Skip to content

v0.27.4

Choose a tag to compare

@yuanhao yuanhao released this 07 Nov 16:04
· 16 commits to main since this release
85c7b67

A fix release, with no breaking changes.

  • No temperature for GPT-6. OpenAI's latest-model guide says: "When reasoning effort is not none, remove temperature, top_p, and top_logprobs."
    • No GPT-6 model defaults to none. Astra and 6.1 Sol do not support it at all.
    • yoagent never sends none.
    • So with_temperature on GPT-6 failed every request.
    • ModelConfig::openai and openai_responses now set accepts_sampling_params from the id, and a gpt-6… id turns it off. That covers the presets, generic("openai", ..), OpenCode and OpenRouter. The agent loop then leaves the temperature out.
    • OpenAI's other reasoning models keep the flag on. If yours rejects a temperature, call with_accepts_sampling_params(false).

yoagent-rutis 0.13.0 and yoagent-workers 0.4.0 are unchanged.

Full notes: CHANGELOG.md