alan-turing-institute/t0-1

Implements a modular retrieval-augmented generation pipeline with vector store (Chroma/FAISS), dense retriever, and LLM components served via FastAPI, supporting multi-model configurations through environment variables. Demonstrates the architecture on NHS health condition data with synthetic query generation and RAG evaluation capabilities. Integrates vLLM for model serving and supports both the custom t0-1.1 models and tool-calling with Qwen2.5, leveraging Azure AI Foundry endpoints for inference.