chat-with-pdf

an pdf chat AI Agent made using Smolagents Framework

Python ⭐ 0

README

A conversational AI agent that lets you upload any PDF and chat with its contents!
Built with smolagents, Google Gemini, pypdf, and Gradio.

Layer	Technology	Purpose
LLM Orchestration	smolagents	Agent framework, tool integration
Language Model	Google Gemini 1.5 Flash	Large Language Model for Q&A
LLM API Adapter	LiteLLM	Unified LLM API interface
PDF Processing	pypdf	Extracts text from uploaded PDF files
Web UI	Gradio	Interactive chat interface
Environment Mgmt	python-dotenv	Loads environment variables (local dev)
Deployment	Hugging Face Spaces	Cloud hosting (public demo)

Upload any PDF and extract its text instantly.
Ask questions about the document—get answers powered by Google Gemini (1.5 Flash).
Conversational interface using Gradio.
Runs locally or on Hugging Face Spaces.

git clone https://github.com/KunalGupta25/chat-with-pdf.git
cd chat-with-pdf

pip install -r requirements.txt

Get your key from Google AI Studio.
Create a .env file in the project root:
```
GEMINI_API_KEY=your_actual_key_here
```

python app.py

Open http://localhost:7860 in your browser.

Go to your Space: https://huggingface.co/spaces/your-username/your-space-name
Click "Duplicate Space" to make your own copy, or use it directly if public.
Add your GEMINI_API_KEY as a secret in the Space settings (if you duplicate or deploy privately).

Model: Uses "gemini/gemini-1.5-flash" by default. If you have access to "gemini/gemini-1.5-pro", change the model_id in app.py.
API Key: Must be a Google AI Studio key, not a Vertex AI or GCP key.
No api_base needed for Gemini AI Studio keys.

Upload a PDF: The app extracts all text using pypdf.
Ask a question: Your question and the extracted text are sent to the Gemini model via smolagents.
Get answers: The agent uses Gemini to answer your question, referencing the PDF content.

Add more tools: Use the @tool decorator from smolagents to add custom functions.
Customize the UI: Edit the Gradio blocks in app.py.
Chunking or RAG: For large PDFs, consider splitting text into chunks and using retrieval-augmented generation.

MIT License.
See LICENSE for details.

Enjoy chatting with your PDFs!
For questions or contributions, open an issue or pull request.