AI Gateway

OpenAI-compatible chat completions endpoint. Proxies requests through a flowgen flow, enabling pre/post processing, logging, and tool execution.

Exposes a /v1/chat/completions endpoint that accepts the OpenAI request format and streams responses as Server-Sent Events.

Configuration

- ai_gateway:
    name: gateway
    credentials_path: /path/to/credentials.json

The AI gateway task requires the HTTP server to be enabled in the worker config:

worker:
  http_server:
    enabled: true
    port: 3000

Requests are sent to POST /v1/chat/completions with the standard OpenAI chat format.

Format: JSON

Field	Type	Description
`prompt`	string	User prompt merged from messages.
`system_prompt`	string / null	System message if present.
`model`	string	Requested model name.
`temperature`	number / null	Sampling temperature if specified.
`max_tokens`	int / null	Token limit if specified.
`stream`	bool	Whether streaming was requested.