Got tired of wiring up vector stores, embedding models, and chunking logic every time I needed RAG. So I built piragi.

  from piragi import Ragi

  kb = Ragi(\["./docs", "./code/\*\*/\*.py", "https://api.example.com/docs"\])

  answer = kb.ask("How do I deploy this?")

That's the entire setup. No API keys required - runs on Ollama + sentence-transformers locally.

What it does:

  - All formats - PDF, Word, Excel, Markdown, code, URLs, images, audio

  - Auto-updates - watches sources, refreshes in background, zero query latency

  - Citations - every answer includes sources

  - Advanced retrieval - HyDE, hybrid search (BM25 + vector), cross-encoder reranking

  - Smart chunking - semantic, contextual, hierarchical strategies

  - OpenAI compatible - swap in GPT/Claude whenever you want

Quick examples:

  # Filter by metadata
  answer = kb.filter(file_type="pdf").ask("What's in the contracts?")

  #Enable advanced retrieval

    kb = Ragi("./docs", config={
     "retrieval": {
        "use_hyde": True,
        "use_hybrid_search": True,
        "use_cross_encoder": True
     }
   })


  # Use OpenAI instead  
  kb = Ragi("./docs", config={"llm": {"model": "gpt-4o-mini", "api_key": "sk-..."}})

Install:

  pip install piragi
  PyPI: https://pypi.org/project/piragi/

Would love feedback. What's missing? What would make this actually useful for your projects?