clm-system/AGENTS.md

CLM System - Agent Guidelines

Important Notes

• Always use uv with --active flag for dependency management
• Read docs from context7 whenever in doubt or needs confirmation on how to do things the right way

Build/Run Commands

# Install dependencies
uv add --active streamlit langchain langchain-community pypdf2 python-docx pytesseract lancedb

# Run Streamlit app
streamlit run app.py

# Manual scan
python scripts/manual_scan.py

# Generate reports
python scripts/generate_reports.py

Code Style

• Framework: Streamlit + LangChain + LanceDB
• Structure: Monolithic with modular components in src/
• Imports: Standard library first, then third-party, then local modules
• Naming: snake_case for functions/variables, PascalCase for classes
• Error Handling: Use try/except blocks with logging to Logger singleton
• Types: Use type hints where beneficial, focus on readability

Key Patterns

• Document Processing Pipeline: FileValidator → OCRProcessor → TextExtractor → Chunker → Embedder → VectorStore
• Singletons: ConfigurationManager, VectorDatabaseConnection, Logger
• Strategy Pattern: ChunkingStrategy (basic fixed-size), EmbeddingModel (single model)
• Direct File Operations: Simple utility functions for file I/O

Testing

# Run basic tests
python -m pytest tests/

# Test single component
python -m pytest tests/test_ingestion.py -v

Linting and Type Checking

# Run ruff linter (auto-fix issues)
ruff check --fix .

# Run pyright type checker
pyright

# Run both after making changes
cd clm-system && ruff check --fix . && pyright

Vector DB Choice

Use LanceDB - lightweight, local, no server setup required for this scope

STRICT RULES

• Do not make sys.path.append fixes to any code. Always understand where you are executing codes from.
• Do not make use of pathlib or os.path always use importlib.resources and define resources in pyproject.toml.