Review concerns first

Default settings: "source"

include_evidence — Show exact source notes

Include exact source notes in the table and exports, or retain only record identifiers. Source validation always remains active.

yes — Enabled · no — Disabled

Default settings: "yes"

Maximum UTF-8 input bytes: 160000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "crop_planning",
    "text": "{\"beds\":\"id,name,length,width,unit,note\\nB1,North bed,4,1.2,m,Rectangular measured bed.\\nB2,South bed,3,1,m,Separate planting area.\",\"plantings\":\"id,bed_id,crop,family,start,end,row_spacing,plant_spacing,note\\nC1,B1,Carrots,APIACEAE,2027-03-01,2027-06-30,30,10,Check the supplied row spacing.\\nC2,B1,Beetroot,AMARANTHACEAE,2027-06-15,2027-09-30,30,15,The planned dates overlap the earlier crop.\\nC3,B2,Peas,FABACEAE,2027-03-01,2027-06-01,40,10,Review the available history.\",\"history\":\"id,bed_id,family,start,end,note\\nH1,B1,APIACEAE,2026-03-01,2026-08-31,Recorded previous crop.\",\"rules\":\"id,family,break_days,note\\nT1,APIACEAE,365,Explicit supplied interval.\\nT2,AMARANTHACEAE,180,Explicit supplied interval.\",\"spacing_unit\":\"cm\",\"edge_margin\":\"0\",\"path_percent\":\"10\",\"rotation_mode\":\"rules\",\"default_break_days\":365,\"overlap_policy\":\"flag\",\"history_status\":\"incomplete\",\"date_from\":\"\",\"date_to\":\"\",\"rounding_digits\":3,\"sort_order\":\"source\",\"include_evidence\":\"yes\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "beds": {
            "type": "string",
            "maxLength": 100000
        },
        "plantings": {
            "type": "string",
            "maxLength": 100000
        },
        "history": {
            "type": "string",
            "maxLength": 100000
        },
        "rules": {
            "type": "string",
            "maxLength": 100000
        },
        "spacing_unit": {
            "type": "string",
            "default": "cm",
            "enum": [
                "m",
                "cm",
                "mm"
            ]
        },
        "edge_margin": {
            "type": "string",
            "default": "0",
            "maxLength": 32
        },
        "path_percent": {
            "type": "string",
            "default": "0",
            "maxLength": 32
        },
        "rotation_mode": {
            "type": "string",
            "default": "rules",
            "enum": [
                "rules",
                "default"
            ]
        },
        "default_break_days": {
            "type": "integer",
            "default": 365,
            "minimum": 0,
            "maximum": 3650
        },
        "overlap_policy": {
            "type": "string",
            "default": "flag",
            "enum": [
                "flag",
                "block_count"
            ]
        },
        "history_status": {
            "type": "string",
            "default": "incomplete",
            "enum": [
                "complete",
                "incomplete"
            ]
        },
        "date_from": {
            "type": "string",
            "default": "",
            "maxLength": 32
        },
        "date_to": {
            "type": "string",
            "default": "",
            "maxLength": 32
        },
        "rounding_digits": {
            "type": "integer",
            "default": 3,
            "minimum": 0,
            "maximum": 6
        },
        "sort_order": {
            "type": "string",
            "default": "source",
            "enum": [
                "source",
                "issues_first"
            ]
        },
        "include_evidence": {
            "type": "string",
            "default": "yes",
            "enum": [
                "yes",
                "no"
            ]
        }
    },
    "required": [
        "beds",
        "plantings",
        "history",
        "rules"
    ]
}

Script continuity review

script_continuity

Compare declared scene states against maintained continuity rules and inspect exact narrative evidence with a separate AI support assessment.

Input and configuration fields

Essential settings

scenes — Scenes

CSV columns id,day,order,text. Up to 40 scenes; day and order are integers 1–999, and each day/order pair must be unique. Scene text is limited to 1200 UTF-8 bytes.

Default settings: ""

states — Declared continuity states

CSV columns id,scene,entity,kind,state,quote. Up to 20 state records. kind is role, prop or costume. Each quote must be an exact passage of its referenced scene, at most 240 UTF-8 bytes. One record per scene, entity and kind. State equality uses exact text, including spelling and case.

Default settings: ""

rules — Maintained continuity rules

CSV columns id,entity,kind,scope,policy,description. Up to 20 rules. scope is same_day or whole_story; policy is constant or track. Entity and kind must match exactly. Descriptions are limited to 240 UTF-8 bytes.

Default settings: ""

Advanced settings

day_from — First story day

Inclusive lower story-day boundary. Only selected state records are reviewed.

Default settings: 1

day_to — Last story day

Inclusive upper story-day boundary, not earlier than the first story day.

Default settings: 999

kind_filter — Continuity category

Limit state evidence to one maintained category.

all — All · role — Role · prop — Prop · costume — Costume

Default settings: "all"

entity_filter — Exact entity filter

An optional exact entity name. Empty includes all entities; spelling and case must match maintained records.

Default settings: ""

rule_filter — Rule policy

Constant flags declared state changes; track documents changes without treating them as conflicts. Filtering changes the set of applicable rules.

all — All · constant — Constant state required · track — Track changes

Default settings: "all"

sort_order — Evidence order

Order state evidence chronologically or by exact entity name. Rule comparisons retain maintained rule and scene order.

chronology — Story chronology · entity — Entity

Default settings: "chronology"

show_consistent — Show unchanged comparisons

Include comparisons where the declared state remains identical. Hidden unchanged comparisons are counted explicitly; source evidence always remains visible.

Default settings: true

include_gaps — Identify missing rules

Add a finding for selected state records without an applicable maintained rule.

Default settings: true

detail — AI evidence length

Brief excerpts are limited to 120 characters, detailed excerpts to 240. Exact source language is retained. Very large inputs can still be rejected.

brief — Brief evidence · detailed — Detailed evidence

Default settings: "brief"

baseline — Script baseline

A supplied version or revision label displayed with every evidence and comparison row.

Default settings: "V1"

Maximum UTF-8 input bytes: 90000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "script_continuity",
    "text": "{\"scenes\":\"id,day,order,text\\nS1,1,1,Mara wears a red coat and carries a brass key.\\nS2,1,2,Mara wears a blue coat and carries a brass key.\",\"states\":\"id,scene,entity,kind,state,quote\\nE1,S1,Mara,costume,red coat,Mara wears a red coat\\nE2,S2,Mara,costume,blue coat,Mara wears a blue coat\",\"rules\":\"id,entity,kind,scope,policy,description\\nR1,Mara,costume,same_day,constant,The coat remains unchanged during the same story day.\",\"day_from\":1,\"day_to\":999,\"kind_filter\":\"all\",\"entity_filter\":\"\",\"rule_filter\":\"all\",\"sort_order\":\"chronology\",\"show_consistent\":true,\"include_gaps\":true,\"detail\":\"brief\",\"baseline\":\"V1\"}"
}

{
    "type": "object",
    "properties": {
        "scenes": {
            "type": "string",
            "minLength": 1,
            "maxLength": 40000
        },
        "states": {
            "type": "string",
            "minLength": 1,
            "maxLength": 40000
        },
        "rules": {
            "type": "string",
            "minLength": 1,
            "maxLength": 40000
        },
        "day_from": {
            "type": "integer",
            "minimum": 1,
            "maximum": 999
        },
        "day_to": {
            "type": "integer",
            "minimum": 1,
            "maximum": 999
        },
        "kind_filter": {
            "type": "string",
            "enum": [
                "all",
                "role",
                "prop",
                "costume"
            ],
            "minLength": 1,
            "maxLength": 40000
        },
        "entity_filter": {
            "type": "string",
            "minLength": 0,
            "maxLength": 120
        },
        "rule_filter": {
            "type": "string",
            "enum": [
                "all",
                "constant",
                "track"
            ],
            "minLength": 1,
            "maxLength": 40000
        },
        "sort_order": {
            "type": "string",
            "enum": [
                "chronology",
                "entity"
            ],
            "minLength": 1,
            "maxLength": 40000
        },
        "show_consistent": {
            "type": "boolean"
        },
        "include_gaps": {
            "type": "boolean"
        },
        "detail": {
            "type": "string",
            "enum": [
                "brief",
                "detailed"
            ],
            "minLength": 1,
            "maxLength": 40000
        },
        "baseline": {
            "type": "string",
            "minLength": 1,
            "maxLength": 120
        }
    },
    "required": [
        "scenes",
        "states",
        "rules",
        "day_from",
        "day_to",
        "kind_filter",
        "entity_filter",
        "rule_filter",
        "sort_order",
        "show_consistent",
        "include_gaps",
        "detail",
        "baseline"
    ],
    "additionalProperties": false
}

Reproducible experiment design

experiment_design

Build a complete, bounded non-clinical run matrix from maintained factors and levels, with seeded ordering and source-bound AI factor review.

Input and configuration fields

Essential settings

factors — Factors and purpose

CSV columns id,name,purpose. Between 1 and 6 factors. Names:120 UTF-8 bytes; purpose:600. The AI classifies only the supplied purpose and cites an exact passage.

Default settings: ""

levels — Maintained factor levels

CSV columns id,factor,value. Each factor needs 2–6 distinct values and valid level IDs. Up to 36 levels; each value is at most 120 UTF-8 bytes. Numeric-looking values remain literal labels and are never converted to measurements.

