Official OpenAI SDKs (Python / Go / Java …) against compatible endpoints

When RouterBrain exposes /v1/chat/completions and related OpenAI-compatible routes, any client that supports base_url + api_key can point at the same Base URL without @routerbrain/sdk (for Node/TS the SDK is still recommended for types and multimodal helpers).

Common settings

SettingValue
Base URLhttps://<your-host>/v1 (watch trailing slash rules vs client URL joining)
API KeyTenant sk-…
ModelExact code from GET /v1/models

Smoke-test /v1/models and /v1/chat/completions with curl before the SDK to separate network/TLS from client config.

Python (openai package, illustrative)

Names and parameters depend on your installed version:

from openai import OpenAI

client = OpenAI(
    base_url="https://gw.example.com/v1",
    api_key="sk-…",
)

resp = client.chat.completions.create(
    model="openai/gpt-4o-mini",
    messages=[{"role": "user", "content": "ping"}],
)
print(resp.choices[0].message.content)

Streaming: use stream=True and chunk iterators per docs; mind proxies and SSE (Troubleshooting).

When to prefer @routerbrain/sdk

  • Assemble input_file, audio parts, etc. from local paths / URLs on Node;
  • Want plugins.pdf → pdf_preprocess merged automatically;
  • Want unified GatewayHttpError handling.

See chat.messages.

LangChain, LlamaIndex, LiteLLM, etc.

These usually accept OpenAI-compatible base_url / api_key: point at …/v1, model id = code from GET /v1/models. Field names differ by framework—check their docs on upgrades.

Checklist: Base URL includes /v1; streaming uses SSE (no proxy buffering—Troubleshooting); tool calls may hit unsupported extensions (400/501). Multimodal needs input_file—Chat Completions.

See also

Back to docs home