Extract structured documentation from AI chat logs.
When building with AI, your decisions get buried in chat logs. chat2docs extracts them into a structured /docs folder.
It's not a summarizer — it's a specification extractor. It finds what's explicitly stated and flags what's missing, with zero hallucination.
# Install
pip install -e .
# Extract
chat2docs extract chat.md --out ./docs --provider openai --model gpt-4o
# Or use local Ollama
chat2docs extract chat.md --provider ollama --model llama3.2- Installation - Setup and dependencies
- Usage Guide - Commands and options
- Philosophy - Design principles and no-hallucination approach
- Architecture - System design and modules
- Extending - Custom extractors and LLM providers
10 structured markdown documents:
/docs
├── 00_product_vision.md # Goals and vision
├── 01_requirements.md # Functional & non-functional
├── 02_system_architecture.md # Architecture and design
├── 03_data_models.md # Entities and schema
├── 04_api_contracts.md # Endpoints and contracts
├── 05_user_flows.md # User journeys
├── 06_security_and_rules.md # Security and rules
├── 07_tech_stack.md # Technology choices
├── 08_coding_guidelines.md # Code standards
└── 09_open_questions.md # Gaps and clarifications
- No Hallucination - Guard engine enforces strict extraction only
- Deterministic - Same input always produces same output (temperature=0)
- Multiple LLMs - OpenAI cloud or Ollama local
- Extensible - Plugin-based extractor system
- Production Ready - Type-safe, well-documented, tested
- OpenAI - GPT-4o, GPT-4 Turbo, GPT-3.5
- Ollama - Local models (Llama 3.2, Mistral, CodeLlama, etc.)
Contributions welcome! See Contributing Guide.
git checkout -b feature/amazing-feature
git commit -m 'Add amazing feature'
git push origin feature/amazing-featureMIT License - see LICENSE file
Learn more: Full documentation | GitHub