feat(llm): better thinking mode control

.kimi/prompts/gen-changelog.md

@@ -1 +1 @@
-根据当前分支相对于 main 分支的修改，在的 @CHANGELOG.md 文件中添加更新日志条目，遵循现有的格式和风格。
+根据当前分支相对于 main 分支的修改，在的根目录 CHANGELOG.md 或相应 package/sdk 等目录的 CHANGELOG.md 文件中添加更新日志条目，遵循现有的格式和风格。

CHANGELOG.md

@@ -11,6 +11,10 @@ Only write entries that are worth mentioning to users.
 
 ## Unreleased
 
+- Shell: Use `/model` to toggle thinking mode instead of Tab key
+- Config: Add `default_thinking` config option (auto-migrated from metadata)
+- LLM: Add `always_thinking` capability for models that always use thinking mode
+
 ## 0.76 (2026-01-12)
 
 - Tool: Make `ReadFile` tool description reflect model capabilities for image/video support

docs/en/configuration/config-files.md

@@ -25,6 +25,7 @@ The configuration file contains the following top-level configuration items:
 | Item | Type | Description |
 | --- | --- | --- |
 | `default_model` | `string` | Default model name, must be a model defined in `models` |
+| `default_thinking` | `boolean` | Whether to enable thinking mode by default (defaults to `false`) |
 | `providers` | `table` | API provider configuration |
 | `models` | `table` | Model configuration |
 | `loop_control` | `table` | Agent loop control parameters |
@@ -35,6 +36,7 @@ The configuration file contains the following top-level configuration items:
 
 ```toml
 default_model = "kimi-for-coding"
+default_thinking = false
 
 [providers.kimi-for-coding]
 type = "kimi"

docs/en/configuration/env-vars.md

@@ -53,7 +53,7 @@ export KIMI_MODEL_MAX_CONTEXT_SIZE="262144"
 
 ### `KIMI_MODEL_CAPABILITIES`
 
-Overrides the model's `capabilities` field in the configuration file. Multiple capabilities are comma-separated, supported values are `thinking`, `image_in`, and `video_in`.
+Overrides the model's `capabilities` field in the configuration file. Multiple capabilities are comma-separated, supported values are `thinking`, `always_thinking`, `image_in`, and `video_in`.
 
 ```sh
 export KIMI_MODEL_CAPABILITIES="thinking,image_in"

docs/en/configuration/providers.md

@@ -108,7 +108,8 @@ The `capabilities` field in model configuration declares the capabilities suppor
 
 | Capability | Description |
 | --- | --- |
-| `thinking` | Supports thinking mode (deep reasoning) |
+| `thinking` | Supports thinking mode (deep reasoning), can be toggled |
+| `always_thinking` | Always uses thinking mode (cannot be disabled) |
 | `image_in` | Supports image input |
 | `video_in` | Supports video input |
 
@@ -122,7 +123,11 @@ capabilities = ["thinking", "image_in"]
 
 ### `thinking`
 
-When thinking mode is enabled, the model performs deeper reasoning before answering, suitable for complex problems. In shell mode, you can toggle thinking mode with the `Tab` key, or control it at startup with `--thinking` / `--no-thinking` flags.
+Declares that the model supports thinking mode. When enabled, the model performs deeper reasoning before answering, suitable for complex problems. In shell mode, you can use the `/model` command to switch models and thinking mode, or control it at startup with `--thinking` / `--no-thinking` flags.
+
+### `always_thinking`
+
+Indicates the model always uses thinking mode and cannot be disabled. For example, models with "thinking" in their name like `kimi-k2-thinking-turbo` typically have this capability. When using such models, the `/model` command won't prompt for thinking mode toggle.
 
 ### `image_in`
 

docs/en/faq.md

@@ -33,15 +33,6 @@ If you need to change working directory:
 - **Use `--work-dir` flag**: Specify working directory at startup, like `kimi --work-dir /path/to/project`.
 - **Use absolute paths in commands**: Execute commands with absolute paths directly, like `ls /path/to/dir`.
 
