fix(config): 供应商改名同步重键模型别名 + 删除时清理关联配置（v0.8.3-rc1）

修复「改名后模型别名保留旧前缀，僵尸条目删了又出现」这一类配置不一致。

- 改名时同步重键模型别名、全局默认模型、供应商记住的默认模型、子代理池条目
- 改名冲突（撞已存在供应商名 / 遗留裸别名）加 -2/-3 后缀，不再静默吞模型
- 修复复制供应商复用移动语义 rekey 导致源供应商模型被清空
- 加载时不再合并「供应商已不存在」的 SQLite 僵尸模型
- 删除供应商/模型时清理子代理池条目、池默认回落、default_model 置空、
  SQLite 用量查询配置（不误删另一 agent，读取失败即跳过）
- 防呆：复制供应商改为 dirty-only，放弃修改不再落盘
- 测试基建：统一 crate 级 test_state 锁，消除环境变量并发竞争

测试：前端 146（新增 16）、Rust 197（新增 4）

.github/workflows/release.yml

@@ -74,6 +74,7 @@ jobs:
         shell: bash
         run: |
           FILE="docs/release-notes/${{ github.ref_name }}.md"
+          [ -f "$FILE" ] || FILE="docs/release-notes/release-notes-${{ github.ref_name }}.md"
           [ -f "$FILE" ] || FILE="release-notes-${{ github.ref_name }}.md"
           {
             echo 'body<<RELEASE_BODY_EOF'

CHANGELOG.md

@@ -6,6 +6,53 @@
 
 ---
 
+## v0.8.3-rc1 (2026-10-10)
+
+配置一致性修复预发布版（RC，请先本地验证）。
+
+- **供应商改名时模型别名同步重键**：此前只改模型的 `provider` 字段，别名键保留旧前缀（如 `opencode-go-copy/...`）形成僵尸条目；现在模型别名、全局默认模型、供应商记住的默认模型、子代理池条目全部同步更新
+- **僵尸模型不再复活**：加载时 SQLite 镜像会把 config.toml 中缺失的模型合并回来，导致已删模型保存后重现；现在供应商已不存在的模型不再合并
+- **改名冲突不再静默吞模型**：改名撞已存在的供应商名或遗留裸别名时，目标别名自动加 `-2`/`-3` 后缀
+- **复制供应商不再清空源供应商**：修复复制复用改名逻辑（移动语义）导致源供应商模型被搬走；副本独立、源完整
+- **删除供应商/模型的关联清理**：子代理模型池中引用被删模型的条目一并移除；池默认模型自动回落到剩余条目（池清空则移除整节）；`default_model` 置空；SQLite 中该供应商的用量查询配置随之清理，且不误删另一 agent（Pi）的设置
+- **防呆：放弃修改不落盘**：复制供应商不再立即静默保存，与其他编辑一致，放弃后配置保持原样
+- 修复测试基建中 `KIMI_CODE_HOME` / `KIMI_SWITCH_DB_PATH` 的并发竞争（统一 crate 级共享锁）
+- models.dev 快照同步至 2026-10-10（8,456 模型 / 226 供应商）
+- 测试：前端 146 项（新增 16 项）、Rust 197 项（新增 4 项）
+
+## v0.8.2 (2026-10-09)
+
+性能专项，功能与数据口径不变（用量数字、字段、IPC 接口、导出格式均与 v0.8.1 一致）。
+
+- **仪表盘重启后首次打开：约 22~27 秒 → 约 0.85 秒**（实测 6.4 万条记录 / 1.03 GB）。用量解析结果持久化到 `~/.kimi-switch/scan-cache-*.json`，启动直读快照渲染，后台线程全量校验，有变化时事件推送无感更新（先快后准）
+- **切换时间范围几乎瞬时**：热力图 / 模型总览 / 各范围汇总按记录集缓存；并去掉原「全时段」聚合的整表无用遍历
+- **界面不再闪加载态**：切范围与重进页面先呈现旧数据再静默刷新（stale-while-revalidate）
+- **供应商卡片墙变快**：供应商配置读取缓存（含外部 CLI 改动失效路径），N 张卡片由 N 次全量加载降为 1 次
+- **SQLite 连接复用**：连接与迁移改为每库一次，卡片墙场景由 200+ 次降为 1 次
+- **models.dev 快照更新**：同步至 2026-10-09（8,467 模型 / 225 供应商）
+
+## v0.8.1 (2026-10-04)
+
+- **全局配置新增两个顶层开关**：自动生成会话标题（`auto_session_title`）与重复调用拦截（`repeat_breaker`），默认开启，关闭显式写 `false`、开启删键跟随上游默认（适配 kimi-code #3962 / #3995）
+- **最大输出未匹配 models.dev 时回填默认值 131072**（128K 兜底，编辑页回填与获取模型列表预填同口径）
+- **修复 `[watch] enabled` 默认值显示**：上游 #4015 已翻回默认开启，设置页同步修正
+- **「最大输出 Token」列名精简为「最大输出」**，描述对齐上游 #4091（留空即不发送输出上限）
+
+## v0.8.0 (2026-10-03)
+
+- **模型映射新增「最大输出 Token」列**：逐模型设置 `max_output_size`（写入 config.toml，留空用上游默认），models.dev 有 output 上限时显示可点击参考值一键填入；获取模型列表时自动预填
+- **models.dev 快照新增 output 上限字段**：同步至 2026-10-03（8,385 模型 / 226 供应商，output 覆盖 97.4%）；应用内同步（Rust 侧）同步支持，避免同步后字段丢失
+- **移除无效的「声明支持 1M」复选框**：该控件读时由上下文长度派生、写时不持久化，勾选从不生效，由「最大输出 Token」列取代
+
+## v0.7.23 (2026-09-30)
+
+- **思考等级五档 + 能力感知**：全局配置的思考等级扩展为低 / 中 / 高 / Max / XHigh，按当前默认模型声明的 `support_efforts` 自动置灰不支持档位并提示原因；models.dev 标记不支持思考的模型整体禁用思考区；未声明档位的模型保留可选并提示"上游可能拒绝"
+- **修正过时适配**：删除 max→high 读取迁移（kimi-code 2.1.1 中 max 为合法档位），存储的 max 与任意手写档位值原样保留显示；子代理页继承档位显示同步修正
+
+## v0.7.22 (2026-09-24)
+
+- **修复 Sub2API / NewAPI 模板查询不显示**：为无预设识别类型的自定义供应商配置 Sub2API（或 NewAPI）模板后，弹窗测试可查但供应商列表始终不显示——前端列表的查询/显示门槛只看预设 `usageKinds`，显式模板被漏掉；现在与后端口径对齐（模板查询绕过 usage_kinds），列表正常显示余额 / 配额
+
 ## v0.7.21 (2026-09-24)
 
 - **models.dev 在线同步**：高级设置新增「模型参考数据」卡片——一键在线拉取 models.dev 最新快照（上下文长度 / 能力标记 / 单价），前端与仪表盘计价即时生效，无需等待新版本；支持一键恢复随版本内置数据；显示当前数据来源与版本（同步 / 内置 + 日期 + 模型/供应商数）

