55e83bb4f0176536de9483c951497228ff7e76b6
49%

Ran 29 Nov 2025 07:43PM UTC

Jobs 0

Files 0

Run time –

Badge

Embed ▾

pending completion

cancel

Build # 55e83bb4f0176536de9483c951497228ff7e76b6

Build Type

push

github

Committed by

web-flow

Commit Message

fix: increase default timeout for OpenAI reasoning models (#252)

Fixes #242 - TransportError{reason: :closed} with GPT-5.1 and structured objects.

The issue was that the default 30 second timeout was too short for reasoning
models (o1, o3, gpt-5, etc.) which use the Responses API. These models perform
extended reasoning which can take much longer, especially with large inputs
(40k+ tokens).

Changes:
- OpenAI provider now uses thinking_timeout (300s default) for Responses API models
- Also sets pool_timeout and connect_options for consistent timeout handling
- Updated test config to include api: 'responses' for GPT-5 model variants
- Added tests to verify timeout behavior for reasoning vs chat models

Amp-Thread-ID: https://ampcode.com/threads/T-cb219e55-2431-4d13-a8a9-3f91f57e0c95

Co-authored-by: Amp <amp@ampcode.com>

Run Details

Source Files on build 55e83bb4f0176536de9483c951497228ff7e76b6

Detailed source file information is not available for this build.

agentjido / req_llm / 55e83bb4f0176536de9483c951497228ff7e76b6
49%

README BADGES
x

Markdown

Textile

RDoc

HTML

Rst

Source Files on build 55e83bb4f0176536de9483c951497228ff7e76b6

agentjido / req_llm / 55e83bb4f0176536de9483c951497228ff7e76b6 49%

README BADGES x

Markdown

Textile

RDoc

HTML

Rst

Source Files on build 55e83bb4f0176536de9483c951497228ff7e76b6

agentjido / req_llm / 55e83bb4f0176536de9483c951497228ff7e76b6
49%

README BADGES
x