Kimi Code @moonshot-ai/kimi-code@0.6.0

### 次要更新 - [#212](https://github.com/MoonshotAI/kimi-code/pull/212) [`2bbea75`](https://github.com/MoonshotAI/kimi-code/commit/2bbea75ee4c0b11f12d2921061774426df40479a) - 新增 `KIMI_MODEL_*` 环境变量通道，允许您在不编辑 `config.toml` 的情况下，针对特定模型（提供商类型、基础 URL、API 密钥、上下文大小、能力及思考设置）运行 Kimi Code。 - [#221](https://github.com/MoonshotAI/kimi-code/pull/221) [`bab2da7`](https://github.com/MoonshotAI/kimi-code/commit/bab2da7b1c785d6deba25decb1411f8f5a70de8c) - 支持直接从 GitHub 仓库 URL 安装插件，并在插件管理器中显示每次安装的来源和信任级别（kimi-official、curated、third-party）。 - [#118](https://github.com/MoonshotAI/kimi-code/pull/118) [`8913440`](https://github.com/MoonshotAI/kimi-code/commit/891344054111a05171963cfa524ef749c2855321) - 支持在 `listSessions` 中通过 sessionId 或 workDir 查询会话，并在从不同工作目录恢复会话时显示友好的 `cd` 命令提示。 - [#186](https://github.com/MoonshotAI/kimi-code/pull/186) [`537cf20`](https://github.com/MoonshotAI/kimi-code/commit/537cf20d18b26d4238f963f793f8a8ef085ac97e) - 移除默认的每轮步骤限制（1000 步）。用户仍可在配置中设置 `max_steps_per_turn` 以应用自定义限制。 ### 补丁更新 - [#197](https://github.com/MoonshotAI/kimi-code/pull/197) [`f3269ea`](https://github.com/MoonshotAI/kimi-code/commit/f3269eacb9da9a6b66f578a864d0b9bdfb1d6d81) - 在对话记录中显示后台代理的实际终端状态，使丢失、失败和已终止的代理不再显示为已完成；并在失败通知中包含恢复代理 ID 和恢复说明，以便模型能够可靠地恢复。 - [#211](https://github.com/MoonshotAI/kimi-code/pull/211) [`54590d3`](https://github.com/MoonshotAI/kimi-code/commit/54590d3d464b05eed0837a725b37f3aa491c09af) - 对失败的压缩重试操作采用固定模型上下文窗口片段的退避策略。 - [#167](https://github.com/MoonshotAI/kimi-code/pull/167) [`b5981a5`](https://github.com/MoonshotAI/kimi-code/commit/b5981a523b66ff2fd5f09a7e66075628b94683c8) - 引入 `ModelProvider` 接口和 `SingleModelProvider`，以解耦 `Agent` 与 `ProviderManager`。 - [#213](https://github.com/MoonshotAI/kimi-code/pull/213) [`2388f20`](https://github.com/MoonshotAI/kimi-code/commit/2388f20bb3d039e89caefca159801059b90dc64a) - 统一处理各提供商响应中的上下文溢出错误。 - [#214](https://github.com/MoonshotAI/kimi-code/pull/214) [`caaa6d8`](https://github.com/MoonshotAI/kimi-code/commit/caaa6d83ee262ba4c954386458ee13aacdb26e1a) - 修复原生自更新器在安装命令实际失败时仍报告更新成功的问题。 - [#202](https://github.com/MoonshotAI/kimi-code/pull/202) [`14a0348`](https://github.com/MoonshotAI/kimi-code/commit/14a03488555682dde4bcd74aadf79f60a9827304) - 修复在恢复不存在的会话时页脚溢出到终端的问题。 - [#198](https://github.com/MoonshotAI/kimi-code/pull/198) [`8c77cfa`](https://github.com/MoonshotAI/kimi-code/commit/8c77cfab62617e07b38f8514a8ef7cddfd9f1069) - 修复当临时文件位于其他文件系统时自动安装 ripgrep 的问题。 - [#199](https://github.com/MoonshotAI/kimi-code/pull/199) [`588145d`](https://github.com/MoonshotAI/kimi-code/commit/588145dc9b266456bdb1d739975a5b9cf33d70ae) - 扩展页脚中的轮换提示，展示更多命令和快捷键，并更突出地显示较新和重要的内容。 - [#192](https://github.com/MoonshotAI/kimi-code/pull/192) [`64964a0`](https://github.com/MoonshotAI/kimi-code/commit/64964a0dda98fc2db5e15ba923ea9414c78e0009) - 改进 TUI 中的使用信息展示。 - [#195](https://github.com/MoonshotAI/kimi-code/pull/195) [`3a0e060`](https://github.com/MoonshotAI/kimi-code/commit/3a0e06031ac6dfde148f64906a06cfe820ad9c63) - 将项目持久化钩子（hook）和受阻提示消息（blocked prompt messages）注入到模型上下文中。 - [#221](https://github.com/MoonshotAI/kimi-code/pull/221) [`bab2da7`](https://github.com/MoonshotAI/kimi-code/commit/bab2da7b1c785d6deba25decb1411f8f5a70de8c) - 将插件信任徽章限制为 Kimi 托管的插件 CDN URL 模式。 - [#207](https://github.com/MoonshotAI/kimi-code/pull/207) [`e280f33`](https://github.com/MoonshotAI/kimi-code/commit/e280f33daf7fbf1271c872dcb224737ec9518f73) - 在长对话期间恢复提供商模型令牌限制错误。 - [#201](https://github.com/MoonshotAI/kimi-code/pull/201) [`3da4dae`](https://github.com/MoonshotAI/kimi-code/commit/3da4daeadee39573c7eeede30fa9465b411be3e2) - 当模型响应流在传输过程中中断（`terminated` 错误）时自动重试，而不是直接导致本轮失败。 - [#190](https://github.com/MoonshotAI/kimi-code/pull/190) [`1873859`](https://github.com/MoonshotAI/kimi-code/commit/1873859b0ef093a956dfd19e1530e920e7118160) - 精简 LLM 诊断日志，减少字段数量并使其更紧凑。 - [#185](https://github.com/MoonshotAI/kimi-code/pull/185) [`114777e`](https://github.com/MoonshotAI/kimi-code/commit/114777e859680f807375760271533e2dc396af5d) - 将 `RuntimeConfig` 拆分为 `Kaos` 和 `ToolServices`，并相应更新所有引用。 - [#189](https://github.com/MoonshotAI/kimi-code/pull/189) [`564721f`](https://github.com/MoonshotAI/kimi-code/commit/564721fe16e582b2774835b01dec799cbb1d0122) - 明确子代理和后台任务的停止消息为“用户主动发起”。 - [#206](https://github.com/MoonshotAI/kimi-code/pull/206) [`07d51e4`](https://github.com/MoonshotAI/kimi-code/commit/07d51e4add6ee23a56fb8745aa7754f05f3d6d36) - 将共享工具服务类型定义迁移至工具支持层。 - [#215](https://github.com/MoonshotAI/kimi-code/pull/215) [`b9860e9`](https://github.com/MoonshotAI/kimi-code/commit/b9860e9f6ec65eb5dfdabbad54f1a87d69f4f00a) - 使数据源插件与通用的双工具工作流保持一致。 - [#200](https://github.com/MoonshotAI/kimi-code/pull/200) [`5159af3`](https://github.com/MoonshotAI/kimi-code/commit/5159af341c7d388a158e41afb470a2281333f329) - 使被拦截的 prompt hook 对话在后续模型轮次中保持可用。

### Minor Changes - [#212](https://github.com/MoonshotAI/kimi-code/pull/212) [`2bbea75`](https://github.com/MoonshotAI/kimi-code/commit/2bbea75ee4c0b11f12d2921061774426df40479a) - Add a `KIMI_MODEL_*` environment-variable channel that lets you run Kimi Code against a specific model (provider type, base URL, API key, context size, capabilities, and thinking settings) without editing `config.toml`. - [#221](https://github.com/MoonshotAI/kimi-code/pull/221) [`bab2da7`](https://github.com/MoonshotAI/kimi-code/commit/bab2da7b1c785d6deba25decb1411f8f5a70de8c) - Install plugins directly from GitHub repository URLs, and surface each install's origin and trust level (kimi-official, curated, third-party) in the plugin manager. - [#118](https://github.com/MoonshotAI/kimi-code/pull/118) [`8913440`](https://github.com/MoonshotAI/kimi-code/commit/891344054111a05171963cfa524ef749c2855321) - Support querying sessions by sessionId or workDir in listSessions, and show a helpful cd command when resuming a session from a different working directory. - [#186](https://github.com/MoonshotAI/kimi-code/pull/186) [`537cf20`](https://github.com/MoonshotAI/kimi-code/commit/537cf20d18b26d4238f963f793f8a8ef085ac97e) - Remove the default per-turn step limit of 1000. Users can still set `max_steps_per_turn` in config to enforce a custom limit. ### Patch Changes - [#197](https://github.com/MoonshotAI/kimi-code/pull/197) [`f3269ea`](https://github.com/MoonshotAI/kimi-code/commit/f3269eacb9da9a6b66f578a864d0b9bdfb1d6d81) - Show the real terminal status of background agents in the transcript so lost, failed, and killed ones no longer appear as completed, and include the resume agent id and recovery instructions in the failure notification so the model can resume reliably. - [#211](https://github.com/MoonshotAI/kimi-code/pull/211) [`54590d3`](https://github.com/MoonshotAI/kimi-code/commit/54590d3d464b05eed0837a725b37f3aa491c09af) - Back off failed compaction retries by a fixed slice of the model context window. - [#167](https://github.com/MoonshotAI/kimi-code/pull/167) [`b5981a5`](https://github.com/MoonshotAI/kimi-code/commit/b5981a523b66ff2fd5f09a7e66075628b94683c8) - Introduce `ModelProvider` interface and `SingleModelProvider` to decouple `Agent` from `ProviderManager`. - [#213](https://github.com/MoonshotAI/kimi-code/pull/213) [`2388f20`](https://github.com/MoonshotAI/kimi-code/commit/2388f20bb3d039e89caefca159801059b90dc64a) - Handle context overflow errors consistently across provider