vAnalytics provides a web interface to help easily monitor vLLM instance metrics. It allows users to easily monitor multiple vLLM instances, as well as being easy to setup and configure.
- Specify vLLM backends easily using name and host configuration
- Uses SQLite for easy database management
- Intuitive and includes error handling
- Flexible schemas and data plotting using Plotly
Configure your instances in monitor.py, then use python src/monitor.py. This will start monitoring in a /data folder, where it will store SQLite databases with your model name.
To start the web interface, execute python src/graph.py. The web interface is avaliable at localhost:4412.