Construction-Hazard-Detection

This MCP server exposes a suite of agent-ready tools designed to monitor and manage construction-hazard detection workflows. Built on FastMCP, it provides capabilities for real-time inference, hazard analysis, notification delivery, violation record management, streaming control, and model management. Key entry points include inference.detect_frame for frame-wise detection, hazard.detect_violations for analyzing detected hazards, and violations-related endpoints (search, get, get_image, my_sites) for auditing safety events. It also supports notification channels via notify.line_push, telegram_send, and broadcast_send, along with record operations like send_violation, batch_send_violations, and sync_pending. For model lifecycle, model.fetch/update/list_available/get_local is available, plus geometry utilities for spatial reasoning. The server can transport data over stdio, SSE, or streamable-HTTP, with HTTP as the default transport for typical usage. To get started, run the server in HTTP mode and interact with its MCP endpoints to perform detection, analyze hazards, push alerts, and manage records in real time.

Prerequisites:

• Python 3.12.x installed (the project targets Python 3.12.x as shown by the supported badges).
• Redis server available if you plan to store streams or results in Redis (recommended for multi-stream management).
• Optional: Docker, if you prefer running auxiliary services (e.g., Redis) via containers.

Step-by-step installation:

1. Create and activate a Python virtual environment

   • macOS/Linux: python3 -m venv venv source venv/bin/activate
   • Windows: python -m venv venv venv\Scripts\activate

2. Install required Python dependencies

   • Ensure you are in the project root or navigate to examples/mcp_server as needed
   • Install via pip (requirements are typically listed in the project, e.g. pip install -r requirements.txt or pip install fastmcp redis ultralytics etc.)

3. Verify Python environment and package availability

   • python --version should show 3.12.x
   • pip install fastmcp redis ultralytics

4. Run the MCP server (HTTP transport by default)

   • python -m examples.mcp_server.main

5. Optional: run with stdio transport for debugging

   • MCP_TRANSPORT=stdio python -m examples.mcp_server.main

If you prefer an isolated containerized setup, you can run a Redis instance via Docker and then start the MCP server with the appropriate environment variables configured for your setup.

Notes and tips:

• The MCP server exposes multiple tools via a single HTTP/streaming interface; explore endpoints such as inference.detect_frame and hazard.detect_violations to perform real-time analysis.
• If using Redis for stream storage, ensure Redis is reachable from the MCP server and that the appropriate keys/data structures are configured.
• For multi-language notifications, ensure tokens/credentials for LINE, Telegram, WeChat, etc., are configured in the environment or the server configuration as described in the MCP docs.
• When testing, you can run in stdio mode to simplify input/output during development and debugging; switch back to HTTP transport for production usage.
• Review the provided Quick Start JSON example to understand how streams and detections are configured and stored; adapt it to your site and model_key as needed.