Skip to content

README.md

Latest commit

 

History

History
78 lines (55 loc) · 3.64 KB

README.md

File metadata and controls

78 lines (55 loc) · 3.64 KB

NVIDIA-NIM Chatbot with FAISS and LangChain Integration

Overview:

The NVIDIA-NIM Chatbot is a Streamlit-based application designed to enable advanced question-answering using NVIDIA's Llama 3.1 Nemotron-70B model. This project integrates LangChain for seamless document processing, FAISS for efficient vector storage and retrieval, and NVIDIA embeddings for high-performance vectorization.

The chatbot processes PDF documents, builds a searchable vector database, and delivers accurate responses based on the provided context. It also includes a feature for displaying document similarity, making it a valuable tool for research, education, and enterprise-level document analysis. In this project I have added a book on CCNA guide, therefore this helps me in clearing my querries in computer networks

Screenshot 2024-11-18 at 1 59 10 PM

Features:

  • NVIDIA Embeddings: Utilizes NVIDIA's advanced embeddings for vectorization.
  • FAISS Vector Store: Efficiently indexes and retrieves relevant document chunks.
  • Document Loader: Automatically processes and parses PDF documents from a directory.
  • Custom Text Splitting: Handles long documents with chunking and overlap strategies for optimal retrieval.
  • Interactive Chatbot Interface: Streamlit-powered interface for a user-friendly experience.
  • Context-Based Responses: Provides answers strictly within the provided context.
  • Document Similarity Search: Displays related document sections for enhanced transparency.

CHatbot1-drawio

How It Works:

--Document Embedding: Loads PDFs from the ./books directory. Splits documents into smaller chunks using LangChain's RecursiveCharacterTextSplitter. Embeds the chunks using NVIDIA's embeddings and stores them in a FAISS vector database.

--Retrieval Chain: Retrieves relevant document chunks using FAISS. Feeds the chunks into a LangChain stuff document chain. Uses NVIDIA's Llama 3.1 model to generate a response based on the context.

--Interactive Q&A: Users input questions through the Streamlit interface. The application retrieves the most relevant context and generates an answer.

Getting Started:

##Prerequisites

  • Python 3.8 or higher
  • Dependencies: Install using the provided requirements.txt file.
  • NVIDIA API Key: Obtain an API key from NVIDIA's AI endpoints.

Code Structure:

  • app.py: Main Streamlit application.
  • books/: Directory for storing PDF files to be processed.
  • requirements.txt: List of dependencies required for the project.
  • .env: Configuration file for environment variables (API key).

Customization:

  • Modify the chunk_size and chunk_overlap in the RecursiveCharacterTextSplitter configuration for different document types.
  • Adjust the LangChain prompt template in the ChatPromptTemplate for specific use cases.
  • Add additional preprocessing steps for documents if required.

Future Enhancements:

  • Add support for additional file formats (e.g., DOCX, TXT).
  • Enhance scalability for larger datasets using distributed FAISS.
  • Implement user authentication and role-based access.
  • Add GPU acceleration for faster embeddings and inference.

Contributing: Contributions are welcome! Please open an issue or submit a pull request with detailed information about your changes.

License: This project is licensed under the MIT License.

Acknowledgments: NVIDIA for their Llama 3.1 Nemotron-70B model and embeddings. LangChain for its robust document processing tools. FAISS for efficient vector search capabilities. Streamlit for an intuitive UI framework.