README.md

@@ -79,7 +79,7 @@
 | **连通性测试** | 实测 `base_url` 延迟，绿/橙/红彩色气泡，6 秒自动消失 |
 | **重复供应商** | 一键深拷贝供应商 + 全部模型，key 改 `xxx-copy` |
 | **图标按钮操作** | 启用 / 编辑 / 复制 / 测试连通 / 删除 全图标化（lucide-react） |
-| **模型映射** | 别名（"provider/model" 形式）↔ 实际请求模型 ID，自定义显示名、上下文长度、1M 上下文声明、能力 |
+| **模型映射** | 别名（"provider/model" 形式）↔ 实际请求模型 ID，自定义显示名、上下文长度、最大输出 Token（`max_output_size`）、能力（仅 Kimi Code） |
 | **自动上下文** | 拉取模型时 **API 返回 > models.dev ref > 正则兜底** 三级优先级自动适配 |
 | **能力自动推导** | `image_in / video_in / tool_use` 全部由 models.dev 推得；UI 仅暴露 `thinking` 一个手动开关 |
 | **全局设置** | `[thinking]` 表完整支持（enabled / effort / keep），仅 Kimi Code 生效 |
@@ -111,7 +111,7 @@
 
 ![编辑供应商-模型映射](docs/screenshots/provider-model-mapping.png)
 
-一张表管理全部模型映射：显示名、实际请求模型、上下文长度、1M 上下文声明、能力（仅"思考"）、设为默认、删除。
+一张表管理全部模型映射：显示名、实际请求模型、上下文长度、最大输出 Token（`max_output_size`，可点参考值填 models.dev 上限）、能力（仅"思考"）、设为默认、删除（后两项仅 Kimi Code）。
 
 **用量仪表盘**
 

README_EN.md

@@ -73,7 +73,7 @@ Build instructions: [`docs/BUILD.md`](./docs/BUILD.md).
 | **Connectivity test** | Real `base_url` latency, coloured bubble (green / orange / red), auto-dismiss in 6 seconds |
 | **Duplicate provider** | Deep-copy a provider + all its models; key auto-suffixed to `xxx-copy` |
 | **Iconified actions** | Activate / Edit / Duplicate / Test Connectivity / Delete via lucide-react |
-| **Model mapping** | Alias (`"provider/model"`) ↔ real model ID, with display name, context size, 1M-context flag, and capabilities |
+| **Model mapping** | Alias (`"provider/model"`) ↔ real model ID, with display name, context size, max output tokens (`max_output_size`), and capabilities (Kimi Code only) |
 | **Auto context size** | On model fetch: **API response > models.dev ref > regex fallback** — three-tier priority |
 | **Auto capabilities** | `image_in / video_in / tool_use` all derived from models.dev; UI only exposes `thinking` as a manual toggle |
 | **Global settings** | Full `[thinking]` table (enabled / effort / keep); Kimi Code only |
@@ -105,7 +105,7 @@ Provider name, notes, official URL, managed-provider toggle, API format, API key
 
 ![Edit provider — model mapping](docs/screenshots/provider-model-mapping.png)
 
-A single table for all model mappings: display name, real model ID, context length, 1M-context flag, capability (thinking only), default toggle, delete.
+A single table for all model mappings: display name, real model ID, context length, max output tokens (`max_output_size`, with a click-to-fill models.dev reference), capability (thinking only), default toggle, delete (the last two are Kimi Code only).
 
 **Usage dashboard**
 
@@ -175,7 +175,7 @@ Per-workspace Kimi Code session browsing, active / archived / all filters; strea
 - **`config.toml` is the authoritative source for Kimi Code**: all providers and models are always written in full; `default_model` selects the active one (matching the CLI's native `/provider` behaviour). Switching only changes `default_model`; newly added providers are auto-promoted to the top of the list and never get overwritten
 - **SQLite holds Kimi Switch-private metadata only**: notes, official URLs, per-agent remembered default model (the `settings` table), ordering. Theme / language / last-update-check live in frontend `localStorage` (WebView2), not under `~/.kimi-switch`. It acts as a fallback when `config.toml` is incomplete
 - **`raw_other` passes unknown fields through untouched**, including `[oauth]` blocks — round-trips never drop fields
-- **models.dev snapshot**: derived from `https://models.dev/api.json`, cached to a local JSON; `capabilitiesFromRef` derives `thinking / image_in / video_in / tool_use`, `getModelRef` derives `max_context_size / display_name`
+- **models.dev snapshot**: derived from `https://models.dev/api.json`, cached to a local JSON; `capabilitiesFromRef` derives `thinking / image_in / video_in / tool_use`, `getModelRef` derives `max_context_size / display_name / max output tokens`
 - **Override env vars**: `KIMI_CODE_HOME` / `PI_CODING_AGENT_DIR` override the Kimi Code / Pi dirs; Kimi Switch's own data dir is fixed at `~/.kimi-switch` (no env override yet). See [Data Storage Locations](#data-storage-locations)
 
 ## Feature Details

docs/LOCAL-VERIFY-fixes.md

@@ -0,0 +1,81 @@
+# 本地验证清单 — 配置一致性修复（待发 v0.8.3）
+
+构建版本：0.8.2（含未提交改动）。自动化测试已全绿（前端 146 / Rust 197），
+以下是**必须人工验证**的场景，重点是引擎侧对配置结构的实际反应。
+
+验证前备份：`~/.kimi-code/config.toml`、`~/.kimi-switch/kimi-switch.db`
+
+---
+
+## 0. 前置：确认基线干净
+
+- [ ] 打开 `~/.kimi-code/config.toml`，搜索 `opencode-go-copy` / `opencode-go-1` → 应为 0 命中
+- [ ] KimiSwitch 供应商列表中 `opencode-go` 正常显示，模型 3 条（longcat / mimo / step-5）
+- [ ] 高级设置 → 子代理模型池中 3 条 opencode-go 条目都在
+- [ ] kimi-code 会话能正常启动（`kimi` 命令），确认 `default_model` 生效
+
+---
+
+## 1. 复制供应商（本次修复的严重回归点）
+
+- [ ] 供应商列表 → `opencode-go` 点「复制」
+- [ ] **源供应商 `opencode-go` 的 3 个模型仍在**（这是关键：旧实现会清空）
+- [ ] 副本 `opencode-go-copy` 也有 3 个模型，别名带 `-copy` 前缀
+- [ ] 副本默认状态为「未启用 / 未使用中」
+- [ ] 点返回 → 弹出未保存确认 → 选「放弃」→ **列表里副本消失**（验证防呆：放弃不落盘）
+- [ ] 重新复制 → Ctrl+S 保存 → 重启应用 → 副本仍在（验证保存生效）
+- [ ] 删除副本 → Ctrl+S → 重启 → 副本消失，`opencode-go` 完好
+
+## 2. 供应商改名（僵尸别名根源）
+
+- [ ] 编辑 `opencode-go-copy` → 名称改为 `opencode-renamed` → 保存 → 返回
+- [ ] `config.toml` 中该供应商的模型别名变为 `opencode-renamed/*`，**无 `opencode-go-copy/*` 残留**
+- [ ] 全局默认模型、该供应商记住的默认模型都指向新别名
+- [ ] 子代理池中若有副本的条目，键同步变为新别名（描述保留）
+- [ ] kimi-code 会话能启动，默认模型正确
+
+## 3. 删除供应商（关联清理）
+
+- [ ] 先把某个模型加入子代理池，并设为池默认
+- [ ] 删除该供应商 → Ctrl+S → 打开 `config.toml` 检查：
+  - [ ] 该供应商的 `[models.*]` 段落全部消失
+  - [ ] `[secondary_model.models]` 中引用它的条目已消失
+  - [ ] 池 `default_model` / `model` 已回落到剩余条目（不是悬空）
+  - [ ] 池被清空时整个 `[secondary_model]` 节消失
+  - [ ] `default_model` 若指向被删模型则为 `""` 或不再指向它
+  - [ ] **kimi-code 会话能正常启动**（重点：验证引擎对 `default_model = ""` 的容忍度）
+- [ ] SQLite 检查：`~/.kimi-switch/kimi-switch.db` 的 `settings` 表中
+      `usage_kinds:<该供应商名>` / `usage_config:<该供应商名>` 行已删除
+- [ ] 无关的 settings 行（如 `app.language`）未受影响
+
+## 4. 删除单个模型
+
+- [ ] 把某模型加入子代理池 → 编辑该供应商 → 删除该模型 → 保存
+- [ ] `config.toml` 中 `[secondary_model.models]` 对应条目消失
+- [ ] 池默认为被删模型时已回落到其他条目
+
+## 5. 防呆：放弃修改不落盘
+
+- [ ] 编辑供应商改名 → 点返回 → 选「放弃」→ 重新打开：名称为旧名，**config.toml 未变**
+- [ ] 删除模型 → 点返回 → 放弃 → 重新打开：模型仍在
+- [ ] 复制供应商 → 放弃 → 副本消失
+- [ ] 任何改动后不点保存就关闭应用 → 重启后配置不变
+
+## 6. Pi agent 不受牵连
+
+- [ ] 切到 Pi agent 面板，确认其供应商与用量查询配置正常
+- [ ] 切回 KimiCode 保存一次 → 再看 Pi 的用量查询设置，应仍在
+- [ ] 检查 SQLite `settings` 表：Pi 供应商的 `usage_*` 行未被误删
+
+## 7. 防呆：改名冲突
+
+- [ ] 存在 `opencode-go` 和 `opencode-go-copy` 时，把后者改名为 `opencode-go` 并保存
+- [ ] 结果：两个供应商条目与各自模型都保留，冲突方别名带 `-2` 后缀，**无静默丢模型**
+
+---
+
+## 已知遗留（不属于本次范围）
+
+- config.toml 中若存在 provider 已删除的僵尸 `[models.*]` 段落，应用不会自动删除
+  （用户权威文件，只防复发），需手动清理一次
+- `default_model = ""` 时 kimi-code CLI 的确切回退行为，建议实测确认