-### Thinking mode unavailable
-
-When pressing `Tab` to toggle thinking mode, if you see "Thinking mode requires a model with thinking capability", it means the current model doesn't support thinking mode.
-
-Solutions:
-
-- **Switch to a supported model**: Use `/setup` to select a model that supports thinking (like `kimi-k2-thinking-turbo`).
-- **Check model capability configuration**: If you're sure the model supports thinking, you can force enable it with the environment variable `KIMI_MODEL_CAPABILITIES=thinking`.
-
 ### Image paste fails
 
 When using `Ctrl-V` to paste an image, if you see "Current model does not support image input", it means the current model doesn't support image input.

docs/en/guides/interaction.md

@@ -29,14 +29,14 @@ In shell mode, each command executes independently. Commands that change the env
 
 Thinking mode allows the AI to think more deeply before responding, suitable for handling complex problems.
 
-In agent mode, press `Tab` to toggle thinking mode on or off. The status bar at the bottom will show a notification after switching. You can also enable it at startup with the `--thinking` flag:
+You can use the `/model` command to switch models and thinking mode. After selecting a model, if the model supports thinking mode, the system will ask whether to enable it. You can also enable it at startup with the `--thinking` flag:
 
 ```sh
 kimi --thinking
 ```
 
 ::: tip
-Thinking mode requires support from the current model.
+Thinking mode requires support from the current model. Some models (like `kimi-k2-thinking-turbo`) always use thinking mode and cannot be disabled.
 :::
 
 ## Multi-line input

docs/en/reference/keyboard.md

@@ -7,7 +7,6 @@ Kimi CLI shell mode supports the following keyboard shortcuts.
 | Shortcut | Function |
 |----------|----------|
 | `Ctrl-X` | Toggle agent/shell mode |
-| `Tab` | Toggle thinking mode |
 | `Ctrl-/` | Show help |
 | `Ctrl-J` | Insert newline |
 | `Alt-Enter` | Insert newline (same as `Ctrl-J`) |
@@ -25,20 +24,9 @@ Press `Ctrl-X` in the input box to switch between two modes:
 - **Shell mode**: Input is executed as local shell command
 
 The prompt changes based on current mode:
-- Agent mode: `✨` or `💫` (thinking mode)
+- Agent mode: `✨` (normal) or `💫` (thinking mode)
 - Shell mode: `$`
 
-### `Tab`: Toggle thinking mode
-
-In agent mode, when the input box is empty or no completion menu is displayed, press `Tab` to toggle thinking mode.
-
-- When thinking mode is enabled, the prompt changes to `💫`
-- A notification is displayed when switching
-
-::: tip
-Thinking mode requires model support.
-:::
-
 ### `Ctrl-/`: Show help
 
 Press `Ctrl-/` in the input box to quickly display help information, equivalent to entering the `/help` command.

docs/en/reference/slash-commands.md

@@ -43,11 +43,11 @@ After configuration, settings are automatically saved to `~/.kimi/config.toml` a
 
 ### `/model`
 
-Switch the default model.
+Switch models and thinking mode.
 
-This command first refreshes the available models list from the API platform. When called without arguments, displays an interactive selection interface; you can also specify a model name directly, e.g., `/model kimi-k2`.
+This command first refreshes the available models list from the API platform. When called without arguments, displays an interactive selection interface where you first select a model, then choose whether to enable thinking mode (if the model supports it).
 
-After selecting a new model, Kimi CLI will automatically update the configuration file and reload.
+After selection, Kimi CLI will automatically update the configuration file and reload.
 
 ::: tip
 This command is only available when using the default configuration file. If a configuration was specified via `--config` or `--config-file`, this command cannot be used.

docs/en/release-notes/changelog.md

@@ -4,6 +4,10 @@ This page documents the changes in each Kimi CLI release.
 
