Chat With Data (LangChain + RAG)

Ask complex questions about your CSV/documents and get grounded answers—runs locally with Ollama.

Overview

CSV RAG Analyst is a lightweight Retrieval-Augmented Generation system for structured data. It pairs semantic retrieval with LLM-generated pandas code so you can ask natural-language questions like “Wie viel Umsatz hatten wir im Februar 2025?” and get precise, reproducible answers. Runs fully offline via Ollama.

Key Features

Conversational Q&A over your CSVs; English or German.
Auto-generated and executed pandas queries (filters, groups, dates).
Streamlit web UI option; no cloud APIs—100% local.

Architecture

Loader → chunking with overlap → embeddings → vector store (e.g., Chroma/FAISS).
Retriever + LLM → validated pandas code → execution → formatted results.

Run Locally

Install ollama and pull a local model (e.g., gemma3:27b).
pip install -r requirements.txt
streamlit run app.py then open http://localhost:8501.

Repository

GitHub – Chat With Data (LangChain + RAG)

Roadmap

Dynamic CSV upload
Role-based access & per-department memory
Long-form explanations & visualizations