# t0-1

t0-1 is a research demonstration from the Alan Turing Institute's t0 initiative, applying retrieval-augmented reasoning with lean language models to a domain-specific knowledge base built from NHS condition pages.

t0-1 is an example application from the Alan Turing Institute's t0 research initiative, which develops lean but capable language models. The repository shows retrieval-augmented reasoning applied to a specific body of knowledge: text scraped from NHS A-to-Z condition pages, converted into JSONL records that pair a condition title with its content.

Setup uses a Python virtual environment managed with uv, and the retrieval-augmented model can be served through several scripts: a FastAPI endpoint for conversational RAG, a vLLM endpoint for the project's own fine-tuned model, and a vLLM endpoint for a Qwen model with tool calling. Azure endpoint credentials are required for serving.

The project is a Jupyter Notebook and Python codebase released under the MIT license.

Website: https://github.com/alan-turing-institute/t0-1