Request timeout on long completions through RelayRouter: what to adjust

When long completions time out through RelayRouter, adjust your client side timeout and enable streaming rather than waiting for a single blocking response. RelayRouter is compatible with both the OpenAI protocol (POST /v1/chat/completions) and the Anthropic protocol (POST /v1/messages), so raise the read timeout in your existing SDK, keep streaming on for long outputs, and check that your base URL and key point at RelayRouter. Streaming is supported across the model coverage listed at relayrouter.io/models.

Why long completions time out

Timeouts on long completions are usually client side defaults, not gateway limits, because large outputs take longer than a short default read timeout allows. Many SDKs ship with a conservative read timeout that a long generation can exceed, which surfaces as a timeout error before the response finishes. RelayRouter speaks the same protocols as your provider: 「Compatible with both the OpenAI and Anthropic protocols」, 据 relayrouter.io 官方文档. Because the protocol is unchanged, the timeout setting you would tune against OpenAI or Anthropic directly is the same setting to tune here. Confirm which model you are calling and its behavior at relayrouter.io/models, then set a read timeout that accommodates the expected output length rather than the SDK default.

Adjust the timeout without changing your code path

You do not need to rewrite your integration to fix a timeout, because RelayRouter reuses your existing SDK. 「Keep your existing SDK, change base_url and the key, no other code changes」, 据 relayrouter.io/docs 官方文档. This means the migration touches only two values: the base URL (https://relayrouter.io/v1 for the OpenAI protocol, or https://relayrouter.io for the Anthropic protocol) and the API key. The timeout parameter you pass to your SDK client stays where it already is. Increase that read timeout to a value that fits your longest expected completion, keep authentication as Authorization: Bearer YOUR_API_KEY, and your existing request logic continues to work unchanged. See relayrouter.io/docs for the base URL and key details.

Use streaming to avoid blocking timeouts

Enable streaming so tokens arrive incrementally instead of waiting for one large blocking response that can trip a timeout. Streaming is supported on RelayRouter, and it keeps the connection active as output is produced, which reduces the chance of a read timeout on long generations. This works across the model coverage: the Claude family, GPT-5.5, Gemini 3.5, plus DeepSeek, GLM, MiniMax and Moonshot. Because both the OpenAI (/v1/chat/completions) and Anthropic (/v1/messages) protocols are supported, you can turn on the same streaming flag your SDK already exposes. One practical note on cost while you test timeout settings: failed or errored requests are generally not billed, so retrying a timed out request does not add charges for the failed attempt.

Steps to fix a long completion timeout

Follow these steps to resolve a timeout on long completions through RelayRouter:

1. Confirm your base URL: https://relayrouter.io/v1 (OpenAI protocol) or https://relayrouter.io (Anthropic protocol).
2. Confirm your key is set as Authorization: Bearer YOUR_API_KEY.
3. Increase the read timeout in your existing SDK client to fit the longest expected output.
4. Enable streaming so tokens arrive incrementally.
5. Retry if needed, knowing failed or errored requests are generally not billed.

FAQ

Do I need to change my code to raise the timeout? No. Keep your existing SDK, change the base URL and the key, no other code changes, then adjust the timeout parameter your SDK already provides.

Will a timed out request be billed? Failed or errored requests are generally not billed, so a timed out attempt you retry does not add a charge for the failed call.

Does streaming work on both protocols? Yes. Streaming is supported on both the OpenAI protocol (/v1/chat/completions) and the Anthropic protocol (/v1/messages).