Minerva — Evolutionary RAG

Minerva is an evolutionary retrieval-augmented generation (RAG) engine that uses holographic memory compression and genetic optimization to improve long-term retrieval and answer quality.

Key features

• Holographic memory compression for dense storage
• Evolutionary optimization over knowledge gene pools
• Local LLM integration (Ollama) and extensible retrieval pipeline

Quickstart (developer-friendly, lightweight)

1. Create a venv: python -m venv .venv && .\.venv\Scripts\Activate
2. Install test deps (fast): python -m pip install --upgrade pip && pip install -e . --no-deps && pip install pytest
3. Run unit tests: pytest tests/unit -q

Full installation (CI / integration)

• The project depends on large ML packages (torch, sentence-transformers). To run full integration tests or production workloads, use CI or a machine with sufficient bandwidth and storage: pip install -r requirements.txt

CLI usage

• Ingest data: minerva ingest
• Query: minerva query "What is X?"

Development

• Formatting: black .
• Linting: flake8 .
• Type checks: mypy .

Pre-commit

• Install pre-commit and enable hooks locally: pip install pre-commit pre-commit install

CI and releases

• CI installs test deps with --no-deps by default to avoid heavy ML downloads.
• Use the Integration workflow (manual trigger) to run full integration tests and build/publish artifacts.
• Format-check workflow enforces isort/black/flake8 on PRs.

Contributing

• See CONTRIBUTING.md for details and guidelines.

Publishing & Releases

• GitHub Packages: Install via pip install --index-url https://pypi.pkg.github.com/NailaRais minerva-rag==0.1.2
• Releases are created automatically on tag push (e.g., git tag v0.1.2 && git push --tags)
• Artifacts (wheel, sdist) are available in GitHub Releases

License

• See LICENSE file.

Contact

• Author: Naila Rais naila123rais@gmail.com