Default settings: ""

combinations — Explicit allowed combinations

Used only for allowed_combinations mode. CSV starts with id, followed by every factor ID in the selected factor order. Each cell contains a level ID belonging to that factor. Duplicate combinations are rejected. Leave empty for full_factorial.

Default settings: ""

design_mode — Combination mode

Full factorial enumerates every maintained level combination. Explicit mode includes only the supplied allowed combinations. Neither mode infers safety or feasibility.

full_factorial — Full factorial · allowed_combinations — Explicit allowed combinations

Default settings: "full_factorial"

objective — Maintained objective

A supplied non-clinical objective used as context for source-bound factor review and displayed unchanged in the result. It cannot authorize new levels or experiments.

Default settings: ""

Advanced settings

replicates — Replicates per combination

Repeat every included combination equally. Replication does not establish statistical power or independence.

Default settings: 2

block_size — Runs per block

Consecutive runs are grouped into blocks of this size; the final block may be shorter. These are scheduling groups, not inferred scientific blocking factors.

Default settings: 4

randomization — Run ordering

None preserves generated order. Within blocks shuffles only inside the original blocks. Global shuffles all runs before block assignment. A versioned SHA256 ranking and seed produce reproducible ordering.

none — No randomization · within_blocks — Randomize within blocks · global — Randomize globally

Default settings: "within_blocks"

seed — Reproducibility seed

Integer 0–2147483647. The same complete inputs and ordering algorithm produce the same matrix. This seed is not a security mechanism.

Default settings: 42

run_limit — Maximum complete run count

Reject the whole design if combinations multiplied by replicates exceed this limit. No random subset or silent truncation is produced.

Default settings: 180

factor_order — Factor order

Preserve CSV factor order or sort by factor ID. Explicit combination headers must follow this selected order.

source — Source order · id — ID order

Default settings: "source"

level_order — Level order

Preserve level CSV order within each factor or sort by level ID. This changes full-factorial enumeration and may change block membership.

source — Source order · id — ID order

Default settings: "source"

replicate_order — Replication order

Grouped places replicates of each combination together. Interleaved traverses all combinations once per replicate before randomization.

grouped — Group by combination · interleaved — Interleave replicates

Default settings: "interleaved"

domain — Non-clinical work area

The supported scope is non-clinical process, material, software or educational work. The tool does not establish safety, approvals or suitability for human experimentation.

manufacturing — Manufacturing · materials — Materials · software — Software · education — Education

Default settings: "manufacturing"

design_name — Design reference

A supplied reference displayed with all runs and factor reviews.

Default settings: "DESIGN1"

detail — AI evidence length

Brief excerpts are limited to 120 characters, detailed excerpts to 240. Exact source language is retained. Very large inputs can still be rejected.

brief — Brief evidence · detailed — Detailed evidence

Default settings: "brief"

Maximum UTF-8 input bytes: 90000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "experiment_design",
    "text": "{\"factors\":\"id,name,purpose\\nF1,Print speed,Compare print speed settings for visible edge quality.\\nF2,Cooling setting,Compare controlled cooling settings while reviewing print quality.\",\"levels\":\"id,factor,value\\nL1,F1,Slow\\nL2,F1,Fast\\nL3,F2,Low\\nL4,F2,High\",\"combinations\":\"\",\"design_mode\":\"full_factorial\",\"replicates\":2,\"block_size\":4,\"randomization\":\"within_blocks\",\"seed\":42,\"run_limit\":180,\"factor_order\":\"source\",\"level_order\":\"source\",\"replicate_order\":\"interleaved\",\"domain\":\"manufacturing\",\"objective\":\"Compare maintained printer settings on non-clinical sample parts.\",\"design_name\":\"DESIGN1\",\"detail\":\"brief\"}"
}

{
    "type": "object",
    "properties": {
        "factors": {
            "type": "string",
            "minLength": 1,
            "maxLength": 40000
        },
        "levels": {
            "type": "string",
            "minLength": 1,
            "maxLength": 40000
        },
        "combinations": {
            "type": "string",
            "minLength": 0,
            "maxLength": 40000
        },
        "design_mode": {
            "type": "string",
            "enum": [
                "full_factorial",
                "allowed_combinations"
            ],
            "minLength": 1,
            "maxLength": 40000
        },
        "replicates": {
            "type": "integer",
            "minimum": 1,
            "maximum": 10
        },
        "block_size": {
            "type": "integer",
            "minimum": 1,
            "maximum": 180
        },
        "randomization": {
            "type": "string",
            "enum": [
                "none",
                "within_blocks",
                "global"
            ],
            "minLength": 1,
            "maxLength": 40000
        },
        "seed": {
            "type": "integer",
            "minimum": 0,
            "maximum": 2147483647
        },
        "run_limit": {
            "type": "integer",
            "minimum": 1,
            "maximum": 180
        },
        "factor_order": {
            "type": "string",
            "enum": [
                "source",
                "id"
            ],
            "minLength": 1,
            "maxLength": 40000
        },
        "level_order": {
            "type": "string",
            "enum": [
                "source",
                "id"
            ],
            "minLength": 1,
            "maxLength": 40000
        },
        "replicate_order": {
            "type": "string",
            "enum": [
                "grouped",
                "interleaved"
            ],
            "minLength": 1,
            "maxLength": 40000
        },
        "domain": {
            "type": "string",
            "enum": [
                "manufacturing",
                "materials",
                "software",
                "education"
            ],
            "minLength": 1,
            "maxLength": 40000
        },
        "objective": {
            "type": "string",
            "minLength": 1,
            "maxLength": 600
        },
        "design_name": {
            "type": "string",
            "minLength": 1,
            "maxLength": 120
        },
        "detail": {
            "type": "string",
            "enum": [
                "brief",
                "detailed"
            ],
            "minLength": 1,
            "maxLength": 40000
        }
    },
    "required": [
        "factors",
        "levels",
        "combinations",
        "design_mode",
        "replicates",
        "block_size",
        "randomization",
        "seed",
        "run_limit",
        "factor_order",
        "level_order",
        "replicate_order",
        "domain",
        "objective",
        "design_name",
        "detail"
    ],
    "additionalProperties": false
}

Press release

press_release

Write a press release from supplied facts, with approved quotes kept word for word and no invented figures.

Input and configuration fields

Essential settings

facts — Facts

What happened, who is involved, why it matters. Plain notes are enough.

Default settings: ""

organisation — Organisation

Name of the issuing organisation.

Default settings: ""

location — Location

Place of the dateline.

Default settings: ""

release_date — Date

Date as it should appear.

Default settings: ""

approved_quotes — Approved quotes

Only quotes that were approved. Added word for word; the AI writes no quotes.

Default settings: ""

Advanced settings

boilerplate — About the organisation

Standard paragraph, added word for word.

Default settings: ""

contact — Contact details

Added word for word at the end.

Default settings: ""

length — Length

Approximate length of the draft.

short — Short · medium — Medium · long — Long

Default settings: "medium"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "press_release",
    "text": "{\"facts\":\"Riverside Bakery opens a second shop on Mill Street on 12 May. The shop creates 6 jobs. Bread is baked on site with flour from Hartmann Mill.\",\"organisation\":\"Riverside Bakery\",\"location\":\"Bridgeport\",\"release_date\":\"2 May\",\"approved_quotes\":\"\\\"We wanted to bake where our customers live.\\\" Anna Berg, owner\",\"boilerplate\":\"Riverside Bakery has baked bread in Bridgeport since 1994.\",\"contact\":\"Anna Berg, press@riverside-bakery.example\",\"length\":\"short\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "facts": {
            "type": "string",
            "maxLength": 8000,
            "minLength": 1
        },
        "organisation": {
            "type": "string",
            "maxLength": 120,
            "minLength": 1
        },
        "location": {
            "type": "string",
            "maxLength": 80
        },
        "release_date": {
            "type": "string",
            "maxLength": 40
        },
        "approved_quotes": {
            "type": "string",
            "maxLength": 1800
        },
        "boilerplate": {
            "type": "string",
            "maxLength": 1600
        },
        "contact": {
            "type": "string",
            "maxLength": 600
        },
        "length": {
            "type": "string",
            "enum": [
                "short",
                "medium",
                "long"
            ]
        }
    },
    "required": [
        "facts",
        "organisation",
        "length"
    ]
}

Job advertisement

job_ad_create

Write a job advertisement from role facts, with salary and application details kept word for word and exclusionary wording reported.

Input and configuration fields

Essential settings

facts — Facts

Tasks, requirements, what is offered, working hours. Plain notes are enough.

Default settings: ""

job_title — Job title

Added word for word as the title.

Default settings: ""

company — Company

Name of the employer.

Default settings: ""

location — Location

Place of work or remote.

Default settings: ""

salary — Salary

Added word for word. Leave empty if not published.

