Hybrid RAG Lab Demo 01

Upload a small .txt, .pdf, or .docx document, choose a retrieval technique, ask a query, and inspect both the ranked evidence and the generated grounded answer.

Demo 1 is intentionally lightweight:

  • best with 1 short document
  • supports up to 3 files
  • retrieval works without external keys
  • generation can use the local fallback mode or a Hugging Face LLM if HF_TOKEN is configured
Retrieval Technique
Generation Mode
Hugging Face mode uses CohereLabs/tiny-aya-global:cohere and requires HF_TOKEN.
1 5

Loaded Documents

Ranked Retrieval Results