docs/release-notes/release-notes-v0.7.22.md

@@ -0,0 +1,9 @@
+# KimiSwitch v0.7.22
+
+## 修复
+
+- **Sub2API / NewAPI 模板查询在供应商列表不显示**：为没有预设识别类型的自定义供应商（如自家中转站）配置「Sub2API 中转站」或「NewAPI 中转站」模板后，配置弹窗里「测试查询」正常返回余额，但回到供应商列表却始终空白。根因：列表的查询与显示门槛只认预设注入的 `usageKinds`，而后端模板查询本就绕过 `usageKinds` 直接执行——前端口径与后端不一致。现在二者对齐，显式模板的供应商在列表正常显示余额 / 配额窗口、支持自动查询间隔与预警阈值。
+
+## 备注
+
+- 仅前端修复（`useUsageQuery` 支持判定 + 卡片渲染门槛），后端查询逻辑无变化；已补 3 个单测覆盖支持判定。

docs/release-notes/release-notes-v0.7.23.md

@@ -0,0 +1,18 @@
+# KimiSwitch v0.7.23
+
+## 新功能
+
+- **思考等级五档 + 能力感知**：全局配置的思考等级由「低 / 中 / 高」三档扩展为「低 / 中 / 高 / Max / XHigh」五档，并按当前默认模型的实际能力自适应——
+  - 模型在 config.toml 声明了 `support_efforts` 时，不支持的档位自动置灰并给出原因提示（如 deepseek-v4.1-flash 仅支持「高」）
+  - models.dev 标记不支持思考（reasoning=false）的模型，整个思考等级区禁用并给出警告
+  - 未声明档位的模型保留全部可选，Max / XHigh 附带「上游可能拒绝此档位」提示
+  - 等级行下方新增说明：实际生效档位取决于当前默认模型
+
+## 修复
+
+- **max 档位被误显示为「高」**：此前代码基于过时信息（"上游已移除 max 档位"）把存储的 `effort = "max"` 读取时强制改为 `high`；实测 kimi-code 2.1.1 中 max 为合法档位（如 kimi-for-coding 默认即为 max），现已原样保留与显示。手写的任意档位值（上游本就是自由字符串）也会作为额外可选项正常显示，不再被丢弃
+- **子代理设置页**：继承档位显示同步修正，max / xhigh 正常显示，未知值按原样展示
+
+## 备注
+
+- 仅前端改动（新增 `thinking-efforts` 能力判定模块 + `Segmented` 控件逐选项禁用/tooltip），无 Rust 变更；新增 14 个单测覆盖三层能力判定规则，全量 117 个测试通过

docs/release-notes/release-notes-v0.8.0.md

@@ -0,0 +1,14 @@
+# KimiSwitch v0.8.0
+
+## 新功能
+
+- **模型映射表新增「最大输出 Token」列**（Kimi Code）：为每个模型单独设置 `max_output_size`，写入 config.toml 的 `[models."<alias>"]`，留空则用上游默认值。models.dev 收录了该模型的输出上限时，输入框下方显示可点击的参考值（如 `参考 128000`），一键填入；「获取模型列表」添加模型时也会自动预填参考值
+- **models.dev 快照新增 output 上限字段**：内置快照同步至 2026-10-03（8,385 模型 / 226 供应商，97.4% 模型带 output 上限）；应用内「同步 models.dev」（Rust 侧）同步支持该字段，同步后参考值不丢失
+
+## 修复
+
+- **移除无效的「声明支持 1M」复选框**（Kimi Code）：该控件在读取时由上下文长度派生、写入时不持久化，勾选永远不会生效，已由「最大输出 Token」列取代
+
+## 备注
+
+- 前端 + 抓取脚本 + Rust 快照构建逻辑；新增 `max_output_size` 读写纯函数 9 个单测 + config.toml 全链路往返测试，全量测试通过（前端 126 / Rust 162）

docs/release-notes/release-notes-v0.8.1.md

@@ -0,0 +1,17 @@
+# KimiSwitch v0.8.1
+
+## 新功能
+
+- **全局配置新增两个顶层开关**（适配 kimi-code 近期版本）：
+  - **自动生成会话标题**（`auto_session_title`，上游 #3962）：允许客户端自动生成会话标题，默认开启；关闭后显式写入 `false`，开启即删键跟随上游默认
+  - **重复调用拦截**（`repeat_breaker`，上游 #3995）：连续重复相同工具调用时提醒并强制停止，默认开启；可被环境变量 `KIMI_CODE_REPEAT_BREAKER` 覆盖
+- **最大输出未匹配时回填默认值 131072**（Kimi Code）：打开供应商编辑页自动回填时，models.dev 未收录的模型按 128K（131072）兜底；「获取模型列表」新增模型同口径预填
+
+## 修复
+
+- **`[watch] enabled` 默认值修正为开启**：kimi-code #4015 已把文件监听默认值翻回开启（2.0.2 曾短暂默认关闭），设置页的开关显示默认值同步修正
+- **「最大输出 Token」列名精简为「最大输出」**；列描述修正为「留空则不发送输出上限」（对齐上游 #4091：未显式配置时不发送 `max_tokens`，避免严格服务栈如 bare vLLM 报 400）
+
+## 备注
+
+- 前端 130 测试 / Rust 162 测试全量通过

