title	Template Final Assignment
emoji	🕵🏻‍♂️
colorFrom	indigo
colorTo	indigo
sdk	gradio
sdk_version	5.25.2
app_file	app.py
pinned	false
hf_oauth	true
hf_oauth_expiration_minutes	480

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

LangGraph Agent with OpenAI

This project implements an AI agent using LangGraph and OpenAI models. The agent can answer questions, perform calculations, and search for information using various tools.

Features

• Uses LangGraph for agent workflow management
• Integrates with OpenAI models
• Includes various tools:
  • Basic math operations (add, subtract, multiply, divide, modulus)
  • Web search via Tavily
  • Wikipedia search
  • ArXiv paper search
  • Similar question search from vector database
• Vector store for retrieving similar questions and answers
• Gradio web interface for easy interaction

Setup

1. Environment Variables

Create a .env file in the root directory with the following variables:

OPENAI_API_KEY=your_openai_api_key_here
TAVILY_API_KEY=your_tavily_api_key_here

Optionally, you can also add:

GOOGLE_API_KEY=your_google_api_key_here

2. Install Dependencies

pip install -r requirements.txt

3. Run the Application

python app.py

This will start the Gradio web interface where you can interact with the agent.

Customizing the Agent

Changing the LLM Provider

You can modify the LLM provider and model in app.py:

agent = BasicAgent(provider="openai", model_name="gpt-4.1-2025-04-14")

Supported providers:

• openai (default): Uses OpenAI models
• google: Uses Google Generative AI models

Adding Custom Tools

You can add new tools in agent.py by creating functions decorated with @tool:

@tool
def your_tool_name(param: type) -> return_type:
    """Tool description.
    
    Args:
        param: Parameter description
    """
    # Tool implementation
    return result

Then add your tool to the tools list:

tools = [
    multiply, add, subtract, divide, modulus,
    wiki_search, web_search, arvix_search,
    your_tool_name,  # Add your new tool here
]

How It Works

1. The agent uses LangGraph to manage the workflow between different components.
2. When a question is received, it first checks for similar questions in the vector database.
3. The LLM (OpenAI by default) processes the question and decides what tools to use.
4. If tools are needed, they are executed and the results are fed back to the LLM.
5. The process continues until the LLM provides a final answer.

Architecture

The agent is built using a state graph with the following components:

• Retriever Node: Searches for similar questions in the vector database
• Assistant Node: Processes the question and decides what to do next
• Tools Node: Executes tools requested by the assistant

The graph flow is:

1. START → Retriever
2. Retriever → Assistant
3. Assistant → Tools (if needed) or END
4. Tools → Assistant

License

This project is open source and available under the MIT License.