https://github.com/anupam0202/contextual-rag-chatbot
Contextual RAG Chatbot that processes PDF documents using the Google Gemini API
https://github.com/anupam0202/contextual-rag-chatbot
google-generativeai numpy pypdf2 scikit-learn streamlit
Last synced: 5 months ago
JSON representation
Contextual RAG Chatbot that processes PDF documents using the Google Gemini API
- Host: GitHub
- URL: https://github.com/anupam0202/contextual-rag-chatbot
- Owner: Anupam0202
- Created: 2025-02-23T09:21:10.000Z (11 months ago)
- Default Branch: main
- Last Pushed: 2025-08-10T14:22:20.000Z (6 months ago)
- Last Synced: 2025-08-10T16:23:08.526Z (6 months ago)
- Topics: google-generativeai, numpy, pypdf2, scikit-learn, streamlit
- Language: Python
- Homepage: https://contextual-rag-chatbot.streamlit.app/
- Size: 42 KB
- Stars: 1
- Watchers: 1
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Contextual RAG Chatbot (Gemini 2.5)
An advanced, Streamlit-based Retrieval-Augmented Generation (RAG) chatbot that analyzes multiple PDF documents and answers questions with grounded, context-aware responses powered by Google's Gemini 2.5.
## Features
- Multi-document PDF ingestion (up to 100MB each)
- Fast chunking + embedding-based retrieval
- Contextual conversation memory with configurable context window
- Gemini 2.5 responses with streaming output and markdown formatting
- Source references for retrieved passages
- Advanced UI with avatars, status chips, and analytics dashboard
- Secure API key management via `.env`
## Demo (What You’ll See)
- Left sidebar: configuration (model parameters, chunking settings) and document upload
- Main chat area: user/assistant messages with avatars and streaming responses
- Right panel: document overview and analytics (chunks per document, file size distribution, chunk size histogram)
## Requirements
- Python 3.9 – 3.11 recommended
- Google Gemini API key
## Quick Start
1) Clone this repository and enter the project directory.
2) Create a virtual environment and activate it:
```
python -m venv venv
venv\Scripts\activate # Windows
# source venv/bin/activate # macOS/Linux
```
3) Install dependencies:
```
pip install -r requirements.txt
```
4) Create a `.env` file in the project root with your Gemini API key:
```
GEMINI_API_KEY=your_gemini_api_key_here
```
5) Run the app:
```
streamlit run app.py
```
Open the provided local URL in your browser.
## Usage
- In the sidebar, upload one or more PDF files (each ≤ 100MB).
- Click “Process Documents” to extract text, chunk, and embed.
- Ask questions in the chat input. The assistant streams its answer.
- Expand “Source References” below an answer to see relevant passages grouped by document.
## Configuration
- Temperature: 0.0–1.0 (creativity vs determinism)
- Context Window: how many past turns to consider
- Chunk Size / Overlap: controls chunking granularity and continuity
- Top K Results: how many most relevant chunks feed RAG
These are set in `app.py` and saved in Streamlit session state.
## How It Works (Architecture)
- PDF ingestion: `PyPDF2` reads pages in batches with progress tracking
- Chunking: sentence-aware splitting with configurable size/overlap
- Embeddings: `google-generativeai` embedding model `models/embedding-001`
- Retrieval: cosine similarity over in-session vector store
- Generation: `gemini-2.5-pro` with a RAG-oriented system instruction, markdown responses, and streaming
- UI: Streamlit chat with avatars, analytics using Plotly
## Environment and Security
- The API key is loaded from `.env` using `python-dotenv` and never shown in the UI.
- The app will warn in the sidebar if `GEMINI_API_KEY` is missing and disable processing/chat.
## Troubleshooting
- Streaming error / empty response:
- Ensure your key is valid and has access to Gemini 2.5.
- Re-run with a fresh environment; check network stability.
- PDF parsing issues:
- Some scanned PDFs may have no extractable text. Consider OCR preprocessing.
- Memory / large loads:
- Reduce `chunk_size`, `top_k_results`, or process fewer PDFs at a time.
## Extending the App
- Swap vector store: integrate FAISS/Chroma/Pinecone for persistence at scale.
- Add observability: integrate LangSmith/TruLens for evaluation and tracing.
- Add authentication: gate the UI with Streamlit auth or a reverse proxy.
## Project Structure
```
rag-chatbot/
app.py # Main Streamlit app
requirements.txt # Python dependencies
venv/ # Virtual environment (optional)
```
## License
This project is provided as-is, without warranties. Adapt and extend for your use case.