docs/release-notes/release-notes-v0.8.2.md

@@ -0,0 +1,21 @@
+# KimiSwitch v0.8.2
+
+本版为**性能专项**：把主要操作的等待时间从「秒级」压到「感知不到」，功能与数据口径完全不变。
+
+## 性能优化
+
+- **仪表盘重启后首次打开：约 22~27 秒 → 约 0.85 秒**（实测 1065 个 `wire.jsonl` / 1.03 GB / 6.4 万条记录）。用量解析结果持久化到本地快照（`~/.kimi-switch/scan-cache-*.json`），启动时直接读快照渲染，后台线程同时做全量校验；校验发现新记录时自动推送事件、无感更新数字（先快后准，无需手动刷新）
+- **切换时间范围几乎瞬时**：热力图、模型总览、各范围汇总结果按记录集与日期缓存，来回切换不再重复全量聚合；未命中时也不再做无用的整表遍历（原「全时段」统计会为取一张模型表而白算日/明细/供应商三份数据）
+- **界面不再闪加载态**：仪表盘切范围与重新进入页面时先呈现上次数据再后台静默刷新，不再出现数据已经拿过却先显示骨架屏的情况
+- **供应商卡片墙显著变快**：供应商配置读取结果加缓存（含敏感字段写入与外部 CLI 改动两条失效路径），N 张卡片从 N 次全量读库 + 配置解析降为 1 次
+- **SQLite 连接复用**：库连接与建表迁移从「每次调用各做一遍」改为每库一次，卡片墙场景下由 200+ 次降为 1 次
+
+## 其他
+
+- **models.dev 快照更新**：同步至 2026-10-09（8,467 模型 / 225 供应商），价格与上下文长度随之刷新
+
+## 备注
+
+- 数据口径与功能行为与 v0.8.1 完全一致：用量数字、字段、IPC 接口、导出格式均无变化；「强制刷新」仍会忽略一切缓存重新全量解析
+- 测试全量通过：Rust 193 项（新增 16 项，覆盖缓存失效、磁盘快照往返等价、损坏快照降级等）、前端 130 项
+- 首次使用本版本时会重建一次磁盘快照（沿用原有全量扫描耗时），此后每次启动均为快速加载

docs/release-notes/release-notes-v0.8.3-rc1.md

@@ -0,0 +1,43 @@
+# KimiSwitch v0.8.3-rc1
+
+配置一致性修复版：解决「供应商改名后模型别名没跟着改，产生的僵尸条目删了又出现」这一类问题，并补齐删除供应商时的关联清理与防呆。
+
+## 修复
+
+- **供应商改名时模型别名同步重键**：此前改名只更新模型的 `provider` 字段，模型别名键仍保留旧前缀（如 `opencode-go-copy/...`），形成永远指向新供应商的僵尸条目；现在别名、全局默认模型、该供应商记住的默认模型、子代理池条目全部同步更新
+- **僵尸模型不再复活**：应用加载时会把 SQLite 镜像中缺失于 config.toml 的模型合并回来（迁移保护），导致已删除的模型在保存后重新出现；现在供应商已不存在的模型不再被合并
+- **改名冲突不再静默吞模型**：改名为已存在的供应商名、或遗留裸别名与新名冲突时，目标别名自动加 `-2`/`-3` 后缀，与「应用 JSON」路径的既有规则一致
+- **复制供应商不再清空源供应商**：修复复制时复用改名逻辑导致源供应商模型被整体搬走的问题；副本为独立副本，源供应商模型完整保留
+
+## 删除供应商 / 模型时的关联清理
+
+- 子代理模型池（`[secondary_model]`）中引用被删模型的条目一并移除
+- 池默认模型指向被删模型时自动回落到池中剩余条目；池被清空时整个 `[secondary_model]` 节移除
+- 全局 `default_model` 指向被删模型时置空
+- SQLite 中该供应商的用量查询配置（`usage_kinds:` / `usage_config:`）随之清理；保留另一 agent（Pi）的设置，读取失败时跳过清理而非误删
+- 模型占位符自动改名为真实模型 id 时，池条目同步跟随
+
+## 防呆
+
+- **放弃修改不落盘**：复制供应商不再立即静默保存，与其他编辑一致——点「返回」并放弃后配置保持原样
+- 改名/删除/JSON 应用等所有编辑路径统一遵守「必须保存才写盘」
+
+## 其他
+
+- 修复测试基建中进程级环境变量（`KIMI_CODE_HOME` / `KIMI_SWITCH_DB_PATH`）的并发竞争：dashboard 与 oauth 测试此前各自持锁，改为统一的 crate 级共享锁
+- models.dev 快照同步至 2026-10-10（8,456 模型 / 226 供应商）
+
+## 测试
+
+- 前端 146 项（新增 16 项：改名重键、复制保留源、冲突去重、子代理池清理与空表防护）
+- Rust 197 项（新增 4 项：删除后双存储持久化与不复活、孤儿配置修剪、跨 agent 修剪保护）
+
+## 已知限制
+
+- config.toml 中若已存在「供应商已删除」的僵尸 `[models.*]` 段落，升级后不会自动删除（config.toml 是用户权威文件，本版只防复发），需在界面中手动删除一次
+- 删除供应商后 `default_model` 置空时的 kimi-code CLI 回退行为，建议实测确认
+
+## 下载
+
+- **Windows**：`.exe`（NSIS 安装包，本版仅提供）
+- **macOS / Linux**：见 [Releases](https://github.com/billowliu2/KimiSwitch/releases)（由 CI 构建）

package.json

@@ -1,7 +1,7 @@
 {
   "name": "kimiswitch",
   "private": true,
-  "version": "0.7.21",
+  "version": "0.8.3-rc1",
   "type": "module",
   "scripts": {
     "dev": "vite",

scripts/fetch-models-dev.mjs

@@ -133,6 +133,11 @@ async function main() {
       if (typeof m.limit?.context === "number" && m.limit.context > 0) {
         entry.context = m.limit.context;
       }
+      // Max output tokens — the reference value behind the max_output_size
+      // column (model mapping table).
+      if (typeof m.limit?.output === "number" && m.limit.output > 0) {
+        entry.output = m.limit.output;
+      }
       if (m.reasoning === true) entry.reasoning = true;
       if (m.tool_call === true) entry.tool_call = true;
       if (m.structured_output === true) entry.structured_output = true;

scripts/publish-gitea.sh

@@ -17,6 +17,7 @@ TAG="${1:?usage: publish-gitea.sh <tag> <asset> [asset...]}"
 shift
 
 NOTES_FILE="docs/release-notes/${TAG}.md"
+[ -f "$NOTES_FILE" ] || NOTES_FILE="docs/release-notes/release-notes-${TAG}.md"
 [ -f "$NOTES_FILE" ] || NOTES_FILE="release-notes-${TAG}.md"
 json_escape() {
   python - "$1" <<'PYEOF'

src-tauri/Cargo.lock

@@ -1978,7 +1978,7 @@ dependencies = [
 
 [[package]]
 name = "kimiswitch"
-version = "0.7.21"
+version = "0.8.3-rc1"
 dependencies = [
  "anyhow",
  "chrono",

src-tauri/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "kimiswitch"
-version = "0.7.21"
+version = "0.8.3-rc1"
 description = "Kimi Switch - model config manager"
 authors = ["codingplan.site"]
 edition = "2021"

src-tauri/src/commands.rs

@@ -6,7 +6,7 @@ use indexmap::IndexMap;
 use serde::Serialize;
 use std::collections::HashMap;
 use std::sync::{Mutex, OnceLock};
-use std::time::{Duration, Instant};
+use std::time::{Duration, Instant, UNIX_EPOCH};
 use tauri::Manager;
 use tauri_plugin_opener::OpenerExt;
 
@@ -79,8 +79,61 @@ fn load_pi_native_config() -> Result<Config, String> {
     Ok(config)
 }
 
