Aller au contenu
Jefferson Gonçalves
← Retour aux projets

Ai Docs

#ai.docs

ai.docs is a structured, documentation-first system designed to guide AI-assisted software development from intent to execution — without ambiguity.

It provides a clear, ordered documentation flow that helps humans and AI models collaborate reliably when designing, planning and building software systems.

#Why this exists

AI-assisted development often fails not because of code quality, but because of missing definitions:

  • unclear scope
  • undefined data models
  • vague API contracts
  • undocumented decisions
  • UX disconnected from rules and permissions

ai.docs addresses this by enforcing:

  • explicit schemas and contracts
  • a strict execution order
  • separation between intent, constraints and implementation
  • traceable and reviewable decisions

The result is a system that can be executed, not guessed.

#What this repository is

This repository is a startkit / reference implementation of the ai.docs methodology.

It demonstrates how to structure and use an ai.docs/ directory as the canonical source of truth for:

  • product intent
  • requirements
  • architecture
  • security constraints
  • execution rules

It is not a code boilerplate.
It is a documentation and reasoning framework designed to be embedded into real projects.

The example used throughout the documentation represents a small system (e.g. a mini CRM), but the methodology applies to projects of any size.

#How to use

  1. Clone or copy this repository into your project.
  2. Ensure the ai.docs/ directory exists at the root of your project.
  3. Follow the documentation in order, starting from:
  4. Complete each document before moving to the next.
  5. Treat the documentation as a contract, not as optional notes.

The numbered files inside ai.docs/ represent the canonical execution order.

#Documentation flow (canonical order)

All documents below live inside the ai.docs/ directory:

  1. 01_README_AI.md — Operational contract for AI / LLMs
  2. 02_CONTEXT.md — Purpose, scope and boundaries
  3. 03_PRD.md — Product requirements and rules
  4. 04_ARCHITECTURE.md — System architecture
  5. 05_SECURITY_AND_COMPLIANCE.md — Security and compliance constraints
  6. 06_BOILERPLATE_STRUCTURE.md — Project structure and conventions
  7. 07_API_CONTRACTS.md — API contracts and versioning
  8. 08_DATA_SCHEMA.md — Data models and relationships
  9. 09_DECISIONS_ADR.md — Architectural decisions and rationale
  10. 10_UX_UI_INDEX.md — UX/UI flows, permissions and routes
  11. 11_ROADMAP.md — Future evolution
  12. 12_RELEASE_CHECKLIST.md — Release and delivery gate

#CI Documentation Gate

To ensure our living contracts and diagrams do not become outdated (a "guess generator"), we have integrated a CI Gate in GitHub Actions.

Before any merge into the main branch, it is mandatory to:

  • Fill out the Pull Request Template, checking whether the Documentation was updated (covering docs/c4, docs/adr, docs/prd, docs/api, and docs/schema) or if the change does not require documentation.
  • The Docs CI Gate workflow will automatically fail and block the merge if no documentation option is validated in the Pull Request body.

We are starting "small", focusing on traceability and expanding as the pain appears.

#Important principles

  • Documentation comes before code.
  • If something is not defined, it must not be implemented.
  • AI models must not invent schemas, rules or behavior.
  • Decisions must be recorded, not remembered.
  • UX is a contract, not decoration.

#Contributions

This repository is open to improvements and refinements.

If you contribute:

  • respect the canonical execution order
  • keep examples consistent
  • update related documents when changing templates

Any structural or behavioral change must be reflected inside ai.docs/.

#License

Open-source. Use, adapt and improve freely.

Nouvelle version disponible.