Default settings: ""

apply_contact — How to apply

Contact or address for applications, added word for word.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "job_ad_create",
    "text": "{\"facts\":\"Support the workshop team. Plan repair orders, order parts, talk to customers. Completed training as a car mechanic required, driving licence required. 38 hours per week, 30 days of holiday.\",\"job_title\":\"Service advisor (all genders)\",\"company\":\"Martens Garage\",\"location\":\"Lindenau\",\"salary\":\"\",\"apply_contact\":\"jobs@martens-garage.example\",\"tone\":\"neutral\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "facts": {
            "type": "string",
            "maxLength": 8000,
            "minLength": 1
        },
        "job_title": {
            "type": "string",
            "maxLength": 120,
            "minLength": 1
        },
        "company": {
            "type": "string",
            "maxLength": 120,
            "minLength": 1
        },
        "location": {
            "type": "string",
            "maxLength": 120
        },
        "salary": {
            "type": "string",
            "maxLength": 120
        },
        "apply_contact": {
            "type": "string",
            "maxLength": 300
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        }
    },
    "required": [
        "facts",
        "job_title",
        "company",
        "tone"
    ]
}

Cover letter for an offer

offer_cover_letter

Write the letter that accompanies an offer: amount, validity and offer number stay exactly as supplied.

Input and configuration fields

Essential settings

facts — Facts

What is offered, for whom, what matters to this customer.

Default settings: ""

customer — Customer

Name of the customer or contact person.

Default settings: ""

offer_number — Offer number

Added word for word.

Default settings: ""

amount — Amount

Added word for word, for example with currency.

Default settings: ""

valid_until — Valid until

Added word for word.

Default settings: ""

sender — Sender and signature

Name, role and contact details. Added word for word.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "offer_cover_letter",
    "text": "{\"facts\":\"Offer for repainting the stairwell in the Maple Street building. Two coats, work during office hours, the customer asked for low-odour paint.\",\"customer\":\"Ms Keller, Maple Street Property\",\"offer_number\":\"A-2041\",\"amount\":\"4,180 EUR plus VAT\",\"valid_until\":\"30 June\",\"sender\":\"Jonas Weber, Weber Painters\",\"tone\":\"friendly\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "facts": {
            "type": "string",
            "maxLength": 6000,
            "minLength": 1
        },
        "customer": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "offer_number": {
            "type": "string",
            "maxLength": 60
        },
        "amount": {
            "type": "string",
            "maxLength": 60
        },
        "valid_until": {
            "type": "string",
            "maxLength": 60
        },
        "sender": {
            "type": "string",
            "maxLength": 600
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        }
    },
    "required": [
        "facts",
        "customer",
        "tone"
    ]
}

Customer notice about a change

customer_change_notice

Announce a price, terms or service change clearly: what changes, from when, what customers need to do.

Input and configuration fields

Essential settings

facts — Facts

What changes, why, what stays the same, what customers can do.

Default settings: ""

organisation — Organisation

Name of the sender.

Default settings: ""

effective_date — Effective from

Added word for word.

Default settings: ""

contact — Contact details

Added word for word at the end.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "customer_change_notice",
    "text": "{\"facts\":\"The monthly membership rises from 29 to 32 EUR because of higher energy costs. Contracts can be cancelled until the change takes effect. Opening hours stay the same.\",\"organisation\":\"North Gym\",\"effective_date\":\"1 March\",\"contact\":\"service@north-gym.example\",\"tone\":\"neutral\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "facts": {
            "type": "string",
            "maxLength": 6000,
            "minLength": 1
        },
        "organisation": {
            "type": "string",
            "maxLength": 120,
            "minLength": 1
        },
        "effective_date": {
            "type": "string",
            "maxLength": 60,
            "minLength": 1
        },
        "contact": {
            "type": "string",
            "maxLength": 600
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        }
    },
    "required": [
        "facts",
        "organisation",
        "effective_date",
        "tone"
    ]
}

Polite decline

polite_decline

Write a respectful decline for a request, a supplier or an application, stating only the reason that was supplied.

Input and configuration fields

Essential settings

situation — Situation

Who asked for what. Plain notes are enough.

Default settings: ""

reason — Reason

Only the reason that may be stated. Leave empty to give no reason.

Default settings: ""

alternative — Alternative

Optional offer or next step.

Default settings: ""

sender — Sender and signature

Name, role and contact details. Added word for word.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "polite_decline",
    "text": "{\"situation\":\"Lena Hoffmann applied for the trainee position in our print shop. Interview on 4 April.\",\"reason\":\"Another applicant had more experience with offset printing.\",\"alternative\":\"We keep the application for six months if she agrees.\",\"sender\":\"Tom Richter, Richter Print\",\"tone\":\"friendly\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "situation": {
            "type": "string",
            "maxLength": 4000,
            "minLength": 1
        },
        "reason": {
            "type": "string",
            "maxLength": 1000
        },
        "alternative": {
            "type": "string",
            "maxLength": 1000
        },
        "sender": {
            "type": "string",
            "maxLength": 600
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        }
    },
    "required": [
        "situation",
        "tone"
    ]
}

Payment reminder

payment_reminder

Write a payment reminder at the chosen level. Invoice number, amount and date stay exact; no invented fees or threats.

Input and configuration fields

Essential settings

customer — Customer

Name of the customer.

Default settings: ""

invoice_number — Invoice number

Added word for word.

Default settings: ""

invoice_date — Invoice date

Added word for word.

Default settings: ""

amount — Open amount

Added word for word, with currency.

Default settings: ""

due_date — New due date

Added word for word.

Default settings: ""

level — Level

How firm the reminder is.

friendly — Friendly reminder · second — Second reminder · final — Final reminder

Default settings: "friendly"

payment_details — Payment details

Added word for word.

Default settings: ""

sender — Sender and signature

Name, role and contact details. Added word for word.

Default settings: ""

Advanced settings

notes — Notes

Optional: earlier contact, agreed fees, consequences that apply.

Default settings: ""

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "payment_reminder",
    "text": "{\"customer\":\"Baker & Sons\",\"invoice_number\":\"2026-118\",\"invoice_date\":\"3 February\",\"amount\":\"1,240.00 EUR\",\"due_date\":\"20 March\",\"level\":\"friendly\",\"notes\":\"\",\"payment_details\":\"IBAN DE00 0000 0000 0000 0000 00\",\"sender\":\"Accounts team, Lumen Print\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "customer": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "invoice_number": {
            "type": "string",
            "maxLength": 60,
            "minLength": 1
        },
        "invoice_date": {
            "type": "string",
            "maxLength": 40
        },
        "amount": {
            "type": "string",
            "maxLength": 60,
            "minLength": 1
        },
        "due_date": {
            "type": "string",
            "maxLength": 40,
            "minLength": 1
        },
        "level": {
            "type": "string",
            "enum": [
                "friendly",
                "second",
                "final"
            ]
        },
        "notes": {
            "type": "string",
            "maxLength": 2000
        },
        "payment_details": {
            "type": "string",
            "maxLength": 500
        },
        "sender": {
            "type": "string",
            "maxLength": 600
        }
    },
    "required": [
        "customer",
        "invoice_number",
        "amount",
        "due_date",
        "level"
    ]
}

Event invitation

event_invitation

Write an invitation with the date, place and reply details exactly as supplied.

Input and configuration fields

Essential settings

event — Event

Name or kind of event.

Default settings: ""

facts — Facts

Purpose, programme, who is invited, what to bring.

Default settings: ""

date_time — Date and time

Added word for word.

Default settings: ""

place — Place

Added word for word.

Default settings: ""

rsvp — Reply

How and until when to reply. Added word for word.

Default settings: ""

host — Host

Who invites.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "event_invitation",
    "text": "{\"event\":\"Summer party of the Lindenhof tenants\",\"facts\":\"Grill and drinks are provided, please bring a salad or a cake. Children welcome. Music from 18:00.\",\"date_time\":\"Saturday 21 June, 16:00\",\"place\":\"Courtyard, Lindenhof 4\",\"rsvp\":\"Reply to Mrs Yilmaz by 14 June\",\"host\":\"Tenants committee\",\"tone\":\"friendly\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "event": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "facts": {
            "type": "string",
            "maxLength": 5000,
            "minLength": 1
        },
        "date_time": {
            "type": "string",
            "maxLength": 80,
            "minLength": 1
        },
        "place": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "rsvp": {
            "type": "string",
            "maxLength": 160
        },
        "host": {
            "type": "string",
            "maxLength": 160
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        }
    },
    "required": [
        "event",
        "facts",
        "date_time",
        "place",
        "tone"
    ]
}

Cancellation letter

cancellation_letter

Write a clear cancellation letter for a contract with a request for written confirmation. No legal advice.

Input and configuration fields

Essential settings

contract — Contract

Kind of contract and contract number.

Default settings: ""

provider — Provider

Name of the company.

Default settings: ""

end_date — End of contract

For example a date or "next possible date". Added word for word.

Default settings: ""

customer_block — Customer details

Name, address and customer number. Added word for word.

Default settings: ""

Advanced settings

notes — Notes

Optional: reason, direct debit to stop, data to delete.

Default settings: ""

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "cancellation_letter",
    "text": "{\"contract\":\"Mobile contract, customer number 88213\",\"provider\":\"Telecom North\",\"end_date\":\"next possible date\",\"customer_block\":\"Maria Lang, Oak Street 7, 12345 Riverton\",\"notes\":\"Please stop the direct debit after the last invoice.\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "contract": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "provider": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "end_date": {
            "type": "string",
            "maxLength": 80,
            "minLength": 1
        },
        "customer_block": {
            "type": "string",
            "maxLength": 500,
            "minLength": 1
        },
        "notes": {
            "type": "string",
            "maxLength": 1500
        }
    },
    "required": [
        "contract",
        "provider",
        "end_date",
        "customer_block"
    ]
}

Complaint letter

complaint_letter

Turn notes into a factual complaint with a clear request and deadline, without threats or invented details.

Input and configuration fields

Essential settings

facts — Facts

What happened, when, what was agreed, what has been tried.

Default settings: ""

recipient — Recipient

Company or person.

Default settings: ""

desired_outcome — Requested outcome

What should happen now.

Default settings: ""

reference — Reference

Order or customer number. Added word for word.

Default settings: ""

deadline — Deadline

Added word for word.

Default settings: ""

sender — Sender and signature

Name, role and contact details. Added word for word.

Default settings: ""

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "complaint_letter",
    "text": "{\"facts\":\"Washing machine ordered on 3 March, delivered on 20 March instead of 10 March. Door seal damaged on delivery. Reported by phone on 21 March, no answer since.\",\"recipient\":\"HomeTech Service\",\"desired_outcome\":\"Replacement of the door seal or of the machine.\",\"reference\":\"Order 55821\",\"deadline\":\"14 April\",\"sender\":\"Paul Novak, Elm Road 2\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "facts": {
            "type": "string",
            "maxLength": 6000,
            "minLength": 1
        },
        "recipient": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "desired_outcome": {
            "type": "string",
            "maxLength": 800,
            "minLength": 1
        },
        "reference": {
            "type": "string",
            "maxLength": 80
        },
        "deadline": {
            "type": "string",
            "maxLength": 60
        },
        "sender": {
            "type": "string",
            "maxLength": 600
        }
    },
    "required": [
        "facts",
        "recipient",
        "desired_outcome"
    ]
}

Company profile

company_profile

Write an "about us" text in two lengths from supplied facts, without invented figures or superlatives.

Input and configuration fields

Essential settings

facts — Facts

What the company does, for whom, since when, how it works.

Default settings: ""

company — Company

Name of the company.

Default settings: ""

location — Location

Optional.

Default settings: ""

contact — Contact details

Added word for word at the end.

Default settings: ""

Advanced settings

length — Length

Approximate length of the draft.

short — Short · medium — Medium · long — Long

Default settings: "medium"

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "company_profile",
    "text": "{\"facts\":\"Family business for garden furniture, founded 2003. Own workshop with 14 people. Wood from regional forests. Sells to private customers and hotels. Repairs for 10 years.\",\"company\":\"Waldholz Furniture\",\"location\":\"Freiberg\",\"contact\":\"hello@waldholz.example\",\"length\":\"short\",\"tone\":\"friendly\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "facts": {
            "type": "string",
            "maxLength": 6000,
            "minLength": 1
        },
        "company": {
            "type": "string",
            "maxLength": 120,
            "minLength": 1
        },
        "location": {
            "type": "string",
            "maxLength": 120
        },
        "contact": {
            "type": "string",
            "maxLength": 600
        },
        "length": {
            "type": "string",
            "enum": [
                "short",
                "medium",
                "long"
            ]
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        }
    },
    "required": [
        "facts",
        "company",
        "length",
        "tone"
    ]
}

Speech draft

speech_draft

Draft a speech for an occasion from supplied facts, with no invented anecdotes, names or numbers.

Input and configuration fields

Essential settings

occasion — Occasion

For example anniversary, farewell, opening.

Default settings: ""

speaker — Speaker

Who speaks, in which role.

Default settings: ""

audience — Audience

Who listens.

Default settings: ""

facts — Facts

Names, milestones, what should be said. Only what is written here can appear.

Default settings: ""

Advanced settings

length — Length

Approximate length of the draft.

short — Short · medium — Medium · long — Long

Default settings: "medium"

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "speech_draft",
    "text": "{\"occasion\":\"Farewell of Helga Sommer after 31 years\",\"speaker\":\"Managing director\",\"audience\":\"Colleagues and family\",\"facts\":\"Helga Sommer started in 1995 in accounting and later led the team of 6. She introduced the first online invoices and trained 12 apprentices. She will now travel to Norway.\",\"length\":\"short\",\"tone\":\"friendly\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "occasion": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "speaker": {
            "type": "string",
            "maxLength": 160
        },
        "audience": {
            "type": "string",
            "maxLength": 160
        },
        "facts": {
            "type": "string",
            "maxLength": 6000,
            "minLength": 1
        },
        "length": {
            "type": "string",
            "enum": [
                "short",
                "medium",
                "long"
            ]
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        }
    },
    "required": [
        "occasion",
        "facts",
        "length",
        "tone"
    ]
}

Grant application summary

grant_summary

Write the project summary of a grant application from supplied facts, with no invented figures or effects.

Input and configuration fields

Essential settings

project — Project title

Added word for word.

Default settings: ""

organisation — Organisation

Applicant.

Default settings: ""

funder — Funder

Optional, for choosing the wording.

Default settings: ""

facts — Facts

Problem, goals, activities, target group, expected effects.

Default settings: ""

amount — Requested amount

Added word for word.

Default settings: ""

period — Period

Added word for word.

Default settings: ""

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "grant_summary",
    "text": "{\"project\":\"Reading afternoons in the district library\",\"organisation\":\"Friends of the East Library\",\"funder\":\"City culture fund\",\"facts\":\"Children from 6 to 10 read aloud with volunteers on two afternoons per week. Many families have few books at home. 8 volunteers are ready, books are provided by the library. Goal: more children use the library regularly.\",\"amount\":\"6,500 EUR\",\"period\":\"September to June\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "project": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "organisation": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "funder": {
            "type": "string",
            "maxLength": 160
        },
        "facts": {
            "type": "string",
            "maxLength": 8000,
            "minLength": 1
        },
        "amount": {
            "type": "string",
            "maxLength": 60
        },
        "period": {
            "type": "string",
            "maxLength": 60
        }
    },
    "required": [
        "project",
        "organisation",
        "facts"
    ]
}

Letter to parents

parent_letter

Write a clear letter from a school or nursery to parents: what happens, when, and what parents need to do.

Input and configuration fields

Essential settings

school — School or nursery

Name of the institution.

Default settings: ""

facts — Facts

What happens, what is needed, what it costs.

Default settings: ""

date_time — Date and time

Added word for word.

Default settings: ""

deadline — Reply deadline

Added word for word.

Default settings: ""

sender — Sender and signature

Name, role and contact details. Added word for word.

Default settings: ""

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "parent_letter",
    "text": "{\"school\":\"Sunflower Primary School\",\"facts\":\"Class 3b visits the zoo. Children need a packed lunch and rain clothes. The trip costs 8 EUR, to be paid in class. We return by 14:00.\",\"date_time\":\"Tuesday 13 May, 8:30\",\"deadline\":\"Friday 9 May\",\"sender\":\"Ms Aydin, class teacher\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "school": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "facts": {
            "type": "string",
            "maxLength": 5000,
            "minLength": 1
        },
        "date_time": {
            "type": "string",
            "maxLength": 80
        },
        "deadline": {
            "type": "string",
            "maxLength": 60
        },
        "sender": {
            "type": "string",
            "maxLength": 600
        }
    },
    "required": [
        "school",
        "facts"
    ]
}

Notice for a board or door

notice_board

Write a short notice for a board, door or stairwell: what happens, when, what to do.

Input and configuration fields

Essential settings

facts — Facts

What happens, who is affected, what to do.

Default settings: ""

date_time — Date and time

Added word for word.

Default settings: ""

place — Place

Added word for word.

Default settings: ""

contact — Contact details

Added word for word at the end.

Default settings: ""

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "notice_board",
    "text": "{\"facts\":\"The water is switched off because the main valve is being replaced. Please fill some water beforehand and do not use the washing machine.\",\"date_time\":\"Thursday 8 May, 9:00 to 12:00\",\"place\":\"House numbers 4 to 8\",\"contact\":\"Caretaker Mr Lang, 0151 000000\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "facts": {
            "type": "string",
            "maxLength": 3000,
            "minLength": 1
        },
        "date_time": {
            "type": "string",
            "maxLength": 80,
            "minLength": 1
        },
        "place": {
            "type": "string",
            "maxLength": 160
        },
        "contact": {
            "type": "string",
            "maxLength": 600
        }
    },
    "required": [
        "facts",
        "date_time"
    ]
}

Application cover letter

cover_letter

Write a job application letter from the job advertisement and a profile. Only stated skills and experience appear.

Input and configuration fields

Essential settings

job_ad — Job advertisement

Text of the advertisement.

Default settings: ""

profile — Own profile

Experience, skills, education, motivation. Only this may be claimed.

Default settings: ""

company — Company

Name of the employer.

Default settings: ""

applicant — Applicant details

Name and contact details. Added word for word.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "cover_letter",
    "text": "{\"job_ad\":\"We are looking for a team assistant with experience in event organisation. Good English required.\",\"profile\":\"Six years as office manager at a trade fair company. Organised 40 customer events per year. English at B2 level. Looking for a role with more responsibility.\",\"company\":\"Norden Events\",\"applicant\":\"Sara Jensen, sara@example.com\",\"tone\":\"neutral\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "job_ad": {
            "type": "string",
            "maxLength": 6000,
            "minLength": 1
        },
        "profile": {
            "type": "string",
            "maxLength": 6000,
            "minLength": 1
        },
        "company": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "applicant": {
            "type": "string",
            "maxLength": 400,
            "minLength": 1
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        }
    },
    "required": [
        "job_ad",
        "profile",
        "company",
        "applicant",
        "tone"
    ]
}

Property listing

property_listing

Write a property listing from supplied facts. Size and price stay exact; no invented features or neighbourhood claims.

