Projects that keep structure and context
Each project can record a title, code, category, version, author, notes and status. Search by name, filter by category and mark projects as open or closed.
Technical content has its own model
Alongside free-form text, enter structured information and let TICdoc turn it into a consistent part of the report.
- Functional and non-functional requirementsRecord, sort and format every requirement consistently.
- Use cases and actorsDescribe flows and build the related diagram without leaving the project.
- Data templatesDefine text, number, date, list, boolean or image fields and repeat a template for each record.
- Bibliography and glossaryMaintain entries as data and let TICdoc compose the section.
- Automatic indexesBuild contents, image and table lists with page numbers.
- Rich contentInsert images, tables, formulas, code, TICdoc diagrams, PlantUML, links and internal bookmarks.
A local library of code and diagrams
The code and diagram editors are independent of a document, so their content remains available to other projects.
Print, share or continue elsewhere
Direct PDF project export exists in the code but remains disabled while page layout is validated. The stable route is print preview or HTML/RTF export.
AI-assisted documentation from your code
TICdoc can build a complete first draft using your chosen AI provider. It supports OpenAI, Anthropic Claude, Google Gemini and services offering an OpenAI-compatible API, including compatible local servers.
What changes in version 1.2 is where the context comes from. Until now the model only received the project title and notes. You can now attach source folders and TICdoc builds a local index of what they declare, which becomes part of every request.
The model is explicitly told that the code is the source of truth and that it must not invent units, classes or methods absent from the index. Only declarations and signatures are sent, never routine bodies.
Structured sections: requirements, use cases and glossary
A requirement is not a paragraph. TICdoc stores it as a record, with its code, title and description, and the same goes for use cases — with their steps, actors and exceptions — and glossary terms. So in those sections the AI does not write prose: it fills in the records.
The context menu of a functional requirements, non-functional requirements, use cases or glossary section offers “Generate … with AI”. The model replies in a structured format, TICdoc creates the records and then rebuilds the document from them, exactly as if you had typed them yourself. From there you edit, reorder and renumber them with the usual tools.
- Codes (RF-001, RNF-001, CU-001…) are assigned by TICdoc using the format configured in the project, not by the model.
- Use case actors are reused: if “User” already exists in the project, it is not duplicated.
- Generated items are appended after any existing ones, without deleting or altering anything.
- When generating the full document, these sections are filled in a second pass too. Combined with project templates, a new project comes out with its outline, its content, its requirements, its use cases and its glossary.
Documenting a single unit
Beyond generating the whole document, “Document a unit with AI” lets you pick a file from the index and create a section explaining what the unit is for, what each type handles and how it is used. Here the full detail of that unit is sent, including its private part, because the request is precisely to describe how it works inside.
Long generations
- Resume. If generation is interrupted, launch the same option again: it detects the sections left empty and continues with them, without rebuilding the outline or overwriting existing text.
- Retries. Transient provider errors — no connection, per-minute rate limits or a momentary outage — are retried automatically. Running out of credit is not retried, since the answer would be the same.
- Progress and cancellation. A window shows the current section and overall progress, and lets you cancel. Cancellation takes effect once the section in progress finishes; everything done up to then is saved.
- Optimise requests. Enabling this option in Settings sends each section only the project's units and the declarations related to its title, instead of the whole code structure on every request. It greatly reduces consumption per document in exchange for a narrower context.
- 1Configure the providerUnder Settings → AI provider, enter your key, review the URL and choose a model. “Models” retrieves those available.
- 2Attach the codeThe better the folder represents the real project, the more specific the documentation. It can be re-read at any time.
- 3Describe the project wellThe code index says what exists; the title and notes say what it is for. Together they produce the best result.
- 4Always reviewCheck accuracy, sources, privacy and possible costs. AI creates a draft, not a technical validation.
API keys are encrypted with Windows data protection and bound to your account and computer. TICdoc also migrates and encrypts keys stored by previous versions.
Generation sends the title, notes, section names, prompts and — if you have attached source folders — your code structure: unit, type and signature names, to the selected provider. Modifying a section also sends its current HTML and your instruction. For confidential code, configure a local model or a compatible server on your network: nothing then leaves your computer.
What stays local and when information leaves
By default, projects, code and diagrams are saved under C:\Users\Public\Documents\TicDoc, including the database. TICdoc does not require its own account.
Recommended backup: close TICdoc and copy the complete data folder, not only the exported document. This preserves the database and linked resources.
See the folders created during installation