AI Gateway¶
The AI Gateway is the single entry point for all LLM and agent requests. It handles authentication, tenant isolation, rate limiting, and model provider routing.
Responsibilities¶
- AuthN/AuthZ: API key and JWT validation
- Tenant isolation: Per-tenant configuration, rate limits, and model access
- Rate limiting: Token bucket algorithm with configurable RPM per tenant
- Model routing: Abstraction over OpenAI, Anthropic, and local/stub providers
Endpoint¶
POST /v1/gateway/complete
Source¶
src/agentguard/gateway/
Configuration¶
| Variable | Description | Default |
|---|---|---|
API_KEY_HEADER |
Header name for API key | X-API-Key |
JWT_SECRET |
JWT signing secret | (required in production) |
TENANT_HEADER |
Header for tenant ID | X-Tenant-ID |
DEFAULT_MODEL_PROVIDER |
Default LLM provider | openai |
DEFAULT_MODEL_NAME |
Default model name | gpt-4o |
RATE_LIMIT_REQUESTS_PER_MINUTE |
Per-tenant RPM | 60 |