Input and configuration fields

Essential settings

facts — Facts

Size, rooms, condition, year, features, what is near.

Default settings: ""

location — Location

Added word for word.

Default settings: ""

price — Price or rent

Added word for word.

Default settings: ""

contact — Contact details

Added word for word at the end.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "property_listing",
    "text": "{\"facts\":\"3 rooms, 74 square metres, second floor, balcony facing south, renovated 2021, fitted kitchen, cellar, tram stop 5 minutes on foot.\",\"location\":\"Schillerstrasse, Leipzig\",\"price\":\"980 EUR plus utilities\",\"contact\":\"Immobilien Kern, 0341 000000\",\"tone\":\"neutral\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "facts": {
            "type": "string",
            "maxLength": 6000,
            "minLength": 1
        },
        "location": {
            "type": "string",
            "maxLength": 160
        },
        "price": {
            "type": "string",
            "maxLength": 80
        },
        "contact": {
            "type": "string",
            "maxLength": 600
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        }
    },
    "required": [
        "facts",
        "tone"
    ]
}

Classified ad

classified_ad

Write a short, honest classified ad for an item, with defects named and the price kept exact.

Input and configuration fields

Essential settings

facts — Facts

What it is, age, condition, defects, what is included.

Default settings: ""

price — Price

Added word for word.

Default settings: ""

location — Location

Added word for word.

Default settings: ""

contact — Contact

Added word for word.

Default settings: ""

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "classified_ad",
    "text": "{\"facts\":\"Child bike 20 inch, bought 2022, 6 gears, blue. Small scratch on the frame, new brakes last year. Lights included.\",\"price\":\"120 EUR\",\"location\":\"Potsdam\",\"contact\":\"Message on the platform\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "facts": {
            "type": "string",
            "maxLength": 3000,
            "minLength": 1
        },
        "price": {
            "type": "string",
            "maxLength": 60
        },
        "location": {
            "type": "string",
            "maxLength": 120
        },
        "contact": {
            "type": "string",
            "maxLength": 200
        }
    },
    "required": [
        "facts"
    ]
}

Donation appeal

donation_appeal

Write a donation appeal from supplied facts, with no invented impact figures and no pressure.

Input and configuration fields

Essential settings

organisation — Organisation

Name of the organisation.

Default settings: ""

facts — Facts

What the money is for, who benefits, what has been done so far.

Default settings: ""

goal — Donation goal

Added word for word.

Default settings: ""

donation_details — How to donate

Account or link. Added word for word.

Default settings: ""

contact — Contact details

Added word for word at the end.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "donation_appeal",
    "text": "{\"organisation\":\"Animal Shelter Bergen\",\"facts\":\"The roof of the dog house leaks and has to be replaced before winter. The shelter cares for about 40 dogs. A roofer quoted 12,000 EUR. 5,000 EUR has been collected so far.\",\"goal\":\"7,000 EUR by October\",\"donation_details\":\"IBAN DE00 0000 0000 0000 0000 00, reference Roof\",\"contact\":\"info@shelter-bergen.example\",\"tone\":\"friendly\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "organisation": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "facts": {
            "type": "string",
            "maxLength": 6000,
            "minLength": 1
        },
        "goal": {
            "type": "string",
            "maxLength": 80
        },
        "donation_details": {
            "type": "string",
            "maxLength": 500
        },
        "contact": {
            "type": "string",
            "maxLength": 600
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        }
    },
    "required": [
        "organisation",
        "facts",
        "tone"
    ]
}

Release notes

release_notes

Turn a list of changes into readable release notes. Every change appears once, and numbers stay as supplied.

Input and configuration fields

Essential settings

changes — Changes

One change per row. Type is added, changed, fixed, removed or security.

Default settings: ""

product — Product

Name of the product.

Default settings: ""

version — Version

Added word for word.

Default settings: ""

release_date — Date

Added word for word.

Default settings: ""

audience — Audience

Who reads the notes.

customers — Customers · developers — Developers · internal — Internal team

Default settings: "customers"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "release_notes",
    "text": "{\"changes\":\"id,type,text\\nC1,added,CSV export for invoices\\nC2,fixed,Login failed when the e-mail contained capital letters\\nC3,changed,Reports load in the background\",\"product\":\"Billy Invoices\",\"version\":\"3.2\",\"release_date\":\"12 May\",\"audience\":\"customers\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "changes": {
            "type": "string",
            "maxLength": 30000,
            "minLength": 1
        },
        "product": {
            "type": "string",
            "maxLength": 80,
            "minLength": 1
        },
        "version": {
            "type": "string",
            "maxLength": 40
        },
        "release_date": {
            "type": "string",
            "maxLength": 40
        },
        "audience": {
            "type": "string",
            "enum": [
                "customers",
                "developers",
                "internal"
            ]
        }
    },
    "required": [
        "changes",
        "product",
        "audience"
    ]
}

Video chapters

video_chapters

Create chapter marks from subtitles or a transcript with time codes. The times come from the file, only titles from the AI.

Input and configuration fields

Essential settings

transcript — Transcript with time codes

SRT, VTT or lines that begin with a time such as 01:12.

Default settings: ""

max_chapters — Maximum chapters

From 3 to 20.

Default settings: 8

Advanced settings

min_gap_seconds — Minimum distance in seconds

From 10 to 600.

Default settings: 60

Maximum UTF-8 input bytes: 70000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "video_chapters",
    "text": "{\"transcript\":\"00:00 Welcome to the workshop and a short overview of today.\\n00:40 We start with the tools you need for changing a tyre.\\n01:30 First loosen the wheel nuts while the car is on the ground.\\n02:20 Now jack up the car at the marked point.\\n03:10 Remove the wheel and mount the spare.\\n04:05 Tighten the nuts crosswise and lower the car.\\n05:00 Check the pressure and store the tools.\",\"max_chapters\":4,\"min_gap_seconds\":60}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "transcript": {
            "type": "string",
            "maxLength": 60000,
            "minLength": 1
        },
        "max_chapters": {
            "type": "integer",
            "minimum": 3,
            "maximum": 20
        },
        "min_gap_seconds": {
            "type": "integer",
            "minimum": 10,
            "maximum": 600
        }
    },
    "required": [
        "transcript",
        "max_chapters",
        "min_gap_seconds"
    ]
}

Meeting agenda

meeting_agenda

Build a timed meeting agenda from topics and durations. The server calculates the times, and the AI words the items and expected results.

Input and configuration fields

Essential settings

meeting_title — Meeting title

Added word for word.

Default settings: ""

objective — Objective

What the meeting should achieve.

Default settings: ""

start_time — Start time

Format 09:30.

Default settings: "09:00"

topics — Topics

One topic per row with the responsible person and minutes.

Default settings: ""

Advanced settings

available_minutes — Available minutes

0 means no limit. A longer agenda is reported.

Default settings: 0

break_minutes — Break after every third topic

Minutes, 0 means no break.

Default settings: 0

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "meeting_agenda",
    "text": "{\"meeting_title\":\"Spring planning\",\"objective\":\"Decide the order of the three projects\",\"start_time\":\"09:30\",\"topics\":\"id,topic,owner,minutes\\nT1,review of last quarter,Anna,20\\nT2,projects A B C budget,Marek,40\\nT3,who does what,Anna,30\",\"available_minutes\":90,\"break_minutes\":0}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "meeting_title": {
            "type": "string",
            "maxLength": 120,
            "minLength": 1
        },
        "objective": {
            "type": "string",
            "maxLength": 600
        },
        "start_time": {
            "type": "string",
            "maxLength": 5,
            "minLength": 1
        },
        "topics": {
            "type": "string",
            "maxLength": 20000,
            "minLength": 1
        },
        "available_minutes": {
            "type": "integer",
            "minimum": 0,
            "maximum": 1440
        },
        "break_minutes": {
            "type": "integer",
            "minimum": 0,
            "maximum": 60
        }
    },
    "required": [
        "meeting_title",
        "start_time",
        "topics",
        "available_minutes",
        "break_minutes"
    ]
}

Questions to a document

document_questions

Answer questions from a pasted document. Each answer needs an exact quote; otherwise the result says the document does not state it.

Input and configuration fields

Essential settings

document — Document

Paste the text. It is treated as data, never as instructions.

Default settings: ""

questions — Questions

One question per line, up to 12.

