概览
What is MCP CamStream Analyzer?
MCP CamStream Analyzer is a tool for collecting and analyzing content from cameras and RTSP video streams. It integrates with any OpenAI-compatible API to generate descriptions and analyses of images and video frames.
How to use MCP CamStream Analyzer?
Clone the repository, install dependencies with pip install -r requirements.txt, edit the auto-generated config.yaml (especially the api_key field), and run python main.py. The service then processes camera feeds according to the configured templates and video analysis settings.
Key features of MCP CamStream Analyzer
- Image analysis using customizable description templates
- Video analysis with frame-by-frame or full-video mode
- Support for default, USB, and RTSP camera streams
- Integration with any OpenAI-compatible API
- Configurable frame interval and output directory
- Built-in prompt templates (default, detailed, security)
Use cases of MCP CamStream Analyzer
- Security monitoring with automated abnormality detection
- Detailed scene and people analysis for surveillance footage
- General-purpose image or video description for accessibility
- Multi-camera environment monitoring (e.g., IP cameras)
- Real-time event logging from RTSP streams
FAQ from MCP CamStream Analyzer
What dependencies are required?
Python packages listed in requirements.txt are required. The server also needs an API key for an OpenAI-compatible service.
How do I configure cameras?
The config.yaml file lists cameras as default (integer device index), webcam1 (another device index), or ip_camera1 (with rtsp_url, buffer_size). Modify the file to match your hardware.
What is the default video analysis mode?
The default mode is frames, with a frame_interval of 30. You can change it to full video analysis by setting mode: entire.
Does the server support authentication?
The README only mentions an api_key for the OpenAI-compatible API. No other authentication mechanisms are described.
Where is the output stored?
Analysis results are written to the directory specified in the output_dir field of config.yaml (default: output).
