Issue Station
Sign In
Tools hereToolbox

hermes - ✅(Solved) Fix [Bug]: kimi-for-coding 主模型 400 错误 — invalid temperature: only 0.6 is allowed [1 pull requests, 3 comments, 4 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
NousResearch/hermes-agent#11764•Fetched 2026-04-18 05:59:01
View on GitHub
Comments
3
Participants
4
Timeline
6
Reactions
1
Author
zhuyankarlzhuyankarl
Participants
breeze86breeze86gongli0929gongli0929tong333666999tong333666999zhuyankarlzhuyankarl
Timeline (top)
commented ×3cross-referenced ×1referenced ×1subscribed ×1

Fix Action

Fixed

PR fix notes

PR #11830: fix(provider): inject required temperature for kimi-for-coding model

Description (problem / solution / changelog)

Summary

The kimi-for-coding model rejects API requests with HTTP 400 unless temperature=0.6 is set. Hermes did not set any temperature for this model by default, causing all requests to fail.

Changes

  • run_agent.py: add _MODEL_REQUIRED_TEMPERATURES dict mapping model names to their required temperatures
  • run_agent.py: in _build_api_kwargs, inject the required temperature via setdefault before request_overrides so it can still be overridden by the user

Testing

Requests to kimi-for-coding no longer return HTTP 400. The request_overrides mechanism still allows users to override the temperature if needed.

Closes #11764

Changed files

  • run_agent.py (modified, +11/-0)
  • scripts/release.py (modified, +1/-0)

Code Example

invalid temperature: only 0.6 is allowed for this model
RAW_BUFFERClick to expand / collapse

Bug Description

切换 kimi-for-coding（provider: kimi-coding）为主模型后，所有消息均触发 400 错误：

invalid temperature: only 0.6 is allowed for this model

与 #9125 不同，本 issue 涉及的是 主对话循环（conversation loop），而非 auxiliary 调用。#9125 中 Kimi API 要求 temperature=1，而本场景中 kimi-for-coding 要求 temperature=0.6，说明不同 Kimi 子模型的 temperature 约束不一致。

Steps to Reproduce

  1. 设置主模型为 kimi-for-coding：hermes model kimi-coding/kimi-for-coding
  2. 发送任意消息
  3. 触发 400 错误，对话无法进行

Expected Behavior

Gateway 在调用 kimi-for-coding 时自动使用 temperature=0.6，或在请求参数中省略 temperature 让 API 使用默认值。

Actual Behavior

每次 API 调用均返回 400，主模型完全不可用。

Related

  • #9125 — 同类问题，但针对 auxiliary.vision / auxiliary.session_search（报错为 only 1 is allowed）

extent analysis

TL;DR

The most likely fix is to configure the gateway to use temperature=0.6 when calling the kimi-for-coding model.

Guidance

  • Verify that the kimi-for-coding model indeed requires temperature=0.6 by checking its documentation or testing it with a direct API call.
  • Check the gateway configuration to see if there's an option to set the temperature parameter for the kimi-for-coding model.
  • If the gateway allows it, try omitting the temperature parameter in the request to see if the API uses the default value.
  • Review the related issue #9125 to understand how the temperature constraint is handled for other Kimi sub-models.

Example

No code snippet is provided as the issue does not contain sufficient information about the gateway's API or configuration.

Notes

The solution may depend on the specific implementation of the gateway and the Kimi API. It's also unclear if the gateway has a built-in mechanism to handle different temperature constraints for various models.

Recommendation

Apply workaround: configure the gateway to use temperature=0.6 for the kimi-for-coding model, as this is the only value allowed by the model.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#api#autograd error#model save/load#optimization#mixed precision

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING