This project is a production-style, modular rebuild of rag-bot-basic — a Retrieval-Augmented Generation (RAG) chatbot that lets you upload and chat with multiple PDFs.
What’s different in this version? We’ve restructured everything to reflect how you'd build a scalable real-world RAG app. The UI and logic remain familiar, but the under-the-hood design is completely revamped.
| Area | Old Project | This Project |
|---|---|---|
| Modularity | All logic in a single file | ✅ Split into logical modules: chat, sidebar, vectorstore, llm, pdf_handler, etc. |
| PDF Parsing | PyPDF2 |
✅ Switched to pypdf (more modern & maintained) |
| Chain Logic | load_qa_chain |
✅ Now uses RetrievalChain with stuff_documents_chain |
| Vector Store | FAISS | ✅ Now uses ChromaDB (with inspection support) |
| Component Rendering | Conditional rendering | ✅ All components rendered but disabled until their dependencies are met |
| Prompt Design | Static QA prompt | ✅ Custom LangChain prompt template with system/human roles |
| UI Features | Same core UI | ✅ Added live vectorstore inspector for developers (developer_mode.py) |
| Error Handling | Minimal | ✅ Improved error handling and edge case feedback |
- 🔌 Choose Groq or Gemini LLMs
- 📚 Upload multiple PDFs
- 💬 Chat interface powered by LangChain retrieval chains
- 🧠 Contextual embeddings with HuggingFace or Google models
- 🧹 Utilities panel: Reset, Clear, Undo
- 📥 Downloadable chat history
- 🧪 ChromaDB Developer Mode for inspecting embeddings
🛠️ Tech Stack
- UI: Streamlit
- LLMs: Groq & Gemini via LangChain
- Vector DB: ChromaDB (was FAISS in old version)
- Embeddings: HuggingFace & Google GenAI
- PDF Parsing: PyPDF
- Orchestration: LangChain Retrieval Chain
git clone https://github.com/Zlash65/rag-bot-chroma.git
cd rag-bot-chroma
python3 -m venv venv
source venv/bin/activate
pip3 install -r requirements.txt- Groq API key from console.groq.com
- Google Gemini API key from ai.google.dev
Create a .env file:
GROQ_API_KEY=your-groq-key
GOOGLE_API_KEY=your-google-keystreamlit run app.py- Choose your model provider (Groq or Gemini)
- Pick a model
- Upload PDFs
- Click Submit
- Ask anything!
📁 Project Structure
.
├── app.py # Main app logic
├── utils/
│ ├── chat_handler.py # Handles chat, input, history, downloads
│ ├── sidebar_handler.py # Handles sidebar config, upload, utilities
│ ├── llm_handler.py # LLM and chain setup
│ ├── vectorstore_handler.py # Embedding + Chroma vectorstore logic
│ ├── pdf_handler.py # PDF parsing and chunking
│ ├── config.py # API keys and model metadata
│ └── developer_mode.py # Inspector for vectorstore queries
├── data/ # Local vectorstore (Chroma) (not committed)
├── assets/ # GIFs and images for README
├── .env # API keys (not committed)
└── requirements.txt
| Button | Function |
|---|---|
| 🔄 Reset | Clears session state and reruns app |
| 🧹 Clear Chat | Clears chat + PDF submission |
| ↩️ Undo | Removes last question/response |
Chat history is saved in the session state and can be exported as a CSV with the following columns:
| Question | Answer | Model Provider | Model Name | PDF File | Timestamp |
|---|---|---|---|---|---|
| What is this PDF about? | This PDF explains... | Groq | llama3-70b-8192 | file1.pdf, file2.pdf | 2025-07-03 21:00:00 |
Check out the original repo here:
👉 rag-bot-basic
Great for understanding the fundamentals before jumping into modularization.