Multi Tool Agent

This project implements an intelligent Python agent focused on research assistance. It is capable of searching and retrieving scientific papers from arXiv.

Main Features

• Arxiv MCP integration: Incorporates a Model Context Protocol (MCP) tool for searching and retrieving scientific papers from arXiv, enabling the agent to answer research-related queries.
• Extensible agent: Based on Google ADK, you can easily add new tools or models.
• Modular code: Main functions are in multi_tool_agent/agent.py.

Project Structure

• multi_tool_agent/agent.py: Contains the main functions and agent definition.
• multi_tool_agent/__init__.py: Initializes the module.

Requirements

• Python 3.11+
• uv (required for fast and reliable dependency management)
• Ollama (required to run the local LLM)
• Ollama model: qwen3:8b (must be pulled with ollama pull qwen3:8b)
• Packages: httpx, geopy, google-adk (and ADK dependencies)

Installation

1. Install uv if you don't have it:

   pip install uv

2. Install project dependencies using uv:

   uv pip install -r requirements.txt
   # or, for a quick setup:
   uv pip install httpx geopy google-adk

3. (Optional) Set up your virtual environment.
4. Install Ollama and pull the required model:

   # Install Ollama (see https://ollama.com/download)
   ollama pull qwen3:8b

Configuration

To ensure all dependencies are up to date and properly installed, run:

uv sync

.env file configuration

Create a .env file in the project root to store environment variables. Example:

# Example .env file
# Set your Google API key if required by google-adk or other services
GOOGLE_API_KEY=your_google_api_key_here

# Ollama API base URL (required if using Ollama locally)
OLLAMA_API_BASE="http://localhost:11434"

# Set other environment variables as needed by your project

• If you use Ollama locally, make sure to set OLLAMA_API_BASE as shown above.
• If you use any API keys or secrets, add them here and never commit your .env file to version control.
• The project will automatically load variables from .env if you use packages like python-dotenv.

Usage

• Ollama and the qwen3:8b model must be running locally for the agent to work.
• If the city is not found or the API does not respond, a friendly error message is returned.
• If you need to use API keys, set them in the .env file as shown above.
• To run the project, execute:

  adk web

  from the root folder. This will start the ADK web server for your agent.

License

MIT

---

Author: emi-dm