+/// Cached `load_agent_config_command` results, keyed by agent. The full load
+/// parses config.toml, reads all of SQLite and runs `merge_usage_kinds` (two
+/// settings reads per provider); a usage-card wall triggers one load per
+/// provider card, so caching turns N loads into one. Entries are validated
+/// against the db generation counter (bumped by every write path in db.rs)
+/// plus the config.toml file stamp (catches edits made outside the app, e.g.
+/// the CLI's /provider command).
+static CONFIG_CACHE: Mutex<Option<HashMap<String, (u64, Option<(u64, u64)>, Config)>>> =
+    Mutex::new(None);
+
+/// (mtime_ms, len) of the authoritative on-disk config file for the agent, or
+/// None when the file is absent / the agent has no authoritative file (Pi is
+/// SQLite-first, so the db generation alone covers it).
+fn config_file_stamp(agent: &Agent) -> Option<(u64, u64)> {
+    match agent {
+        Agent::KimiCode => {
+            let md = std::fs::metadata(crate::kimi_code_io::kimi_code_config_path()).ok()?;
+            let mtime_ms = md
+                .modified()
+                .ok()?
+                .duration_since(UNIX_EPOCH)
+                .ok()?
+                .as_millis() as u64;
+            Some((mtime_ms, md.len()))
+        }
+        Agent::Pi => None,
+    }
+}
+
 #[tauri::command]
 pub fn load_agent_config_command(agent: Agent) -> Result<Config, String> {
+    let stamp = config_file_stamp(&agent);
+    let gen = db::generation();
+    let key = agent.as_str().to_string();
+
+    {
+        let guard = CONFIG_CACHE.lock().unwrap_or_else(|e| e.into_inner());
+        if let Some(map) = guard.as_ref() {
+            if let Some((cached_gen, cached_stamp, config)) = map.get(&key) {
+                if *cached_gen == gen && *cached_stamp == stamp {
+                    return Ok(config.clone());
+                }
+            }
+        }
+    }
+
+    let config = load_agent_config_uncached(agent)?;
+    let mut guard = CONFIG_CACHE.lock().unwrap_or_else(|e| e.into_inner());
+    guard
+        .get_or_insert_with(HashMap::new)
+        .insert(key, (gen, stamp, config.clone()));
+    Ok(config)
+}
+
+fn load_agent_config_uncached(agent: Agent) -> Result<Config, String> {
     // Load Kimi Switch's own SQLite database (metadata + migration fallback).
     let db_config = db::load_config(&agent).ok();
 
@@ -122,11 +175,18 @@ pub fn load_agent_config_command(agent: Agent) -> Result<Config, String> {
                 }
                 // Migration safety: include providers/models that exist in
                 // SQLite but not in config.toml (e.g. after upgrading from the
-                // old single-provider-write behaviour).
+                // old single-provider-write behaviour). Models whose provider
+                // no longer exists anywhere are skipped — they can never be
+                // shown or selected in the UI, and merging them back would
+                // resurrect zombie `[models."old-name/..."]` entries into
+                // config.toml on the next save.
                 for (name, p) in &db.providers {
                     config.providers.entry(name.clone()).or_insert_with(|| p.clone());
                 }
                 for (alias, m) in &db.models {
+                    if !config.providers.contains_key(&m.provider) {
+                        continue;
+                    }
                     config.models.entry(alias.clone()).or_insert_with(|| m.clone());
                 }
             }
@@ -177,6 +237,25 @@ pub fn save_agent_config_command(agent: Agent, config: Config) -> Result<(), Str
             None => db::delete_setting_pub(&cfg_key).map_err(fmt_anyhow)?,
         }
     }
+    // Prune usage settings of providers that no longer exist (deleted in the
+    // UI). The keys are agent-agnostic, so the live set includes the other
+    // agent's providers too. If the other agent's config cannot be read we
+    // skip this round entirely — pruning against an incomplete live set
+    // could delete the other agent's still-live settings.
+    {
+        let other_agent = match agent {
+            Agent::KimiCode => Agent::Pi,
+            Agent::Pi => Agent::KimiCode,
+        };
+        let other_cfg = match db::load_config(&other_agent) {
+            Ok(cfg) => cfg,
+            Err(_) => return Ok(()),
+        };
+        let mut live: std::collections::HashSet<String> =
+            config.providers.keys().cloned().collect();
+        live.extend(other_cfg.providers.keys().cloned());
+        db::prune_orphan_usage_settings(&live).map_err(fmt_anyhow)?;
+    }
     Ok(())
 }
 
