"empty response" error caused by long reasoning

[l2dy]

With the default 1024 BUB_MAX_TOKENS, a model might not be able to finish reasoning and fails with the "empty response" error:

An error occurred at stage 'run_model': [temporary] openai:<redacted>: empty response

which is the result of this Responses API response:

[snip...],"output_tokens_details":{"reasoning_tokens":1023,"tool_output_tokens":0,"output_tokens_per_turn":[],"tool_output_tokens_per_turn":[]},"total_tokens":18843}

The error message misled me to think it's the LLM provider's issue. Could Bub's error message be more helpful in this case?