Skip to main content

OpenAI

An llm · client with dialect: openai sends requests to the OpenAI Chat Completions API. Requests that arrived in the Anthropic API are translated to the OpenAI format and back.

Configuration

openai_llm_client:
  type: llm
  kind: client
  options:
    dialect: openai
    server: https://api.openai.com
    authorization:
      api_key:
        credentials: "Bearer {credentials}"
  telemetry:
    metrics:
      - llm.*
  exit: sys:http_client

Server URL

Zilla appends /chat/completions to the server URL. When the URL has no path, or only /, the path defaults to /v1, so https://api.openai.com sends requests to https://api.openai.com/v1/chat/completions.

The same applies to any server that implements the OpenAI Chat Completions API: set server to its base URL.

Credentials

options.authorization adds the API key to each request as Authorization: Bearer <key>. With an inline guard, the key is the one the caller sent, forwarded unchanged. See LLM Credentials.

Routing to OpenAI

north_llm_proxy:
  type: llm
  kind: proxy
  routes:
    - when:
        - model:
            - gpt-*
      exit: openai_llm_client

Next Steps