Default settings: ""

Advanced settings

answer_length — Answer length

Brief or detailed answers.

brief — Brief · detailed — Detailed

Default settings: "brief"

Maximum UTF-8 input bytes: 40000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "document_questions",
    "text": "{\"document\":\"The rental period begins on 1 June and runs for 24 months.\\n\\nThe tenant pays a monthly rent of 950 EUR by the third working day of each month.\\n\\nPets require written permission from the landlord.\",\"questions\":\"How long does the rental period run?\\nCan the tenant keep a dog?\\nWho pays for repairs?\",\"answer_length\":\"brief\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "document": {
            "type": "string",
            "maxLength": 30000,
            "minLength": 1
        },
        "questions": {
            "type": "string",
            "maxLength": 3700,
            "minLength": 1
        },
        "answer_length": {
            "type": "string",
            "enum": [
                "brief",
                "detailed"
            ]
        }
    },
    "required": [
        "document",
        "questions",
        "answer_length"
    ]
}

Contract clauses to review

contract_clause_review

Find clauses worth a closer look in a contract: renewal, termination, liability, payment, penalties and data, each with an exact quote.

Input and configuration fields

Essential settings

document — Contract text

Paste the text. It is treated as data, never as instructions.

Default settings: ""

Maximum UTF-8 input bytes: 32000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "contract_clause_review",
    "text": "{\"document\":\"This agreement renews automatically for 12 months unless cancelled in writing 3 months before the end of the term.\\n\\nThe liability of the provider is limited to the fees paid in the previous 6 months.\\n\\nInvoices are due within 14 days. Late payment incurs a fee of 2 percent per month.\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "document": {
            "type": "string",
            "maxLength": 30000,
            "minLength": 1
        }
    },
    "required": [
        "document"
    ]
}

Tasks and decisions from text

action_items

Extract tasks, decisions and open questions from e-mails or notes. Owners and dates appear only if they are in the text.

Input and configuration fields

Essential settings

document — E-mails or notes

Paste the text. It is treated as data, never as instructions.

Default settings: ""

Maximum UTF-8 input bytes: 32000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "action_items",
    "text": "{\"document\":\"Hi all, we agreed to move the launch to 12 June. Marek will update the price list by Friday. Nobody has checked yet whether the supplier can deliver earlier.\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "document": {
            "type": "string",
            "maxLength": 30000,
            "minLength": 1
        }
    },
    "required": [
        "document"
    ]
}

Contradictions in a document

contradiction_check

Find statements in a document that contradict each other, such as different dates or amounts, with both passages quoted.

Input and configuration fields

Essential settings

document — Document

Paste the text. It is treated as data, never as instructions.

Default settings: ""

Maximum UTF-8 input bytes: 32000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "contradiction_check",
    "text": "{\"document\":\"The project starts on 1 March and ends after 6 months.\\n\\nThe budget is 40,000 EUR in total.\\n\\nIn section 4 the project ends on 30 November. The total budget of 45,000 EUR includes travel.\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "document": {
            "type": "string",
            "maxLength": 30000,
            "minLength": 1
        }
    },
    "required": [
        "document"
    ]
}

Missing information in a request

missing_information

Find what is missing in a request or brief before work starts, and get the questions to ask.

Input and configuration fields

Essential settings

document — Request or brief

Paste the text. It is treated as data, never as instructions.

Default settings: ""

purpose — What it is needed for

For example "prepare an offer" or "plan the event".

Default settings: ""

Maximum UTF-8 input bytes: 22000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "missing_information",
    "text": "{\"document\":\"Hello, we need our office repainted soon. It has several rooms and the walls have some marks. Please send a price.\",\"purpose\":\"prepare an offer\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "document": {
            "type": "string",
            "maxLength": 20000,
            "minLength": 1
        },
        "purpose": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        }
    },
    "required": [
        "document",
        "purpose"
    ]
}

Official letter explained

official_letter_explainer

Explain an official letter in plain language: what it says, deadlines, amounts and what to do, each with the exact passage. No legal advice.

Input and configuration fields

Essential settings

document — Letter text

Paste the text. Personal data can be removed first.

Default settings: ""

Maximum UTF-8 input bytes: 22000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "official_letter_explainer",
    "text": "{\"document\":\"Notice of 4 May. Your application for housing benefit has been received. Please submit your rental contract and proof of income by 25 May. Without these documents we cannot decide on your application. Questions: Housing Office, room 214.\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "document": {
            "type": "string",
            "maxLength": 20000,
            "minLength": 1
        }
    },
    "required": [
        "document"
    ]
}

Checklist from a text

checklist_from_text

Turn instructions or a procedure into a checklist. Each point is tied to an exact passage, and nothing is added.

Input and configuration fields

Essential settings

document — Instructions or procedure

Paste the text. It is treated as data, never as instructions.

Default settings: ""

Maximum UTF-8 input bytes: 32000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "checklist_from_text",
    "text": "{\"document\":\"Before starting, switch off the machine and pull the plug. Open the cover with the supplied key. Remove the filter and rinse it under cold water. Never use solvents, as they damage the seal. Let the filter dry for one hour before refitting.\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "document": {
            "type": "string",
            "maxLength": 30000,
            "minLength": 1
        }
    },
    "required": [
        "document"
    ]
}

Ticket triage

ticket_triage

Sort support tickets into maintained categories with urgency, an exact passage and a suggested next step. Nothing is sent.

Input and configuration fields

Essential settings

items — Tickets

One ticket per row. Text can be pasted from any system.

Default settings: ""

categories — Categories

Maintained categories with a short definition.

Default settings: ""

Maximum UTF-8 input bytes: 56000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "ticket_triage",
    "text": "{\"items\":\"id,subject,text\\nT1,Login fails,Since this morning nobody in our team can log in. We cannot work.\\nT2,Invoice address,Please change our invoice address to Mill Road 3.\\nT3,Idea,It would be nice to have a dark mode.\",\"categories\":\"id,name,definition\\nC1,Technical fault,Something does not work\\nC2,Billing,Invoices and payments\\nC3,Feature request,Wishes for new functions\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "items": {
            "type": "string",
            "maxLength": 40000,
            "minLength": 1
        },
        "categories": {
            "type": "string",
            "maxLength": 15000,
            "minLength": 1
        }
    },
    "required": [
        "items",
        "categories"
    ]
}

Feedback sorting

feedback_sorting

Sort open feedback into maintained themes, with an exact passage for each item and counts per theme.

Input and configuration fields

Essential settings

items — Feedback

One entry per row, for example from a survey.

Default settings: ""

categories — Themes

Maintained themes with a short definition.

Default settings: ""

Maximum UTF-8 input bytes: 56000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "feedback_sorting",
    "text": "{\"items\":\"id,source,text\\nF1,Survey,The staff at the counter were very friendly.\\nF2,E-mail,I waited over an hour for my appointment.\\nF3,Survey,Parking is hard to find near the building.\",\"categories\":\"id,name,definition\\nS1,Staff,Behaviour and advice of staff\\nS2,Waiting time,Waiting and appointments\\nS3,Location,Access and parking\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "items": {
            "type": "string",
            "maxLength": 40000,
            "minLength": 1
        },
        "categories": {
            "type": "string",
            "maxLength": 15000,
            "minLength": 1
        }
    },
    "required": [
        "items",
        "categories"
    ]
}

Recommendation letter

recommendation_letter

Write a letter of recommendation from supplied facts, without invented achievements, ratings or comparisons.

Input and configuration fields

Essential settings

person — Person

Who is being recommended.

Default settings: ""

relationship — Relationship

How the writer knows the person, and for how long.

Default settings: ""

purpose — Recommended for

Optional: job, course or programme.

Default settings: ""

facts — Facts

Plain notes are enough. Only what is written here can appear in the draft.

Default settings: ""

sender — Sender and signature

Name, role and contact details. Added word for word.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "recommendation_letter",
    "text": "{\"person\":\"Mira Kovac\",\"relationship\":\"Team lead for three years at Lumen Print\",\"purpose\":\"Project manager role\",\"facts\":\"Mira planned the move of our print workshop in 2025. She trained two new colleagues. She kept customers informed during the move.\",\"tone\":\"friendly\",\"sender\":\"Tom Berg, Head of Production, Lumen Print\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "person": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "relationship": {
            "type": "string",
            "maxLength": 240,
            "minLength": 1
        },
        "purpose": {
            "type": "string",
            "maxLength": 200
        },
        "facts": {
            "type": "string",
            "maxLength": 6000,
            "minLength": 1
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        },
        "sender": {
            "type": "string",
            "maxLength": 600,
            "minLength": 1
        }
    },
    "required": [
        "person",
        "relationship",
        "facts",
        "tone",
        "sender"
    ]
}

CV from notes

cv_from_notes

Turn notes on work and education into a structured CV, with dates and employers exactly as supplied.

Input and configuration fields

Essential settings

full_name — Name

Added word for word.

Default settings: ""

contact — Contact details

Added word for word below the name.

Default settings: ""

target_role — Target role

Optional: the job this CV is for.

Default settings: ""

experience — Work experience

Positions, employers, dates and tasks. Plain notes are enough.

Default settings: ""

education — Education

Optional: schools, courses and dates.

Default settings: ""

skills — Skills

Optional: languages, tools and other skills.