@@ -2314,3 +2393,272 @@ mod usage_key_tests {
         assert_eq!(resolve_usage_api_key(&p).unwrap(), None);
     }
 }
+
+/// End-to-end persistence tests for the Kimi Code agent: a UI edit followed
+/// by save_agent_config_command must land in BOTH stores (config.toml and
+/// SQLite) and must not resurrect deleted models on the next load.
+#[cfg(test)]
+mod config_persistence_tests {
+    use super::*;
+
+    /// Serialize env mutation with every other env-mutating test in the
+    /// binary (shared crate-level lock; see lib.rs `test_state`).
+    fn with_env_homes<T>(f: impl FnOnce(&std::path::Path) -> T) -> T {
+        let _guard = crate::test_state::lock();
+        let home = tempfile::tempdir().unwrap();
+        let db_dir = tempfile::tempdir().unwrap();
+        std::env::set_var("KIMI_CODE_HOME", home.path());
+        std::env::set_var("KIMI_SWITCH_DB_PATH", db_dir.path().join("test.db"));
+        db::close_cached_conn_for_tests();
+        let out = f(home.path());
+        db::close_cached_conn_for_tests();
+        std::env::remove_var("KIMI_CODE_HOME");
+        std::env::remove_var("KIMI_SWITCH_DB_PATH");
+        out
+    }
+
+    const TOML_WITH_STALE_MODELS: &str = r#"
+default_model = "opencode-go/longcat-2.5-preview-free"
+
+[providers.opencode-go]
+type = "openai"
+base_url = "https://opencode.ai/zen/go/v1"
+api_key = "sk-test"
+
+[models."opencode-go-copy/deepseek-v4-flash"]
+provider = "opencode-go"
+model = "deepseek-v4-flash"
+max_context_size = 512000
+
+[models."opencode-go-1/space-bunny-free"]
+provider = "opencode-go"
+model = "space-bunny-free"
+max_context_size = 512000
+
+[models."opencode-go/longcat-2.5-preview-free"]
+provider = "opencode-go"
+model = "longcat-2.5-preview-free"
+max_context_size = 1000000
+"#;
+
+    #[test]
+    fn ui_delete_then_save_persists_to_both_stores_and_never_resurrects() {
+        with_env_homes(|home| {
+            std::fs::write(home.join("config.toml"), TOML_WITH_STALE_MODELS).unwrap();
+
+            // Load from the seeded config.toml, then mirror it into SQLite to
+            // reproduce the user's real state: stale aliases in BOTH stores.
+            let mut config = load_agent_config_uncached(Agent::KimiCode).unwrap();
+            assert!(config.models.contains_key("opencode-go-copy/deepseek-v4-flash"));
+            assert!(config.models.contains_key("opencode-go-1/space-bunny-free"));
+            db::save_config(&Agent::KimiCode, &config).unwrap();
+
+            // Simulate the UI's model delete (onModelDelete): drop the stale
+            // aliases from the in-memory config, then save.
+            config.models.retain(|alias, _| {
+                !alias.starts_with("opencode-go-copy/") && !alias.starts_with("opencode-go-1/")
+            });
+            save_agent_config_command(Agent::KimiCode, config).unwrap();
+
+            // config.toml: stale sections gone, live model kept.
+            let toml_text = std::fs::read_to_string(home.join("config.toml")).unwrap();
+            assert!(!toml_text.contains("opencode-go-copy"));
+            assert!(!toml_text.contains("opencode-go-1"));
+            assert!(toml_text.contains("opencode-go/longcat-2.5-preview-free"));
+
+            // SQLite mirror: stale rows gone.
+            let db_config = db::load_config(&Agent::KimiCode).unwrap();
+            assert!(!db_config.models.contains_key("opencode-go-copy/deepseek-v4-flash"));
+            assert!(!db_config.models.contains_key("opencode-go-1/space-bunny-free"));
+
+            // Reload through the real load path: no resurrection from SQLite.
+            let reloaded = load_agent_config_uncached(Agent::KimiCode).unwrap();
+            assert!(!reloaded.models.contains_key("opencode-go-copy/deepseek-v4-flash"));
+            assert!(!reloaded.models.contains_key("opencode-go-1/space-bunny-free"));
+            assert!(reloaded.models.contains_key("opencode-go/longcat-2.5-preview-free"));
+        });
+    }
+
+    #[test]
+    fn sqlite_models_whose_provider_was_deleted_do_not_resurrect() {
+        with_env_homes(|home| {
+            // config.toml already clean: only the live provider + model.
+            let toml = r#"
+default_model = "opencode-go/longcat-2.5-preview-free"
+
+[providers.opencode-go]
+type = "openai"
+base_url = "https://opencode.ai/zen/go/v1"
+api_key = "sk-test"
+
+[models."opencode-go/longcat-2.5-preview-free"]
+provider = "opencode-go"
+model = "longcat-2.5-preview-free"
+max_context_size = 1000000
+"#;
+            std::fs::write(home.join("config.toml"), toml).unwrap();
+
+            // Seed SQLite with a zombie model whose provider no longer exists
+            // (left behind by an old rename-then-delete).
+            let mut providers = IndexMap::new();
+            providers.insert(
+                "opencode-go".to_string(),
+                Provider {
+                    name: "opencode-go".to_string(),
+                    provider_type: ProviderType::Openai,
+                    base_url: Some("https://opencode.ai/zen/go/v1".to_string()),
+                    api_key: Some("sk-test".to_string()),
+                    api_key_env: None,
+                    env: IndexMap::new(),
+                    note: None,
+                    official_url: None,
+                    managed: false,
+                    enabled: true,
+                    active: true,
+                    icon: None,
+                    icon_color: None,
+                    raw_other: serde_json::Value::Null,
+                    usage_kinds: None,
+                    usage_config: None,
+                },
+            );
+            let mut models = IndexMap::new();
+            models.insert(
+                "opencode-go-copy/deepseek-v4-flash".to_string(),
+                Model {
+                    alias: "opencode-go-copy/deepseek-v4-flash".to_string(),
+                    provider: "opencode-go-copy".to_string(),
+                    model: "deepseek-v4-flash".to_string(),
+                    max_context_size: 512_000,
+                    display_name: None,
+                    supports_1m: false,
+                    capabilities: vec![],
+                    raw_other: serde_json::Value::Null,
+                },
+            );
+            let seeded = Config {
+                default_model: Some("opencode-go/longcat-2.5-preview-free".to_string()),
+                providers,
+                models,
+                raw_other: serde_json::Value::Null,
+                imported_section_keys: vec![],
+            };
+            db::save_config(&Agent::KimiCode, &seeded).unwrap();
+
+            // Loading must NOT merge the zombie back: its provider is gone,
+            // so it can never be shown or selected, and merging it would
+            // rewrite it into config.toml on the next save.
+            let loaded = load_agent_config_uncached(Agent::KimiCode).unwrap();
+            assert!(!loaded.models.contains_key("opencode-go-copy/deepseek-v4-flash"));
+            assert!(loaded.models.contains_key("opencode-go/longcat-2.5-preview-free"));
+
+            // And a save after that load must keep config.toml clean.
+            save_agent_config_command(Agent::KimiCode, loaded).unwrap();
+            let toml_text = std::fs::read_to_string(home.join("config.toml")).unwrap();
+            assert!(!toml_text.contains("opencode-go-copy"));
+        });
+    }
+
+    fn plain_provider(name: &str) -> Provider {
+        Provider {
+            name: name.to_string(),
+            provider_type: ProviderType::Openai,
+            base_url: Some("https://opencode.ai/zen/go/v1".to_string()),
+            api_key: Some("sk-test".to_string()),
+            api_key_env: None,
+            env: IndexMap::new(),
+            note: None,
+            official_url: None,
+            managed: false,
+            enabled: true,
+            active: true,
+            icon: None,
+            icon_color: None,
+            raw_other: serde_json::Value::Null,
+            usage_kinds: None,
+            usage_config: None,
+        }
+    }
+
+    #[test]
+    fn deleting_a_provider_prunes_its_orphan_usage_settings() {
+        with_env_homes(|home| {
+            std::fs::write(home.join("config.toml"), TOML_WITH_STALE_MODELS).unwrap();
+            let mut config = load_agent_config_uncached(Agent::KimiCode).unwrap();
+
+            // Provider present: its usage settings survive the save.
+            config.providers.get_mut("opencode-go").unwrap().usage_kinds =
+                Some(vec!["plan:opencode_go".to_string()]);
+            save_agent_config_command(Agent::KimiCode, config.clone()).unwrap();
+            assert_eq!(
+                db::get_setting_pub("usage_kinds:opencode-go").unwrap().as_deref(),
+                Some(r#"["plan:opencode_go"]"#)
+            );
+
+            // Seed orphans the way history did (renamed/deleted providers).
+            db::set_setting_pub("usage_kinds:opencode-go-copy", r#"["plan:opencode_go"]"#)
+                .unwrap();
+            db::set_setting_pub("usage_config:opencode-go-1", "{}").unwrap();
+            // Non-usage settings must never be touched by the prune.
+            db::set_setting_pub("app.language", "zh").unwrap();
+
+            // Delete the provider (UI semantics) and save.
+            config.providers.shift_remove("opencode-go");
+            config.models.retain(|_, m| m.provider != "opencode-go");
+            config.default_model = None;
+            save_agent_config_command(Agent::KimiCode, config).unwrap();
+
+            // The provider's own settings and the historical orphans are gone.
+            assert_eq!(db::get_setting_pub("usage_kinds:opencode-go").unwrap(), None);
+            assert_eq!(db::get_setting_pub("usage_kinds:opencode-go-copy").unwrap(), None);
+            assert_eq!(db::get_setting_pub("usage_config:opencode-go-1").unwrap(), None);
+            assert_eq!(db::get_setting_pub("app.language").unwrap().as_deref(), Some("zh"));
+
+            // Re-adding a provider with the same name starts clean (no
+            // resurrection of the pruned rows).
+            let mut config = load_agent_config_uncached(Agent::KimiCode).unwrap();
+            config
+                .providers
+                .insert("opencode-go".to_string(), plain_provider("opencode-go"));
+            save_agent_config_command(Agent::KimiCode, config).unwrap();
+            assert_eq!(db::get_setting_pub("usage_kinds:opencode-go").unwrap(), None);
+            assert_eq!(db::get_setting_pub("app.language").unwrap().as_deref(), Some("zh"));
+        });
+    }
+
+    #[test]
+    fn pruning_never_touches_the_other_agents_usage_settings() {
+        with_env_homes(|home| {
+            std::fs::write(home.join("config.toml"), TOML_WITH_STALE_MODELS).unwrap();
+
+            // Pi (the other agent) has its own provider with usage settings.
+            let mut pi_providers = IndexMap::new();
+            pi_providers.insert("pi-provider".to_string(), plain_provider("pi-provider"));
+            db::save_config(
+                &Agent::Pi,
+                &Config {
+                    default_model: None,
+                    providers: pi_providers,
+                    models: IndexMap::new(),
+                    raw_other: serde_json::Value::Null,
+                    imported_section_keys: vec![],
+                },
+            )
+            .unwrap();
+            db::set_setting_pub("usage_kinds:pi-provider", r#"["balance:deepseek"]"#)
+                .unwrap();
+            db::set_setting_pub("usage_config:pi-provider", "{}").unwrap();
+
+            // Save the Kimi Code config (which has no `pi-provider`).
+            let config = load_agent_config_uncached(Agent::KimiCode).unwrap();
+            save_agent_config_command(Agent::KimiCode, config).unwrap();
+
+            // The other agent's settings survive the prune.
+            assert_eq!(
+                db::get_setting_pub("usage_kinds:pi-provider").unwrap().as_deref(),
+                Some(r#"["balance:deepseek"]"#)
+            );
+            assert_eq!(db::get_setting_pub("usage_config:pi-provider").unwrap().as_deref(), Some("{}"));
+        });
+    }
+}

