MultiDoc-AI-Assistant

Live Link: https://multidoc-ai-assistant-project.streamlit.app/

MultiDoc-AI-Assistant is an intelligent application that allows you to build a knowledge base from various document types (PDFs, CSVs, JSON, websites, handwritten notes/images) and then chat with this knowledge base using a conversational AI. It leverages Retrieval Augmented Generation (RAG) to provide contextually relevant answers based on your uploaded sources.

Features ✨

• Multi-Source Ingestion: Upload and process various file types:
  • PDFs (text-based and scanned/image-based via OCR)
  • CSV files
  • JSON files
  • Images (PNG, JPG, JPEG - for OCR)
  • Handwritten notes (PDFs or images - for OCR)
  • Website URLs
• Intelligent Chat Interface: Ask questions and receive answers grounded in the content of your uploaded documents.
• Source Referencing: Assistant's responses can include references to the source documents used to generate the answer.
• Dynamic Knowledge Base: Each processing action creates a fresh, isolated knowledge base for your chat session.
• Rate Limiting: Basic rate limiting for OCR operations to manage resource usage.

Tech Stack 🛠️

This project utilizes a modern stack for document processing, AI, and web application development:

• Backend & Application Logic:
  • Python
  • Streamlit: For the interactive web application interface.
• Large Language Model (LLM) & Orchestration:
  • Langchain: Framework for developing applications powered by language models.
    • ConversationalRetrievalChain: For implementing the RAG pattern.
    • ConversationBufferMemory: To maintain chat history.
  • Groq API (Llama 3): For fast LLM inference.
• Document Processing & Text Extraction:
  • PyPDF2: For extracting text from text-based PDFs.
  • pdf2image & Poppler: For converting PDF pages to images.
  • Google Cloud Vision API: For Optical Character Recognition (OCR) on images and scanned PDFs.
  • Beautiful Soup: For parsing and extracting text from website URLs.
  • Pandas: For handling CSV data.
• Vector Store & Embeddings (RAG Core):
  • ChromaDB: As the vector database to store document embeddings.
  • Sentence Transformers (BAAI/bge-small-en-v1.5): For generating text embeddings.
  • pysqlite3-binary: To ensure SQLite compatibility for ChromaDB in cloud environments.
• Deployment:
  • Streamlit Cloud