[Bug]: Factory Droid (ACP Registry) shows reasoning effort only for its default model since #35458

未关闭
  • 哪里不对？
#36,153 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
4/5
预计耗时
1-2 天
新手友好度
55/100
Issue 类型
缺陷
描述清晰度
描述清楚
活跃度
活跃
技术栈
typescript
领域
backend-api-design

调研方向

Start in the apps/server area and read the changes in #35458, which the issue identifies as the regression. Reproduce with Factory Droid through the ACP Registry, comparing the reasoning-effort options for the default Sonnet 5.5 model and non-default Opus 5.5 or GPT-6.1 Sol. Done when each model shows the reasoning-effort levels Droid reports, without options leaking between models.

由索引模型根据 Issue 内容生成。

描述

Before submitting
  • I searched existing issues and did not find a duplicate.
  • I included enough detail to reproduce or investigate the problem.
Area

apps/server

Steps to reproduce
  1. Set Droid's default model to Sonnet 5.5 (sessionDefaultSettings.model: "claude-sonnet-5-5" in ~/.factory/settings.json).
  2. Start T3 Code nightly 0.1.10-nightly.20261113.4630 with Factory Droid added through the ACP Registry.
  3. In a new thread, pick Factory Droid with Opus 5.5 (or any model other than Sonnet 5.5) and open the model options.
Expected behavior

Opus 5.5 shows a Reasoning Effort picker with the levels Droid reports for it (low, medium, high, xhigh, max), as it did before the update.

Actual behavior

In the composer model picker, Factory Droid's Opus 5.5 and GPT-6.1 Sol show only Autonomy Level. Droid's Sonnet 5.5, the model the startup probe session opens on, shows Reasoning Effort (Off, Low, Medium, High, Extra High, Maximum) plus Autonomy Level.

The provider cache shows the same thing:

  • ~/.t3/caches/acpRegistry_factory_droid.json (checked 2026-11-13): reasoning_effort on 1 of 74 models (claude-sonnet-5-5).
  • Older ~/.t3/caches/droid.json from 2026-10-06, same acpRegistry driver: reasoning_effort on 58 of 59 models.

Droid still reports the option for each model. Driving droid exec --output-format acp directly (initialize, session/new, then session/set_config_option):

  • session/new -> current model claude-sonnet-5-5, reasoning_effort = off/low/medium/high/xhigh/max
  • set model = claude-opus-5-5 -> reasoning_effort = low/medium/high/xhigh/max
  • set reasoning_effort = high -> accepted, current value high
  • set model = gpt-6.1-sol -> reasoning_effort = low/medium/high/xhigh/max

This started with #35458. It keeps model-specific options only for the model that was current in a session. Its "Not in this change" section lists this case: a model that was never current in a session shows no reasoning options. For Droid, the probe always opens on the user's default model, so only that one model gets an effort picker. The fix for #15177 stopped one model's options from leaking to the others, but it also took the effort picker away from every Droid model except one.

Impact

Major degradation or frequent failure

Version or commit

0.1.10-nightly.20261113.4630

Environment

macOS 26.5, T3 Code desktop (Nightly), Factory Droid 1.10.0 via ACP Registry

Workaround

None found in T3. Changing Droid's default model only moves the picker to that one model.

Written with Claude Opus 5.5 in Factory Droid. The initial investigation was done with GPT-6.1 Sol in Codex.

主要语言
TypeScript
星标
24.8k
派生
6.4k
平均合并
59 分钟
30 天内合并 PR
418

环境准备

在 Codespaces 中打开

在浏览器里用你自己的 GitHub 账号启动这个项目的开发容器。

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

pingdotgg/t3code 的其他 Issue

查看 pingdotgg/t3code 的全部 Issue

相似的 Issue

更多 TypeScript Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。