AI · Retrieval experiment
Llama 2 GGML Medical Chatbot
An experiment in answering questions from medical reference documents with a quantized language model.
Overview
An experiment in answering questions from medical reference documents with a quantized language model.
Built the retrieval workflow, model integration and chatbot interface using Python and LangChain.
The problem
Explore document-grounded question answering using local model inference and retrieved source passages.
My contribution
Built the retrieval workflow, model integration and chatbot interface using Python and LangChain.
PythonLlama 2LangChainFAISSChainlit
Inside the build
- Retrieve document passages using sentence-transformer embeddings and FAISS.
- Pass retrieved context to a pretrained Llama 2 model and return source documents with the answer.
Project notes
Research prototype
This is a document-retrieval experiment, not a clinical system or a substitute for professional medical advice. It uses a pretrained model; no model-training or clinical-accuracy claim is made.