TensorZero Gateway
TensorZero Gateway is an AI inference optimization platform that provides model routing, observability, and experimentation capabilities for AI applications.
Features
- Model Routing: Intelligently route requests to different AI models
- Observability: Built-in metrics and monitoring with ClickHouse integration
- Experimentation: A/B testing and variant management for AI models
- Multi-Provider Support: Works with OpenAI, Anthropic, AWS Bedrock, and many other providers
Configuration
The gateway uses a TOML configuration file that defines:
- Model configurations and routing rules
- Function definitions for different AI tasks
- Metrics collection and optimization settings
Required Setup
- API Keys: Configure API keys for the AI providers you want to use
- Configuration Files: The service includes a sample configuration with GPT-4o-mini setup. Configure it in Config Editor.
Health Check
The service includes a built-in health check endpoint at /status
that monitors the gateway's operational status.
Documentation
For more information, visit the TensorZero documentation.