 ## Unreleased
 
+- Shell: Use `/model` to toggle thinking mode instead of Tab key
+- Config: Add `default_thinking` config option (auto-migrated from metadata)
+- LLM: Add `always_thinking` capability for models that always use thinking mode
+
 ## 0.76 (2026-01-12)
 
 - Tool: Make `ReadFile` tool description reflect model capabilities for image/video support

docs/zh/configuration/config-files.md

@@ -25,6 +25,7 @@ kimi --config '{"default_model": "kimi-for-coding", "providers": {...}, "models"
 | 配置项 | 类型 | 说明 |
 | --- | --- | --- |
 | `default_model` | `string` | 默认使用的模型名称，必须是 `models` 中定义的模型 |
+| `default_thinking` | `boolean` | 默认是否开启 Thinking 模式（默认为 `false`） |
 | `providers` | `table` | API 供应商配置 |
 | `models` | `table` | 模型配置 |
 | `loop_control` | `table` | Agent 循环控制参数 |
@@ -35,6 +36,7 @@ kimi --config '{"default_model": "kimi-for-coding", "providers": {...}, "models"
 
 ```toml
 default_model = "kimi-for-coding"
+default_thinking = false
 
 [providers.kimi-for-coding]
 type = "kimi"

docs/zh/configuration/env-vars.md

@@ -53,7 +53,7 @@ export KIMI_MODEL_MAX_CONTEXT_SIZE="262144"
 
 ### `KIMI_MODEL_CAPABILITIES`
 
-覆盖配置文件中模型的 `capabilities` 字段。多个能力用逗号分隔，支持的值为 `thinking`、`image_in` 和 `video_in`。
+覆盖配置文件中模型的 `capabilities` 字段。多个能力用逗号分隔，支持的值为 `thinking`、`always_thinking`、`image_in` 和 `video_in`。
 
 ```sh
 export KIMI_MODEL_CAPABILITIES="thinking,image_in"

docs/zh/configuration/providers.md

@@ -108,7 +108,8 @@ env = { GOOGLE_CLOUD_PROJECT = "your-project-id" }
 
 | 能力 | 说明 |
 | --- | --- |
-| `thinking` | 支持 thinking 模式（深度思考） |
+| `thinking` | 支持 Thinking 模式（深度思考），可开关 |
+| `always_thinking` | 始终使用 Thinking 模式（不可关闭） |
 | `image_in` | 支持图片输入 |
 | `video_in` | 支持视频输入 |
 
@@ -122,7 +123,11 @@ capabilities = ["thinking", "image_in"]
 
 ### `thinking`
 
-启用 Thinking 模式后，模型会在回答前进行更深入的推理，适合复杂问题。在 Shell 模式下，可以通过 `Tab` 键切换 Thinking 模式，或在启动时通过 `--thinking` / `--no-thinking` 参数控制。
+声明模型支持 Thinking 模式。启用后，模型会在回答前进行更深入的推理，适合复杂问题。在 Shell 模式下，可以通过 `/model` 命令切换模型和 Thinking 模式，或在启动时通过 `--thinking` / `--no-thinking` 参数控制。
+
+### `always_thinking`
+
+表示模型始终使用 Thinking 模式，无法关闭。例如 `kimi-k2-thinking-turbo` 等名称中包含 "thinking" 的模型通常具有此能力。使用这类模型时，`/model` 命令不会提示选择 Thinking 模式的开关。
 
 ### `image_in`
 

docs/zh/faq.md

@@ -33,15 +33,6 @@ API 密钥无效可能的原因：
 - **使用 `--work-dir` 参数**：启动时指定工作目录，如 `kimi --work-dir /path/to/project`。
 - **在命令中使用绝对路径**：直接使用绝对路径执行命令，如 `ls /path/to/dir`。
 
