DocuMind LLM is a Generative AI-powered document assistant built with Hugging Face Transformers, FAISS, and LangChain.
It allows users to upload PDF files, intelligently index their contents, and ask natural language questions about the document.
- 📄 PDF Upload & Parsing — Extracts text and chunks it for semantic understanding.
- 🤖 LLM-powered Q&A — Uses a Transformer model (e.g.,
mistralai/Mixtral,google/flan-t5, etc.) to answer questions. - ⚡ FAISS-based Vector Search — Enables fast and accurate document retrieval.
- 💬 Conversational Memory — Keeps track of your recent queries for context-aware responses.
- 🧩 Modular Architecture — Easy to extend with other models, vector stores, or APIs.
| Component | Technology |
|---|---|
| Embeddings | Hugging Face Sentence Transformers |
| Vector Store | FAISS |
| LLM | Hugging Face Transformers |
| Interface | Streamlit / Flask |
| Backend | Python 3.10+ |
git clone https://github.com/ramarav/DocuMind-LLM.git
cd DocuMind-LLM
pip install -r requirements.txtpython app.pyChoose any .pdf document you want to query.
Type natural language questions like:
“What are the main topics covered in this document?”
“Summarize section 3.”
“What are the key takeaways?”
- Research paper summarization
- Legal contract question answering
- Technical documentation assistant
- Corporate report analysis
- AI-based knowledge discovery
DocuMind-LLM/
│
├── app.py # Main entry point
├── utils/ # Helper scripts
│ ├── pdf_loader.py
│ ├── embedder.py
│ ├── vector_store.py
│ └── qa_engine.py
├── sample.pdf # Example document
├── requirements.txt
└── README.md
- Add chat history memory using LangChain.
- Integrate OpenAI API for comparison.
- Enable multi-file document search.
- Add semantic summarization features.
This project is licensed under the MIT License.