src-tauri/src/db.rs

@@ -6,6 +6,8 @@
 //! a provider.
 
 use std::path::PathBuf;
+use std::sync::atomic::{AtomicU64, Ordering};
+use std::sync::Mutex;
 
 use anyhow::Context;
 use indexmap::IndexMap;
@@ -17,6 +19,51 @@ use crate::models::{Agent, Config, Model, Provider, ProviderType};
 
 pub type DbResult<T> = anyhow::Result<T>;
 
+/// Process-wide cached connection, keyed by the resolved db path. Opening a
+/// SQLite connection plus running the DDL migrations costs several ms; before
+/// this cache every public helper paid that cost on every call (a usage-card
+/// wall could open 200+ connections per refresh). `Connection` is `Send` but
+/// not `Sync`, so a `Mutex` is the correct `static` wrapper — all db access is
+/// serialized, which is fine for a single-user desktop app.
+static DB_CONN: Mutex<Option<(PathBuf, Connection)>> = Mutex::new(None);
+
+/// Monotonic generation counter, bumped by every write path. Read-side caches
+/// (config cache, archived-session cache) key on this to detect staleness.
+static DB_GEN: AtomicU64 = AtomicU64::new(0);
+
+pub(crate) fn generation() -> u64 {
+    DB_GEN.load(Ordering::Relaxed)
+}
+
+pub(crate) fn bump_generation() {
+    DB_GEN.fetch_add(1, Ordering::Relaxed);
+}
+
+/// Run `f` against the cached connection for the current `db_path()`, opening
+/// (and migrating) it on first use or whenever the path changed (tests point
+/// `KIMI_SWITCH_DB_PATH` at per-test temp dirs).
+fn with_conn<T>(f: impl FnOnce(&mut Connection) -> DbResult<T>) -> DbResult<T> {
+    let mut guard = DB_CONN.lock().unwrap_or_else(|e| e.into_inner());
+    let path = db_path();
+    let needs_open = match guard.as_ref() {
+        Some((cached_path, _)) => cached_path != &path,
+        None => true,
+    };
+    if needs_open {
+        *guard = Some((path, init_db()?));
+    }
+    let (_, conn) = guard.as_mut().expect("connection just opened");
+    f(conn)
+}
+
+/// Drop the cached connection so tests that switch `KIMI_SWITCH_DB_PATH`
+/// between temp dirs never observe cross-test state.
+#[cfg(test)]
+pub(crate) fn close_cached_conn_for_tests() {
+    let mut guard = DB_CONN.lock().unwrap_or_else(|e| e.into_inner());
+    *guard = None;
+}
+
 pub fn kimi_switch_data_dir() -> PathBuf {
     dirs::home_dir()
         .map(|h| h.join(".kimi-switch"))
@@ -158,7 +205,10 @@ pub fn init_db() -> DbResult<Connection> {
 }
 
 pub fn load_config(agent: &Agent) -> DbResult<Config> {
-    let mut conn = init_db()?;
+    with_conn(|conn| load_config_inner(conn, agent))
+}
+
+fn load_config_inner(conn: &mut Connection, agent: &Agent) -> DbResult<Config> {
     let tx = conn.transaction()?;
 
     let default_model = get_setting_tx(&tx, &default_model_key(agent))?;
@@ -249,7 +299,12 @@ pub fn load_config(agent: &Agent) -> DbResult<Config> {
 }
 
 pub fn save_config(agent: &Agent, config: &Config) -> DbResult<()> {
-    let mut conn = init_db()?;
+    with_conn(|conn| save_config_inner(conn, agent, config))?;
+    bump_generation();
+    Ok(())
+}
+
+fn save_config_inner(conn: &mut Connection, agent: &Agent, config: &Config) -> DbResult<()> {
     let tx = conn.transaction()?;
 
     tx.execute("DELETE FROM providers WHERE agent = ?1", params![agent.as_str()])?;
@@ -340,7 +395,7 @@ fn set_setting_tx(tx: &rusqlite::Transaction, key: &str, value: &str) -> DbResul
 
 /// Public helper: read a single setting without an explicit transaction.
 pub fn get_setting_pub(key: &str) -> DbResult<Option<String>> {
-    let conn = init_db()?;
+    with_conn(|conn| {
         let mut stmt = conn.prepare("SELECT value FROM settings WHERE key = ?1")?;
         let mut rows = stmt.query(params![key])?;
         if let Some(row) = rows.next()? {
@@ -348,22 +403,56 @@ pub fn get_setting_pub(key: &str) -> DbResult<Option<String>> {
         } else {
             Ok(None)
         }
+    })
 }
 
 /// Public helper: write a single setting in its own transaction.
 pub fn set_setting_pub(key: &str, value: &str) -> DbResult<()> {
-    let mut conn = init_db()?;
+    with_conn(|conn| {
         let tx = conn.transaction()?;
         set_setting_tx(&tx, key, value)?;
         tx.commit()?;
         Ok(())
+    })?;
+    bump_generation();
+    Ok(())
 }
 
 /// Public helper: delete a single setting (no-op if the key does not exist).
 pub fn delete_setting_pub(key: &str) -> DbResult<()> {
-    let conn = init_db()?;
+    with_conn(|conn| {
         conn.execute("DELETE FROM settings WHERE key = ?1", params![key])?;
         Ok(())
+    })?;
+    bump_generation();
+    Ok(())
+}
+
+/// Delete `usage_kinds:<name>` / `usage_config:<name>` settings rows whose
+/// provider no longer exists in any agent's config, so deleting a provider in
+/// the UI also clears its usage-query settings instead of leaving orphan rows
+/// that outlive every rename of that provider.
+pub fn prune_orphan_usage_settings(
+    live_providers: &std::collections::HashSet<String>,
+) -> DbResult<()> {
+    with_conn(|conn| {
+        let keys: Vec<String> = {
+            let mut stmt = conn.prepare(
+                "SELECT key FROM settings WHERE key LIKE 'usage_kinds:%' OR key LIKE 'usage_config:%'",
+            )?;
+            let rows = stmt.query_map([], |row| row.get::<_, String>(0))?;
+            rows.collect::<Result<Vec<_>, _>>()?
+        };
+        for key in keys {
+            let name = key.split_once(':').map(|(_, n)| n).unwrap_or("");
+            if !live_providers.contains(name) {
+                conn.execute("DELETE FROM settings WHERE key = ?1", params![key])?;
+            }
+        }
+        Ok(())
+    })?;
+    bump_generation();
+    Ok(())
 }
 