-### Thinking 模式不可用
-
-按 `Tab` 键切换 Thinking 模式时，如果提示 "Thinking mode requires a model with thinking capability"，说明当前模型不支持 Thinking 模式。
-
-解决方法：
-
-- **切换到支持的模型**：使用 `/setup` 选择支持 Thinking 的模型（如 `kimi-k2-thinking-turbo`）。
-- **检查模型能力配置**：如果你确定模型支持 Thinking，可以通过环境变量 `KIMI_MODEL_CAPABILITIES=thinking` 强制启用。
-
 ### 粘贴图片失败
 
 使用 `Ctrl-V` 粘贴图片时，如果提示 "Current model does not support image input"，说明当前模型不支持图片输入。

docs/zh/guides/interaction.md

@@ -29,14 +29,14 @@ Shell 模式中每个命令独立执行，`cd`、`export` 等改变环境的命
 
 Thinking 模式让 AI 在回答前进行更深入的思考，适合处理复杂问题。
 
-在 Agent 模式下按 `Tab` 键可以切换 Thinking 模式的开关状态，切换后底部状态栏会显示提示。也可以在启动时通过 `--thinking` 参数启用：
+你可以通过 `/model` 命令切换模型和 Thinking 模式。在选择模型后，如果模型支持 Thinking 模式，系统会询问是否开启。也可以在启动时通过 `--thinking` 参数启用：
 
 ```sh
 kimi --thinking
 ```
 
 ::: tip 提示
-Thinking 模式需要当前模型支持。
+Thinking 模式需要当前模型支持。部分模型（如 `kimi-k2-thinking-turbo`）始终使用 Thinking 模式，无法关闭。
 :::
 
 ## 多行输入

docs/zh/reference/keyboard.md

@@ -7,7 +7,6 @@ Kimi CLI Shell 模式支持以下键盘快捷键。
 | 快捷键 | 功能 |
 |--------|------|
 | `Ctrl-X` | 切换 Agent/Shell 模式 |
-| `Tab` | 切换 Thinking 模式 |
 | `Ctrl-/` | 显示帮助 |
 | `Ctrl-J` | 插入换行 |
 | `Alt-Enter` | 插入换行（同 `Ctrl-J`） |
@@ -25,20 +24,9 @@ Kimi CLI Shell 模式支持以下键盘快捷键。
 - **Shell 模式**：输入作为本地 Shell 命令执行
 
 提示符会根据当前模式变化：
-- Agent 模式：`✨` 或 `💫`（thinking 模式）
+- Agent 模式：`✨`（普通）或 `💫`（Thinking 模式）
 - Shell 模式：`$`
 
-### `Tab`：切换 Thinking 模式
-
-在 Agent 模式下，当输入框为空或没有显示补全菜单时，按 `Tab` 可切换 Thinking 模式。
-
-- 启用 Thinking 模式时，提示符变为 `💫`
-- 切换时会显示提示信息
-
-::: tip 提示
-Thinking 模式需要模型支持。
-:::
-
 ### `Ctrl-/`：显示帮助
 
 在输入框中按 `Ctrl-/` 可快速显示帮助信息，等价于输入 `/help` 命令。

docs/zh/reference/slash-commands.md

@@ -43,11 +43,11 @@
 
 ### `/model`
 
-切换默认模型。
+切换模型和 Thinking 模式。
 
-此命令会先从 API 平台刷新可用模型列表。不带参数调用时，显示交互式选择界面；也可以直接指定模型名称，如 `/model kimi-k2`。
+此命令会先从 API 平台刷新可用模型列表。不带参数调用时，显示交互式选择界面，首先选择模型，然后选择是否开启 Thinking 模式（如果模型支持）。
 
