How Doc Fetch works
From raw documents to grounded, cited answers in four simple steps. Here's what happens under the hood.
Upload & Organize
Create folders to organize your documents by project, topic, or course. Upload PDFs, text files, and CSVs. Each folder becomes its own isolated knowledge base.
Process & Index
Documents are split into smart chunks, embedded into vectors, and indexed with both semantic (FAISS) and keyword (BM25) search for maximum recall. This hybrid approach ensures no relevant passage is missed.
Ask & Get Cited Answers
Your query is expanded into multiple phrasings for better recall. Results from FAISS and BM25 are fused using Reciprocal Rank Fusion. The LLM generates answers citing specific sources, then a grounding check verifies every claim.
Quiz & Learn
Generate quizzes directly from your documents. Choose multiple choice, true/false, or short answer. Focus on specific topics or get an overview. Instant feedback tells you what you got right and why.
What makes it different
Hybrid Search
Combines semantic understanding (FAISS) with keyword matching (BM25) using Reciprocal Rank Fusion. Catches both meaning and exact terms.
Query Expansion
Your question is automatically rephrased into multiple variants so retrieval covers different angles and phrasings of the same concept.
Grounding Verification
Every claim in the answer is checked against source documents. You see exactly which claims are supported and which aren't.
Ready to try it?
Start for free with no credit card required. Upgrade when you need more.