aihpi/workshop-rag

Implements multiple advanced RAG strategies—Multi Query, RAG Fusion, Decomposition, and HyDE—each using distinct query reformulation or document generation techniques to improve retrieval coverage and ranking. Built on sentence embeddings and vector databases, the tool provides a modular, locally-runnable interface powered by Ollama for on-device LLM inference, with Gradio for the web UI and configurable retrieval/generation parameters like relevance thresholds and temperature control.