Default settings: ""

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "cv_from_notes",
    "text": "{\"full_name\":\"Mira Kovac\",\"contact\":\"mira.kovac@example.org\",\"target_role\":\"Project manager\",\"experience\":\"2022 to today: team coordinator at Lumen Print, planned workshop move, trained new colleagues. 2018 to 2022: print technician at Atlas Druck.\",\"education\":\"2018 vocational training as print technician\",\"skills\":\"Project planning, German, English\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "full_name": {
            "type": "string",
            "maxLength": 120,
            "minLength": 1
        },
        "contact": {
            "type": "string",
            "maxLength": 400
        },
        "target_role": {
            "type": "string",
            "maxLength": 160
        },
        "experience": {
            "type": "string",
            "maxLength": 5000,
            "minLength": 1
        },
        "education": {
            "type": "string",
            "maxLength": 2000
        },
        "skills": {
            "type": "string",
            "maxLength": 1000
        }
    },
    "required": [
        "full_name",
        "experience"
    ]
}

Thank-you letter

thank_you_letter

Write a thank-you letter to customers, donors or partners from the supplied occasion and facts.

Input and configuration fields

Essential settings

recipient — Recipient

Person or organisation.

Default settings: ""

occasion — Occasion

What the thanks are for.

Default settings: ""

facts — Facts

Plain notes are enough. Only what is written here can appear in the draft.

Default settings: ""

sender — Sender and signature

Name, role and contact details. Added word for word.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "thank_you_letter",
    "text": "{\"recipient\":\"Riverside Bakery\",\"occasion\":\"Sponsoring our summer fair\",\"facts\":\"The bakery donated bread for 200 guests. Two of their staff helped at the stand.\",\"tone\":\"friendly\",\"sender\":\"Anna Berg, Board, Riverside Association\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "recipient": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "occasion": {
            "type": "string",
            "maxLength": 240,
            "minLength": 1
        },
        "facts": {
            "type": "string",
            "maxLength": 4000,
            "minLength": 1
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        },
        "sender": {
            "type": "string",
            "maxLength": 600
        }
    },
    "required": [
        "recipient",
        "occasion",
        "facts",
        "tone"
    ]
}

Appointment confirmation

appointment_confirmation

Write an appointment confirmation with date, place and change details exactly as supplied.

Input and configuration fields

Essential settings

recipient — Recipient

Name of the person.

Default settings: ""

subject_matter — Appointment

What the appointment is about.

Default settings: ""

date_time — Date and time

Added word for word.

Default settings: ""

place — Place

Added word for word.

Default settings: ""

notes — Notes

Optional: what to bring, how to prepare.

Default settings: ""

cancellation — Change or cancel

How to change or cancel. Added word for word.

Default settings: ""

sender — Sender and signature

Name, role and contact details. Added word for word.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "appointment_confirmation",
    "text": "{\"recipient\":\"Mr Novak\",\"subject_matter\":\"Heating inspection\",\"date_time\":\"12 May, 10:00\",\"place\":\"Mill Road 4\",\"notes\":\"Please make sure the boiler room is accessible. Bring the last inspection report.\",\"cancellation\":\"Call 030 000000 by the day before.\",\"tone\":\"neutral\",\"sender\":\"Service team, Heizung Berg\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "recipient": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "subject_matter": {
            "type": "string",
            "maxLength": 200,
            "minLength": 1
        },
        "date_time": {
            "type": "string",
            "maxLength": 80,
            "minLength": 1
        },
        "place": {
            "type": "string",
            "maxLength": 200
        },
        "notes": {
            "type": "string",
            "maxLength": 2000
        },
        "cancellation": {
            "type": "string",
            "maxLength": 400
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        },
        "sender": {
            "type": "string",
            "maxLength": 600
        }
    },
    "required": [
        "recipient",
        "subject_matter",
        "date_time",
        "tone"
    ]
}

Out-of-office message

out_of_office

Write an out-of-office message for people outside the organisation and one for colleagues, with the period as supplied.

Input and configuration fields

Essential settings

person — Person

Name of the person who is away.

Default settings: ""

date_range — Period

From and until, as it should appear.

Default settings: ""

substitute — Who takes over

Name and contact. Added word for word.

Default settings: ""

notes — Notes

Optional: urgent matters, what waits until the return.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "out_of_office",
    "text": "{\"person\":\"Mira Kovac\",\"date_range\":\"4 to 15 August\",\"substitute\":\"Tom Berg, tom.berg@example.org\",\"notes\":\"Invoices wait until the return.\",\"tone\":\"friendly\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "person": {
            "type": "string",
            "maxLength": 120,
            "minLength": 1
        },
        "date_range": {
            "type": "string",
            "maxLength": 80,
            "minLength": 1
        },
        "substitute": {
            "type": "string",
            "maxLength": 300
        },
        "notes": {
            "type": "string",
            "maxLength": 1500
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        }
    },
    "required": [
        "person",
        "date_range",
        "tone"
    ]
}

Offer follow-up

offer_follow_up

Write a polite follow-up on an open offer, without pressure, invented deadlines or discounts.

Input and configuration fields

Essential settings

customer — Customer

Name of the customer.

Default settings: ""

offer_reference — Offer

Number or title and date. Added word for word.

Default settings: ""

facts — Facts

What was offered, what was discussed, what the next step would be.

Default settings: ""

sender — Sender and signature

Name, role and contact details. Added word for word.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "offer_follow_up",
    "text": "{\"customer\":\"Atlas Druck\",\"offer_reference\":\"Offer 2026-41 of 3 March\",\"facts\":\"New folding machine with installation. We talked about delivery in May. Next step would be a short call.\",\"tone\":\"neutral\",\"sender\":\"Anna Berg, Sales, Lumen Print\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "customer": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "offer_reference": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "facts": {
            "type": "string",
            "maxLength": 3000,
            "minLength": 1
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        },
        "sender": {
            "type": "string",
            "maxLength": 600
        }
    },
    "required": [
        "customer",
        "offer_reference",
        "facts",
        "tone"
    ]
}

Quick guide

quick_guide

Turn notes into a short step-by-step guide, with warnings only as supplied.

Input and configuration fields

Essential settings

guide_title — Title

Added word for word.

Default settings: ""

audience — For whom

Optional: who uses this guide.

Default settings: ""

notes — Notes

Steps, settings and warnings in any order.

Default settings: ""

support — Help

Optional: who to ask. Added word for word.

Default settings: ""

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "quick_guide",
    "text": "{\"guide_title\":\"Clean the coffee machine\",\"audience\":\"Office team\",\"notes\":\"Turn off and let it cool. Empty the drip tray. Rinse the tray with warm water. Do not use the dishwasher. Press the clean button for 3 seconds.\",\"support\":\"Facility team, room 12\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "guide_title": {
            "type": "string",
            "maxLength": 120,
            "minLength": 1
        },
        "audience": {
            "type": "string",
            "maxLength": 160
        },
        "notes": {
            "type": "string",
            "maxLength": 6000,
            "minLength": 1
        },
        "support": {
            "type": "string",
            "maxLength": 300
        }
    },
    "required": [
        "guide_title",
        "notes"
    ]
}

Safety briefing

safety_briefing

Write a safety briefing from the rules and hazards supplied, without adding hazards, rules or legal statements.

Input and configuration fields

Essential settings

workplace — Workplace

Area or task the briefing is for.

Default settings: ""

notes — Rules and hazards

Own rules, hazards and emergency steps. Nothing else is added.

Default settings: ""

responsible — Responsible

Added word for word.

Default settings: ""

briefing_date — Date of the briefing

Added word for word.

Default settings: ""

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "safety_briefing",
    "text": "{\"workplace\":\"Print workshop\",\"notes\":\"Hazards: moving rollers, solvent vapours. Rules: tie back long hair, wear gloves when cleaning, keep doors open when using solvent. Emergency: press the red stop button, call the shift lead.\",\"responsible\":\"Tom Berg\",\"briefing_date\":\"12 May\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "workplace": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "notes": {
            "type": "string",
            "maxLength": 6000,
            "minLength": 1
        },
        "responsible": {
            "type": "string",
            "maxLength": 160
        },
        "briefing_date": {
            "type": "string",
            "maxLength": 40
        }
    },
    "required": [
        "workplace",
        "notes"
    ]
}

Tenant notice

tenant_notice

Write a notice to tenants about work, access or changes, with date and contact exactly as supplied.

Input and configuration fields

Essential settings

property — Property

Address or building.

Default settings: ""

topic — Topic

What the notice is about.

Default settings: ""

facts — Facts

Plain notes are enough. Only what is written here can appear in the draft.

Default settings: ""

date_time — Date and time

Added word for word.

Default settings: ""

contact — Contact details

Added word for word.

Default settings: ""

sender — Sender and signature

Name, role and contact details. Added word for word.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "tenant_notice",
    "text": "{\"property\":\"Mill Road 4\",\"topic\":\"Water shut-off for pipe repair\",\"facts\":\"The water is off in the whole building for about 4 hours. Please collect water beforehand and keep the access to the cellar free.\",\"date_time\":\"12 May, 08:00\",\"contact\":\"Facility team, 030 000000\",\"tone\":\"neutral\",\"sender\":\"Property management Berg\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "property": {
            "type": "string",
            "maxLength": 200,
            "minLength": 1
        },
        "topic": {
            "type": "string",
            "maxLength": 200,
            "minLength": 1
        },
        "facts": {
            "type": "string",
            "maxLength": 4000,
            "minLength": 1
        },
        "date_time": {
            "type": "string",
            "maxLength": 80
        },
        "contact": {
            "type": "string",
            "maxLength": 400
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        },
        "sender": {
            "type": "string",
            "maxLength": 600
        }
    },
    "required": [
        "property",
        "topic",
        "facts",
        "tone"
    ]
}

Internal announcement

internal_announcement

Write an announcement for colleagues or members from supplied facts, with date and contact exactly as given.