responses. - [#214](https://github.com/MoonshotAI/kimi-code/pull/214) [`caaa6d8`](https://github.com/MoonshotAI/kimi-code/commit/caaa6d83ee262ba4c954386458ee13aacdb26e1a) - Fix the native self-updater reporting a successful update when the install command actually failed. - [#202](https://github.com/MoonshotAI/kimi-code/pull/202) [`14a0348`](https://github.com/MoonshotAI/kimi-code/commit/14a03488555682dde4bcd74aadf79f60a9827304) - Fix footer leaking onto the terminal when resuming a non-existent session. - [#198](https://github.com/MoonshotAI/kimi-code/pull/198) [`8c77cfa`](https://github.com/MoonshotAI/kimi-code/commit/8c77cfab62617e07b38f8514a8ef7cddfd9f1069) - Fix automatic ripgrep installation when temporary files are on another filesystem. - [#199](https://github.com/MoonshotAI/kimi-code/pull/199) [`588145d`](https://github.com/MoonshotAI/kimi-code/commit/588145dc9b266456bdb1d739975a5b9cf33d70ae) - Expand the footer's rotating tips to surface more commands and shortcuts, featuring newer and important ones more prominently. - [#192](https://github.com/MoonshotAI/kimi-code/pull/192) [`64964a0`](https://github.com/MoonshotAI/kimi-code/commit/64964a0dda98fc2db5e15ba923ea9414c78e0009) - Improve the usage information display in the TUI. - [#195](https://github.com/MoonshotAI/kimi-code/pull/195) [`3a0e060`](https://github.com/MoonshotAI/kimi-code/commit/3a0e06031ac6dfde148f64906a06cfe820ad9c63) - Project persisted hook and blocked prompt messages into model context. - [#221](https://github.com/MoonshotAI/kimi-code/pull/221) [`bab2da7`](https://github.com/MoonshotAI/kimi-code/commit/bab2da7b1c785d6deba25decb1411f8f5a70de8c) - Restrict plugin trust badges to Kimi-hosted plugin CDN URL patterns. - [#207](https://github.com/MoonshotAI/kimi-code/pull/207) [`e280f33`](https://github.com/MoonshotAI/kimi-code/commit/e280f33daf7fbf1271c872dcb224737ec9518f73) - Recover from provider model token limit errors during long conversations. - [#201](https://github.com/MoonshotAI/kimi-code/pull/201) [`3da4dae`](https://github.com/MoonshotAI/kimi-code/commit/3da4daeadee39573c7eeede30fa9465b411be3e2) - Automatically retry when a model response stream is dropped mid-flight (a `terminated` error) instead of failing the turn. - [#190](https://github.com/MoonshotAI/kimi-code/pull/190) [`1873859`](https://github.com/MoonshotAI/kimi-code/commit/1873859b0ef093a956dfd19e1530e920e7118160) - Slim the LLM diagnostic logs with fewer, more compact fields. - [#185](https://github.com/MoonshotAI/kimi-code/pull/185) [`114777e`](https://github.com/MoonshotAI/kimi-code/commit/114777e859680f807375760271533e2dc396af5d) - Split `RuntimeConfig` into `Kaos` and `ToolServices` and update all references accordingly. - [#189](https://github.com/MoonshotAI/kimi-code/pull/189) [`564721f`](https://github.com/MoonshotAI/kimi-code/commit/564721fe16e582b2774835b01dec799cbb1d0122) - Clarify subagent and background task stop messages as user-initiated. - [#206](https://github.com/MoonshotAI/kimi-code/pull/206) [`07d51e4`](https://github.com/MoonshotAI/kimi-code/commit/07d51e4add6ee23a56fb8745aa7754f05f3d6d36) - Relocate shared tool service typing to the tool support layer. - [#215](https://github.com/MoonshotAI/kimi-code/pull/215) [`b9860e9`](https://github.com/MoonshotAI/kimi-code/commit/b9860e9f6ec65eb5dfdabbad54f1a87d69f4f00a) - Align the datasource plugin with the generic two-tool workflow. - [#200](https://github.com/MoonshotAI/kimi-code/pull/200) [`5159af3`](https://github.com/MoonshotAI/kimi-code/commit/5159af341c7d388a158e41afb470a2281333f329) - Keep blocked prompt hook conversations available to subsequent model turns.

Kimi Code @moonshot-ai/[email protected]