AI · Retrieval experiment
Llama 2 GGML CSV Chatbot
Ask questions of uploaded CSV data using local model inference and vector retrieval.
Overview
Ask questions of uploaded CSV data using local model inference and vector retrieval.
Built a Streamlit prototype combining CSV ingestion, embedding search and a quantized Llama 2 model.
The problem
Exploring tabular data through natural language requires grounding responses in the uploaded file.
My contribution
Built a Streamlit prototype combining CSV ingestion, embedding search and a quantized Llama 2 model.
PythonLangChainStreamlitLlama 2FAISS
Inside the build
- Chunk CSV data and index it with sentence-transformer embeddings and FAISS.
- Retrieve relevant rows for a Llama 2 question-answering workflow.
Project notes
Research prototype
Llama 2 GGML CSV Chatbot is part of my work in ai · retrieval experiment. The links above provide the available project and implementation details.