-选择新模型后，Kimi CLI 会自动更新配置文件并重新加载。
+选择完成后，Kimi CLI 会自动更新配置文件并重新加载。
 
 ::: tip 提示
 此命令仅在使用默认配置文件时可用。如果通过 `--config` 或 `--config-file` 指定了配置，则无法使用此命令。

docs/zh/release-notes/changelog.md

@@ -4,6 +4,10 @@
 
 ## 未发布
 
+- Shell：使用 `/model` 命令切换 Thinking 模式，取代 Tab 键
+- Config：添加 `default_thinking` 配置项（从 metadata 自动迁移）
+- LLM：为始终使用 Thinking 模式的模型添加 `always_thinking` 能力
+
 ## 0.76 (2026-01-12)
 
 - Tool：让 `ReadFile` 工具描述根据模型能力动态反映图片/视频支持情况

packages/kosong/CHANGELOG.md

@@ -1,5 +1,9 @@
 # Changelog
 
+## Unreleased
+
+- Add `thinking_effort` property to `ChatProvider` protocol to query current thinking effort level
+
 ## 0.37.0 (2026-01-08)
 
 - Change `TokenUsage` from dataclass to pydantic BaseModel.

packages/kosong/src/kosong/chat_provider/__init__.py

@@ -23,6 +23,13 @@ def model_name(self) -> str:
         """
         ...
 
+    @property
+    def thinking_effort(self) -> "ThinkingEffort | None":
+        """
+        The current thinking effort level. Returns None if not explicitly set.
+        """
+        ...
+
     async def generate(
         self,
         system_prompt: str,

packages/kosong/src/kosong/chat_provider/chaos.py

@@ -175,6 +175,10 @@ def model_name(self) -> str:
             return f"chaos({self._provider.model_name})"
         return self._provider.model_name
 
+    @property
+    def thinking_effort(self) -> ThinkingEffort | None:
+        return self._provider.thinking_effort
+
     def with_thinking(self, effort: ThinkingEffort) -> "ChaosChatProvider":
         return ChaosChatProvider(self._provider.with_thinking(effort), self._chaos_config)
 

packages/kosong/src/kosong/chat_provider/echo.py

@@ -72,6 +72,10 @@ class EchoChatProvider:
     def model_name(self) -> str:
         return "echo"
 
+    @property
+    def thinking_effort(self) -> ThinkingEffort | None:
+        return None
+
     async def generate(
         self,
         system_prompt: str,

packages/kosong/src/kosong/chat_provider/kimi.py

@@ -98,6 +98,21 @@ def __init__(
     def model_name(self) -> str:
         return self.model
 
+    @property
+    def thinking_effort(self) -> ThinkingEffort | None:
+        reasoning_effort = self._generation_kwargs.get("reasoning_effort")
+        if reasoning_effort is None:
+            return None
+        match reasoning_effort:
+            case "low":
+                return "low"
+            case "medium":
+                return "medium"
+            case "high":
+                return "high"
+            case _:
+                return "off"
+
     async def generate(
         self,
         system_prompt: str,

packages/kosong/src/kosong/chat_provider/mock.py

@@ -36,6 +36,10 @@ def __init__(
     def model_name(self) -> str:
         return "mock"
 
+    @property
+    def thinking_effort(self) -> ThinkingEffort | None:
+        return None
+
     async def generate(
         self,
         system_prompt: str,

packages/kosong/src/kosong/chat_provider/openai_common.py

@@ -44,6 +44,18 @@ def thinking_effort_to_reasoning_effort(effort: ThinkingEffort) -> ReasoningEffo
             return "high"
 
 
+def reasoning_effort_to_thinking_effort(effort: ReasoningEffort) -> ThinkingEffort:
+    match effort:
+        case "low" | "minimal":
+            return "low"
+        case "medium":
+            return "medium"
+        case "high" | "xhigh":
+            return "high"
+        case "none" | None:
+            return "off"
+
+
 def tool_to_openai(tool: Tool) -> ChatCompletionToolParam:
     """Convert a single tool to OpenAI tool format."""
     # simply `model_dump` because the `Tool` type is OpenAI-compatible