Lightweight RAG benchmarking and retrieval ops that’s headed toward a portable ECS container. It’s more than a dashboard: think of it as a briefcase you can run in an isolated environment, with minimal UI layers and a production-minded backend.
- Bench and debug retrieval quality fast (Recall@k, MRR, RAGAS).
- Generate reproducible indexes you can ship and audit.
- Keep the UI thin while the core stays container-ready for prod systems.
- Ingest PDFs/Markdown, chunk with rich metadata.
- Hybrid retrieval (BM25 + vector), inspect scores and provenance.
- Batch evals with exportable artifacts.
ragops is moving from “demo dashboard” toward a deployable ECS unit you can drop into larger systems. The UI is a window, not the product.
pip install -e .
python -m ragops.core.harness