📚 RAG Q&A Assistant with Multi-Document Support
Upload your documents and ask questions using semantic search + local LLM. Powered by: FAISS + Local Transformers (SmolLM2-135M) + FastEmbed
Upload Your Documents
Upload one or more documents (PDF, DOCX, TXT, MD) to index and search. The system will:
- Process each document
- Split into semantic chunks
- Generate embeddings
- Save to vector store for future queries
No documents uploaded yet
Query Your Documents
Ask questions about the documents you've uploaded. The system will retrieve relevant chunks and generate answers with source citations.
Your answer will appear here... (Make sure to upload documents first!)
ℹ️ How It Works
- Query Processing: Your question is converted to embeddings
- Semantic Search: Top-K most relevant chunks are retrieved from documents
- LLM Generation: Local transformers model (SmolLM2-135M) generates answer based on retrieved chunks
- Source Citation: Answer includes [1], [2] citations linked to source documents
🚀 Performance
- First run: ~5 minutes (document processing + embedding generation)
- Subsequent queries: <5 seconds (embeddings cached)
- Completely private & offline (when using local models)