Skip to content

README.md

Latest commit

 

History

History
304 lines (232 loc) · 8.32 KB

README.md

File metadata and controls

304 lines (232 loc) · 8.32 KB
Raw

Dynamic Agentic System

A comprehensive multi-agent system for document processing, mathematical computation, and database querying using FastAPI, LangGraph, and LangChain.

Features

  • Multi-Agent Orchestration: Uses LangGraph to route queries between different specialized agents
  • Document RAG: Pinecone vector store for document retrieval and search
  • Math Computation: Pandas-based mathematical and statistical analysis
  • SQL Query Engine: DuckDB integration for database operations
  • PDF Processing: PyMuPDF and OCR for document text extraction
  • Persona System: Financial, Legal, and General Assistant personas
  • Suggested Queries: AI-powered follow-up question generation
  • FastAPI Backend: RESTful API with automatic documentation
  • Next.js Frontend: Modern React-based user interface

Project Structure

dynamic_agentic_system/
├── agents/                 # Agent personas and management
│   ├── __init__.py
│   └── personas.py        # Financial, Legal, General personas
├── rag/                   # Retrieval-Augmented Generation
│   ├── __init__.py
│   └── document_store.py  # Pinecone vector store integration
├── math_ops/              # Mathematical computations
│   ├── __init__.py
│   ├── computation.py     # Pandas-based calculations
│   ├── data_processor.py  # CSV data processing
│   └── sql_query.py       # DuckDB database operations
├── ocr/                   # PDF processing and OCR
│   ├── __init__.py
│   └── pdf_processor.py   # PyMuPDF and Tesseract integration
├── router/                # Agent orchestration
│   ├── __init__.py
│   ├── agent_router.py    # LangGraph workflow
│   └── suggested_queries.py # Follow-up query generation
├── api/                   # FastAPI application
│   ├── __init__.py
│   └── main.py           # Main API endpoints
├── frontend/              # Next.js frontend application
│   ├── app/              # Next.js app directory
│   ├── components/       # React components
│   ├── lib/              # Utility functions
│   ├── types/            # TypeScript type definitions
│   └── package.json      # Frontend dependencies
├── data/                  # Data storage
│   ├── docs/             # Uploaded PDF documents
│   └── stocks/           # Stock data files
├── config.py             # Configuration and environment variables
├── requirements.txt      # Python dependencies
├── env_example.txt       # Environment variables template
├── run.py               # Main server entry point
└── README.md            # This file

Prerequisites

  • Python 3.8+
  • OpenAI API key
  • Pinecone API key and environment
  • Tesseract OCR (optional, for enhanced PDF processing)

Installation

  1. Clone the repository:

    git clone <repository-url>
cd dynamic_agentic_system

  2. Create a virtual environment:

    python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

  3. Install dependencies:

    pip install -r requirements.txt

  4. Set up environment variables:

    cp env_example.txt .env
# Edit .env with your API keys and configuration

  5. Install Tesseract OCR (optional):

Configuration

Create a .env file with the following variables:

# OpenAI Configuration
OPENAI_API_KEY=your_openai_api_key_here

# Pinecone Configuration
PINECONE_API_KEY=your_pinecone_api_key_here
PINECONE_ENVIRONMENT=your_pinecone_environment_here
PINECONE_INDEX_NAME=dynamic-agentic-system

# Database Configuration
DATABASE_URL=sqlite:///./data/system.db

# Server Configuration
HOST=0.0.0.0
PORT=8000
DEBUG=True

# OCR Configuration
TESSERACT_CMD_PATH=/usr/bin/tesseract

# File Upload Configuration
MAX_FILE_SIZE=10485760  # 10MB
UPLOAD_DIR=./data/docs

Usage

Starting the Server

cd api
python main.py

The server will start at http://localhost:8000

API Endpoints

1. Process Query

POST /query
Content-Type: application/json

{
  "query": "What are the key findings in the uploaded documents?",
  "persona_type": "financial",
  "context": "Additional context if needed"
}

2. Upload PDF

POST /upload
Content-Type: multipart/form-data

file: [PDF file]

3. Get Available Personas

GET /personas

4. System Status

GET /status

5. Document Statistics

GET /documents/stats

6. Database Tables

GET /database/tables

Example Usage

Financial Analysis Query

import requests

# Query with financial persona
response = requests.post("http://localhost:8000/query", json={
    "query": "Analyze the stock performance data and calculate the average return",
    "persona_type": "financial"
})

print(response.json())

Document Upload and Query

# Upload PDF
with open("document.pdf", "rb") as f:
    files = {"file": f}
    upload_response = requests.post("http://localhost:8000/upload", files=files)

# Query the uploaded document
query_response = requests.post("http://localhost:8000/query", json={
    "query": "What are the main points in the uploaded document?",
    "persona_type": "general"
})

Agent Personas

Financial Advisor

  • Stock market analysis and investment strategies
  • Financial planning and portfolio management
  • Economic trends and market indicators
  • Risk assessment and mitigation

Legal Advisor

  • Contract law and legal document analysis
  • Corporate law and business regulations
  • Intellectual property and compliance
  • Risk assessment and legal implications

General Assistant

  • General knowledge and research
  • Problem-solving and analysis
  • Writing and communication
  • Technology and tools

Workflow

  1. Query Classification: The system analyzes the query to determine the appropriate processing path
  2. Document Search: Searches uploaded PDFs using vector similarity
  3. Math Computation: Performs mathematical calculations and statistical analysis
  4. Database Query: Executes SQL queries on available data
  5. Response Generation: Uses the selected persona to generate a comprehensive response
  6. Query Suggestions: Generates relevant follow-up questions

Development

Adding New Personas

  1. Create a new persona class in agents/personas.py
  2. Add it to the PersonaManager class
  3. Update the PersonaType enum

Adding New Computation Types

  1. Extend the MathComputation class in math/computation.py
  2. Add new patterns to the query classification in router/agent_router.py

Adding New Database Operations

  1. Extend the SQLQueryEngine class in math/sql_query.py
  2. Add new table schemas and sample data

Troubleshooting

Common Issues

  1. Missing API Keys: Ensure all required API keys are set in the .env file
  2. Pinecone Index: The system will create the Pinecone index automatically on first run
  3. Tesseract OCR: If OCR fails, check the Tesseract installation path in the configuration
  4. File Upload: Ensure the upload directory exists and has proper permissions

Logs

Check the console output for detailed error messages and system status information.

Contributing

  1. Fork the repository
  2. Create a feature branch
  3. Make your changes
  4. Add tests if applicable
  5. Submit a pull request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Support

For support and questions, please open an issue on the GitHub repository.

Frontend Setup

Starting the Frontend

cd frontend
npm install
npm run dev

The frontend will start at http://localhost:3000

Frontend Features

  • Modern UI: Built with Next.js 14, Tailwind CSS, and ShadCN components
  • Real-time Chat: Interactive chat interface with persona selection
  • File Upload: Drag-and-drop PDF upload functionality
  • Responsive Design: Works on desktop and mobile devices
  • Dark Mode: Toggle between light and dark themes

License

This project is licensed under the MIT License - see the LICENSE file for details.