DocMate

Your own LLM agent for generating dataset from documents.

Overview

DocMate is an Autonomous Document Intelligent tool that lets you chat with both structured and unstrured documents in your dataset. You can upload files, ask questions and get structured responses instantly. It support schema generation, data extraction, refinement and dynamic follow-ups. The front-end features voice input, multiple file upload and chat history.

Architecture

Agent Architecture

File Handler

Features and Technical Highlights

• Multi Modal Support
  Works with structure and unstructured files and parsed using unstructured library,
• RAG based workflow
  Uses Retrieval Augmented Geneation to enhance accuracy, retrieving releavant documents chunks before extraction.
• LangGraph
  Conditional graph-based flow for managing complex state transitions.
• FAISS Vector Store
  Similarity check for document and chunk retrieval.
• Dataset Wide Statistics
  Generates statistics of the whole dataset which displays and saves locally.
• Speech-to-Text Input
• Multiple File upload

Getting Started

Follow these steps to setup DOCMATE on your local machine.

Prerequisite

• Tauri and npm
• API keys for Azure STT and Gemini

Installation

1. Clone the repository:

   git clone https://github.com/ItsAbhinavM/DocMate.git
   cd DocMate

2. Install dependencies:

   npm install
   pip install requirements.txt

3. Set up environment variables:

   Create a .env file in the root directory with the necessary API keys:

   AZURE_API_KEY=<YOUR_AZURE_API_KEY>
   GOOGLE_API_KEY=<YOUR_GOOGLE_GEMINI_API_KEY>
   

4. Start the development server:

   npm run tauri build
   fastapi run main.py