README.md

metadata

title: SecrectDocumentLocker
app_file: app.py
sdk: gradio
sdk_version: 6.3.0

SmolAgents with FastMCP Demo

This project demonstrates a SmolAgents agent interacting with a local MCP server created using FastMCP.

Features

• MCP Server: (mcp_server.py)
  • Implements a dummy authentication tool.
  • Protects a "treasure" resource (treasure://secret) which is only accessible after authentication.
  • Exposes a read_treasure tool.
• Agent: (agent.py)
  • Uses smolagents.MCPClient to connect to the local MCP server.
  • Dynamically loads tools from the MCP server.
  • Uses meta-llama/Llama-4-Scout model via Hugging Face Inference API.
• UI: (app.py)
  • Provides a Gradio chat interface for the agent.

Setup

1. Install Dependencies:

   pip install -r requirements.txt
   

2. Environment Variables: You need a Hugging Face Token to use the Inference API. Create a .env file in the root directory:

   HF_TOKEN=your_hf_token_here
   

Running Locally

1. Run the Gradio app:

   python app.py
   

2. Open the link provided in the terminal (usually http://127.0.0.1:7860).
3. Chat with the agent! Try:

   "Read the treasure file." (The agent should try, fail, then realize it needs to authenticate. Hint: the password is "open sesame")

Deploying to Hugging Face Spaces

1. Create a new Space on Hugging Face (SDK: Gradio).
2. Upload the files:
   • app.py
   • agent.py
   • mcp_server.py
   • dummy_treasure.txt
   • requirements.txt
3. Set the HF_TOKEN in the Space's Settings > Variables and secrets (if not automatically handled, though Spaces usually have access to the token of the owner if configured, but explicit token is safer for specific model access).
4. The Space should build and run automatically!

Files

• mcp_server.py: The FastMCP server implementation.
• agent.py: Agent logic including MCP connection.
• app.py: Entry point for Gradio.
• dummy_treasure.txt: The protected content.

Deploying to Hugging Face Spaces

1. Create a new Space on Hugging Face (SDK: Gradio).
2. Upload the files:
   • app.py
   • agent.py
   • mcp_server.py
   • dummy_treasure.txt
   • requirements.txt
3. Set the HF_TOKEN in the Space's Settings > Variables and secrets (if not automatically handled, though Spaces usually have access to the token of the owner if configured, but explicit token is safer for specific model access).
4. The Space should build and run automatically!

Files

• mcp_server.py: The FastMCP server implementation.
• agent.py: Agent logic including MCP connection.
• app.py: Entry point for Gradio.
• dummy_treasure.txt: The protected content.

Deploying to Hugging Face Spaces

hf upload --repo-type space ohmp/SecrectDocumentLocker . --exclude ".env" ".venv/**" "__pycache__/**" "*.pyc"