Input and configuration fields

Essential settings

topic — Topic

What the announcement is about.

Default settings: ""

audience — For whom

Optional: team, department or all members.

Default settings: ""

facts — Facts

Plain notes are enough. Only what is written here can appear in the draft.

Default settings: ""

effective_date — Effective from

Added word for word.

Default settings: ""

contact — Contact details

Who answers questions. Added word for word.

Default settings: ""

sender — Sender and signature

Name, role and contact details. Added word for word.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "internal_announcement",
    "text": "{\"topic\":\"New opening hours of the office\",\"audience\":\"All staff\",\"facts\":\"The office closes at 16:00 on Fridays. Calls go to the answering service after that. Please tell customers about this in your next contact.\",\"effective_date\":\"1 June\",\"contact\":\"Anna Berg, office\",\"tone\":\"neutral\",\"sender\":\"Management, Lumen Print\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "topic": {
            "type": "string",
            "maxLength": 200,
            "minLength": 1
        },
        "audience": {
            "type": "string",
            "maxLength": 160
        },
        "facts": {
            "type": "string",
            "maxLength": 5000,
            "minLength": 1
        },
        "effective_date": {
            "type": "string",
            "maxLength": 40
        },
        "contact": {
            "type": "string",
            "maxLength": 300
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        },
        "sender": {
            "type": "string",
            "maxLength": 600
        }
    },
    "required": [
        "topic",
        "facts",
        "tone"
    ]
}

Work report

work_report

Turn notes and a table of hours and material into a clean work report. The server adds up hours and amounts, and the AI words the work done.

Input and configuration fields

Essential settings

customer — Customer

Added word for word.

Default settings: ""

site — Site

Optional: address or job number.

Default settings: ""

work_date — Date of work

Added word for word.

Default settings: ""

technician — Carried out by

Optional: name or team.

Default settings: ""

work_notes — Notes on the work

What was done, what was found, what is still open. Plain notes are enough.

Default settings: ""

lines — Hours and material

One row per item. kind is labour, material or other. unit_price may be empty.

Default settings: ""

Advanced settings

currency — Currency

Short code or sign shown with amounts, for example EUR.

Default settings: "EUR"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "work_report",
    "text": "{\"customer\":\"Hartmann Bakery\",\"site\":\"Mill Street 4\",\"work_date\":\"12 May 2026\",\"technician\":\"Tom Berg\",\"work_notes\":\"Replaced the leaking valve on the dough mixer water line. Tested for leaks, none found. The second valve shows rust and should be replaced at the next service.\",\"lines\":\"id,kind,description,quantity,unit,unit_price\\nL1,labour,Repair and test,2.5,h,48\\nL2,material,Ball valve DN15,1,pc,18.40\\nL3,other,Travel,1,flat,25\",\"currency\":\"EUR\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "customer": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "site": {
            "type": "string",
            "maxLength": 200
        },
        "work_date": {
            "type": "string",
            "maxLength": 40,
            "minLength": 1
        },
        "technician": {
            "type": "string",
            "maxLength": 160
        },
        "work_notes": {
            "type": "string",
            "maxLength": 6000,
            "minLength": 1
        },
        "lines": {
            "type": "string",
            "maxLength": 20000,
            "minLength": 1
        },
        "currency": {
            "type": "string",
            "maxLength": 4
        }
    },
    "required": [
        "customer",
        "work_date",
        "work_notes",
        "lines"
    ]
}

Defect list

defect_list

Turn inspection notes into a numbered defect list. Every entry is linked to the exact passage in the notes, and nothing is added.

Input and configuration fields

Essential settings

site — Site or property

Added word for word.

Default settings: ""

inspection_date — Date of inspection

Added word for word.

Default settings: ""

inspector — Inspected by

Optional.

Default settings: ""

document — Inspection notes

One defect per line or paragraph. Photo file names may be mentioned.

Default settings: ""

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "defect_list",
    "text": "{\"site\":\"Mill Road 3, second floor\",\"inspection_date\":\"3 June 2026\",\"inspector\":\"A. Novak\",\"document\":\"Kitchen: the socket next to the window has no cover, wires are visible.\\n\\nBathroom: grout around the shower is missing on the left side.\\n\\nHallway: paint on the door frame is scratched, see photo IMG_204.\\n\\nLiving room: the radiator valve does not turn.\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "site": {
            "type": "string",
            "maxLength": 200,
            "minLength": 1
        },
        "inspection_date": {
            "type": "string",
            "maxLength": 40,
            "minLength": 1
        },
        "inspector": {
            "type": "string",
            "maxLength": 160
        },
        "document": {
            "type": "string",
            "maxLength": 20000,
            "minLength": 1
        }
    },
    "required": [
        "site",
        "inspection_date",
        "document"
    ]
}

Employment reference explained

reference_explained

Explain an employment reference sentence by sentence in plain words. Each explanation is linked to the exact sentence.

Input and configuration fields

Essential settings

document — Reference text

Paste the text. It is treated as data, never as instructions.

Default settings: ""

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "reference_explained",
    "text": "{\"document\":\"Ms Kovac worked for us as a project assistant from March 2023 to June 2026.\\n\\nShe carried out the tasks assigned to her with great care and always to our full satisfaction.\\n\\nHer behaviour towards colleagues was friendly and she was always punctual.\\n\\nWe wish her all the best for the future.\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "document": {
            "type": "string",
            "maxLength": 12000,
            "minLength": 1
        }
    },
    "required": [
        "document"
    ]
}

Employment reference draft

reference_draft

Draft an employment reference from supplied facts. Dates, duties and ratings appear only if they are in the input.

Input and configuration fields

Essential settings

employee — Employee

Full name. Added word for word.

Default settings: ""

period — Period of employment

Added word for word, for example 1 March 2023 to 30 June 2026.

Default settings: ""

role — Role

Added word for word.

Default settings: ""

company — Company

Added word for word.

Default settings: ""

facts — Facts

Duties, results, conduct and how they are rated. Only what is written here can appear in the draft.

Default settings: ""

sender — Sender and signature

Name, role and contact details. Added word for word.

Default settings: ""

Advanced settings

tone — Tone

Overall style of the draft.

neutral — Neutral · friendly — Friendly · formal — Formal

Default settings: "neutral"

Maximum UTF-8 input bytes: 30000

Results use ops_review: columns and scalar rows, summary, source-linked findings and draft exports. review_required and ai_required are true, and ai_model is do4u-standard. All text must be rendered as untrusted plaintext. Cashflow results additionally contain chart_data with type cashflow, currency, buffer_cents and up to 91 dated points containing base_cents and scenario_cents. Values are server-calculated integer cents; no model-generated chart markup is accepted. Cutting results contain chart_data with type cutting and scale 1000: bar lengths are integer thousandths of the stated unit (mm, cm, m or in). Each bar contains id, unit, length, trim_start, trim_end, kerf, segments (id and length), and remaining. One kerf is counted per finished piece. The server verifies that pieces, kerfs, end trims and remaining material exactly equal the stock length. The drawing includes at most 30 complete bars and 120 pieces; total_bars, omitted_bars and truncated explicitly describe any omission. The complete allocation remains in the table and draft exports. The heuristic does not claim global optimality.

Request example

{
    "tool": "reference_draft",
    "text": "{\"employee\":\"Mira Kovac\",\"period\":\"1 March 2023 to 30 June 2026\",\"role\":\"Project assistant\",\"company\":\"Lumen Print\",\"facts\":\"Planned print schedules for 40 customers. Trained two new colleagues. Her work was very good and reliable. Colleagues and customers valued her friendly manner. She leaves at her own request.\",\"tone\":\"formal\",\"sender\":\"Tom Berg, Head of Production\"}"
}

{
    "type": "object",
    "additionalProperties": false,
    "properties": {
        "employee": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "period": {
            "type": "string",
            "maxLength": 120,
            "minLength": 1
        },
        "role": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "company": {
            "type": "string",
            "maxLength": 160,
            "minLength": 1
        },
        "facts": {
            "type": "string",
            "maxLength": 8000,
            "minLength": 1
        },
        "tone": {
            "type": "string",
            "enum": [
                "neutral",
                "friendly",
                "formal"
            ]
        },
        "sender": {
            "type": "string",
            "maxLength": 600,
            "minLength": 1
        }
    },
    "required": [
        "employee",
        "period",
        "role",
        "company",
        "facts",
        "tone",
        "sender"
    ]
}

Utility bill check

utility_bill_check

Recalculate a service charge or utility statement from its figures and list questions to ask. The server does the arithmetic.

Input and configuration fields

Essential settings

property — Flat or property

Added word for word.

Default settings: ""

period — Billing period

Added word for word.

Default settings: ""

lines — Costs from the statement

One row per cost type: total_amount for the building, key_total and key_share for the allocation, for example 500 and 62 square metres. previous_total may be empty.

Default settings: ""

advance_paid — Advance payments

Total paid during the period. Empty if unknown.

Default settings: ""

billed_balance — Balance on the statement

Amount to pay according to the statement, for comparison. Empty if unknown.

Default settings: ""

statement_total — Notes from the statement

Optional: anything else stated on the statement, such as the billed balance.

Default settings: ""