README.md

metadata

title: First Agent Template
emoji: ⚡
colorFrom: pink
colorTo: yellow
sdk: gradio
sdk_version: 5.23.1
app_file: app.py
pinned: false
tags:
  - smolagents
  - agent
  - smolagent
  - tool
  - agent-course

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

🤖 First Agent Project with smolagents

An intelligent, code-backed AI agent capable of dynamic tool execution, reasoning, and multi-modal tasks (like text-to-image generation and timezone calculations). Built using the smolagents framework, powered by Qwen2.5-Coder-32B-Instruct, and deployed seamlessly on Hugging Face Spaces with a user-friendly Gradio interface.

🚀 What This Project Is About

This project demonstrates the implementation of a full CodeAgent using the smolagents library. Unlike traditional rigid chatbots, this agent operates on a Thought-Action-Observation cycle. When given a complex prompt, it writes execution blocks in Python, leverages specialized tools, evaluates the live feedback, and formulates an accurate final response.

Core Features

• Dynamic Code Execution: Solves problems by interpreting and running Python blocks on the fly.
• Custom Tool Integration: Extensible structure to add your own Python functions as agent tools.
• Text-to-Image Generation: Connects directly to Hugging Face Hub inference models to visualize user prompts.
• Gradio Web UI: A clean, shareable interface for interactive chatting and multi-modal outputs.

🛠️ Tech Stack

• Framework: smolagents (Hugging Face)
• LLM Engine: Qwen/Qwen2.5-Coder-32B-Instruct (via Hugging Face Serverless Inference API)
• UI Framework: Gradio
• Language: Python 3.10+
• Libraries: pytz, pyyaml, requests

📁 Project Structure

├── app.py                 # Main entry point: initializes the model, tools, and Gradio UI
├── prompts.yaml           # System prompts and instruction templates for the agent
├── requirements.txt       # Python dependencies
├── Gradio_UI.py           # Helper script setting up the Gradio layout
└── tools/
    └── final_answer.py    # Core tool defining how the agent concludes its reasoning