# Tenets Project Rules
## Project Overview
Tenets is an MCP server, CLI, and Python library for intelligent code context. It helps AI coding assistants understand codebases by finding, ranking, and aggregating relevant code.
## Architecture
### Core Components
- `tenets/core/` - Core logic (NLP, ranking, analysis)
- `tenets/mcp/` - MCP server implementation
- `tenets/cli/` - CLI application (Typer)
- `tenets/models/` - Pydantic data models
- `tenets/storage/` - SQLite session storage
### Key Classes
- `Tenets` in `tenets/__init__.py` - Main API class
- `TenetsMCP` in `tenets/mcp/server.py` - MCP server
- `ContextResult` in `tenets/models/context.py` - Distill output
## Code Style
### Python
- Use type hints for all function parameters and returns
- Docstrings in Google style
- Pydantic for data models
- Lazy imports for heavy dependencies (ML, viz)
### Testing
- pytest for all tests
- Tests in `tests/` mirroring `tenets/` structure
- Use fixtures from `conftest.py`
### Commits
- Conventional commits: feat, fix, docs, test, chore
- 1-2 line messages, human-readable
## Development Commands
```bash
# Install for development
pip install -e ".[dev,mcp]"
# Run tests
pytest
# Run linting
ruff check tenets/
# Type checking
mypy tenets/
# Build docs
mkdocs serve
```
## MCP Server
### Running
```bash
tenets-mcp # stdio transport
tenets-mcp --sse # SSE transport
tenets-mcp --http # HTTP transport
```
### Testing Tools
Use the MCP inspector or call tools directly via Python:
```python
from tenets import Tenets
t = Tenets()
result = t.distill("find auth code")
```
## Important Notes
1. **No hardcoded paths** - Use pathlib and relative paths
2. **Local-first** - No external API calls for core features
3. **Lazy loading** - Heavy imports (ML, viz) are lazy loaded
4. **Version sync** - Keep pyproject.toml and __init__.py in sync
