A simple example of Retrieval-Augmented Generation (RAG) application with basic functionality.
PDFChat-RAG is a modular, containerized application that enables semantic search and chat over PDF documents using RAG. Built with Flask, Celery, Ollama, ChromaDB, and more, it provides an end-to-end pipeline for uploading, and conversing with PDF documents using local language models.
- 📄 Upload PDFs and extract embedded text
- 🔍 Semantic search over vector embeddings using ChromaDB
- 💬 Chat interface for querying document content via RAG
- 🧠 LLM integration using Ollama
- 🧰 Background processing with Celery & Redis
- 🗂️ Metadata storage in MongoDB
| Container | Role |
|---|---|
flask-app |
PDF upload, RAG chat UI/API |
ollama |
Hosts and serves LLM models |
celery |
Runs background tasks |
redis |
Pub/sub and Celery broker |
chromadb |
Stores and queries vector embeddings |
mongodb |
Stores app metadata and user sessions |
git clone https://github.com/darshanz/pdfchat-rag.git
cd pdfchat-rag- NB. Create relevant directories for storing the uploaded files in the host machine and map them to the flask-app docker ocntainer. Similarly map the directories for mongodb and ollama-models directories.
docker-compose up --buildThis will start all containers. Access the app at http://localhost:5000
- Include support for multiple files
- User loging/auth
- Save chat history for given pdf
- Extracts images and tables from the document and Include in chat responses
This project is licensed under the MIT License.