MoonshotAI / MoonshotAI/kimi-code

[thinking].effort 缺乏 TUI 可视反馈和请求验证手段

Open
#1,051 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
TypeScript
Stars
7.5k
Forks
1.2k
Avg merge
11h 53m
Merged PRs (30d)
350

Description

问题描述

使用第三方模型(OpenAI 兼容协议,如 opencode-go)时,[thinking].effort 配置有三处体验问题:

1. TUI 中无思考强度可视反馈

设置了 [thinking]modeeffort 后,在 Kimi Code TUI 中没有任何地方能查看当前的思考强度(如 effort = max)。用户只能通过编辑 config.toml 文件来判断,容易遗忘当前设置。

期望:在模型切换面板(Ctrl+M)或状态栏中,显示当前 Thinking 状态和 effort 级别,例如 Thinking: ON (max)

2. 日志无法验证 reasoning_effort 是否实际发送

Kimi Code 官方文档明确说对 OpenAI 兼容协议会自动注入 reasoning_effort(参见 Providers and Models 中 "Third-party reasoning models work out of the box: the CLI automatically handles the reasoning_content field and reasoning_effort injection")。但默认 INFO 日志级别不记录 API 请求体,用户无法验证 reasoning_effort: "max" 是否真的发给了第三方 API。需要开启 debug 日志或用抓包工具才能确认。

期望

  • 增加 debug 日志级别,打印完整 API 请求体(或至少关键字段如 reasoning_effort
  • 或在 /config 中添加"测试连接"功能,显示实际发出的请求参数
3. /config 交互界面未暴露 [thinking] 配置段

内置的 /config 命令打开交互式配置界面后,找不到 [thinking] 相关选项。用户只能手动编辑 config.toml/reload

期望:在 /config 交互界面中增加 Thinking 配置项(mode 和 effort)。

环境信息

  • kimi-code 版本:0.19.1
  • 操作系统:macOS (darwin/arm64)
  • 第三方 provider 类型:openai(opencode-go)

配置文件片段

[thinking]
mode = "on"
effort = "max"

[providers.opencode-go]
type = "openai"
base_url = "https://opencode.ai/zen/go/v1"

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by locating the Ctrl+M model-switch panel, the /config interaction, and the API request logging path. Trace how [thinking] mode and effort are loaded and how OpenAI-compatible requests are assembled. Done means the TUI exposes the current setting, the configuration interface can edit it, and users have a supported way to verify the reasoning_effort request field.

Written by the indexing model from the issue text.

Assessment

Tech stack
typescript
Domain
api, cli, developer-experience, observability
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.