Agentic AI Project 🚀

In Agentic AI Exploration, I designed 2 multi-agent systems that leverages the Groq language model alongside specialized tools to perform tasks such as financial data analysis, web searches, and PDF document assistance. The system integrates multiple APIs and libraries—including DuckDuckGo for web searches, yfinance for financial data, and Hugging Face for PDF embeddings—to provide comprehensive solutions.

Project Overview

Agents

• Multi-Agent Setup
  The project features a coordinated multi-agent architecture comprising two main sub-agents:

  • Web Agent

    • Utilizes the DuckDuckGo tool to fetch online information.
    • Configured to always include source citations in its output.

  • Financial Agent

    • Retrieves financial data such as stock prices, analyst recommendations, and company news using the YFinance tool.
    • Displays data in formatted tables.
    • (Refer to the Financial Agent Diagram above for a visual overview.)

• PDF Assistant Agent
  Processes PDF documents by:

  • Retrieving content from specified URLs.
  • Embedding the information using a Hugging Face model.
  • Storing the embedded data in a PostgreSQL vector database for interactive querying.

---

Directory Structure

Yash8745-Agentic-Ai/
├── README.md
├── financial_agent.py
├── pdf_assistant.py
├── requirements.txt
└── test.py

Architecture Explanation

Financial Agent Architecture

The Financial Agent is designed to deliver comprehensive financial insights:

• Tool Calling: Integrates various tools to pull real-time financial data.
• Agent Orchestration:
  1. Manages the agent's workflow and data retrieval.
  2. The Orchestrator Agent coordinates the Web Agent and Financial Agent.
  3. The Web Agent fetches data from DuckDuckGo and the Financial Agent retrieves financial data from YFinance.

See the Financial Agent Diagram at the top of this document

PDF Assistant Architecture

The PDF Assistant Agent focuses on document processing:

• Document Retrieval: Loads PDF content from predefined URLs.
• Embedding: Uses Hugging Face’s model to embed the document content.
• Storage & Querying: Saves the embedded data in a PostgreSQL vector database for later retrieval and analysis.

Setup and Installation

Prerequisites

• Python: Version 3.10 or higher.
• PostgreSQL: Required for the PDF Assistant. Ensure you have a running PostgreSQL instance with pgvector support.

Installation Steps

1. Clone the Repository:

   git clone https://github.com/yourusername/yash8745-agentic-ai.git
   cd yash8745-agentic-ai

2. Install Dependencies:

   pip install -r requirements.txt

3. Configure Environment Variables:

   Create a .env file in the project root and add your configurations. For example:

    PHI_API_KEY=""
    GROQ_API_KEY=""

   Adjust other variables as needed (e.g., database connection strings).

Usage

Running the Agents

python financial_agent.py
python pdf_assistant.py

To-Do List ✅

1. PDF Assistant Issue:
   The PDF assistant is running but encounters issues with Hugging Face embedding. Review the error logs and adjust the configuration.

Contributing

Contributions are welcome! If you find issues or have suggestions:

• Open an issue on the repository.
• Submit a pull request with your improvements.