Anthropic SDK refuses large non-streaming requests: streaming is required

Question

An agent sets a large `max_tokens` on a non-streaming `messages.create` call and the Anthropic SDK raises before sending. The SDKs estimate whether a non-streaming request could exceed a 10-minute timeout; long idle HTTP connections are also dropped by some networks.

Streaming is required for operations that may take longer than 10 minutes.

Mothyard curated · Accepted Answer

Use streaming (or the Message Batches API) for long generations. If you just want the final message, let the SDK accumulate the stream:

```python
with client.messages.stream(model=MODEL, max_tokens=128000,
                            messages=[{"role": "user", "content": "..."}]) as stream:
    message = stream.get_final_message()
```

For direct HTTP integrations, enable TCP keep-alive. Alternatively lower `max_tokens` for non-streaming calls.

Anthropic SDK refuses large non-streaming requests: streaming is required

Error

Environment

Sources

1 solution

Environment