 // ---------------------------------------------------------------------------
@@ -413,7 +502,7 @@ pub struct ArchivedSessionSnapshot {
 /// Insert or replace one archived-session snapshot (keyed by session id, which
 /// is a UUID and never reused).
 pub fn upsert_archived_session(row: &ArchivedSessionSnapshot) -> DbResult<()> {
-    let conn = init_db()?;
+    with_conn(|conn| {
         let day_stats = serde_json::to_string(&row.day_stats)?;
         conn.execute(
             "INSERT OR REPLACE INTO archived_sessions
@@ -432,12 +521,15 @@ pub fn upsert_archived_session(row: &ArchivedSessionSnapshot) -> DbResult<()> {
             ],
         )?;
         Ok(())
+    })?;
+    bump_generation();
+    Ok(())
 }
 
 /// Every stored archived-session snapshot. Unreadable `day_stats` JSON degrades
 /// to an empty breakdown instead of failing the whole list.
 pub fn list_archived_sessions() -> DbResult<Vec<ArchivedSessionSnapshot>> {
-    let conn = init_db()?;
+    with_conn(|conn| {
         let mut stmt = conn.prepare(
             "SELECT session_id, workspace_id, title, archived_at_ms, updated_at_ms, created_at_ms,
                     total_tokens, total_cost_usd, day_stats
@@ -462,6 +554,7 @@ pub fn list_archived_sessions() -> DbResult<Vec<ArchivedSessionSnapshot>> {
             out.push(row?);
         }
         Ok(out)
+    })
 }
 
 fn provider_type_for_str(s: &str) -> ProviderType {

src-tauri/src/kimi_code_io.rs

@@ -648,6 +648,114 @@ api_key = ""
         assert!(caps.iter().any(|v| v.as_str() == Some("thinking")));
     }
 
+    #[test]
+    fn kimi_code_export_writes_and_drops_max_output_size() {
+        // max_output_size is not a first-class Model field: it round-trips
+        // through raw_other and must land in config.toml verbatim, and must be
+        // removable (unset = upstream default) without disturbing its siblings.
+        let mut providers = IndexMap::new();
+        providers.insert(
+            "p".to_string(),
+            Provider {
+                name: "p".to_string(),
+                provider_type: ProviderType::Openai,
+                base_url: Some("https://a.example.com".to_string()),
+                api_key: Some("sk-a".to_string()),
+                api_key_env: None,
+                env: IndexMap::new(),
+                note: None,
+                official_url: None,
+                managed: false,
+                enabled: true,
+                active: true,
+                icon: None,
+                icon_color: None,
+                raw_other: Value::Null,
+                usage_kinds: None,
+                usage_config: None,
+            },
+        );
+        let make_model = |raw_other: Value| Model {
+            alias: "p/m".to_string(),
+            provider: "p".to_string(),
+            model: "m".to_string(),
+            max_context_size: 128_000,
+            display_name: None,
+            supports_1m: false,
+            capabilities: vec![],
+            raw_other,
+        };
+
+        let mut models = IndexMap::new();
+        models.insert(
+            "p/m".to_string(),
+            make_model(serde_json::json!({
+                "max_output_size": 32768,
+                "support_efforts": ["low", "high"],
+            })),
+        );
+        let config = Config {
+            default_model: None,
+            providers,
+            models,
+            raw_other: Value::Null,
+            imported_section_keys: Vec::new(),
+        };
+
+        let exported = config_to_kimi_code(&config, None);
+        let root = exported.as_table().unwrap();
+        let model = root
+            .get("models")
+            .unwrap()
+            .as_table()
+            .unwrap()
+            .get("p/m")
+            .unwrap()
+            .as_table()
+            .unwrap();
+        assert_eq!(model.get("max_output_size").and_then(|v| v.as_integer()), Some(32768));
+        assert!(model.get("support_efforts").is_some());
+
+        // Re-import through real TOML text (the same serialize → parse path
+        // save_kimi_code_config / load_kimi_code_config use): the key must
+        // come back in the model's raw_other as an integer.
+        let toml_text = toml::to_string_pretty(&exported).unwrap();
+        assert!(toml_text.contains("max_output_size = 32768"), "toml: {toml_text}");
+        let reparsed: TomlValue = toml_text.parse().unwrap();
+        let parsed = kimi_code_to_config(&reparsed);
+        let loaded = parsed.models.get("p/m").unwrap();
+        assert_eq!(
+            loaded.raw_other.get("max_output_size").and_then(|v| v.as_i64()),
+            Some(32768)
+        );
+
+        // Clearing the override drops the key entirely.
+        let mut models = IndexMap::new();
+        models.insert(
+            "p/m".to_string(),
+            make_model(serde_json::json!({ "support_efforts": ["low", "high"] })),
+        );
+        let config = Config {
+            default_model: None,
+            providers: IndexMap::new(),
+            models,
+            raw_other: Value::Null,
+            imported_section_keys: Vec::new(),
+        };
+        let exported = config_to_kimi_code(&config, None);
+        let root = exported.as_table().unwrap();
+        let model = root
+            .get("models")
+            .unwrap()
+            .as_table()
+            .unwrap()
+            .get("p/m")
+            .unwrap()
+            .as_table()
+            .unwrap();
+        assert!(model.get("max_output_size").is_none());
+    }
+
     #[test]
     fn kimi_code_export_adds_opencode_go_session_header() {
         let make_provider = |base_url: &str, raw_other: Value| Provider {

src-tauri/src/lib.rs

@@ -10,6 +10,20 @@ pub mod pi_io;
 pub mod plugins;
 pub mod services;
 
+/// Process-global state guard for tests that redirect shared environment
+/// (`KIMI_CODE_HOME`, `KIMI_SWITCH_DB_PATH`) or in-memory caches (scan cache,
+/// cached DB connection). Every env-mutating test must hold this lock so a
+/// parallel test can never observe another one's temp home or redirected
+/// database.
+#[cfg(test)]
+pub(crate) mod test_state {
+    static STATE_LOCK: std::sync::Mutex<()> = std::sync::Mutex::new(());
+
+    pub(crate) fn lock() -> std::sync::MutexGuard<'static, ()> {
+        STATE_LOCK.lock().unwrap_or_else(|e| e.into_inner())
+    }
+}
+
 use tauri::menu::{Menu, MenuItem, PredefinedMenuItem};
 use tauri::tray::{MouseButton, MouseButtonState, TrayIconBuilder};
 use tauri::Manager;

src-tauri/src/models_dev.rs

@@ -138,6 +138,14 @@ fn build_snapshot(raw: &Value) -> Result<Value, String> {
             {
                 entry.insert("context".into(), num(ctx));
             }
+            // Max output tokens — reference for the max_output_size column.
+            if let Some(out) = m
+                .pointer("/limit/output")
+                .and_then(|x| x.as_f64())
+                .filter(|o| *o > 0.0)
+            {
+                entry.insert("output".into(), num(out));
+            }
             for flag in ["reasoning", "tool_call", "structured_output"] {
                 if m.get(flag).and_then(|x| x.as_bool()) == Some(true) {
                     entry.insert(flag.into(), Value::Bool(true));
@@ -286,7 +294,7 @@ mod tests {
                     },
                     "kimi-image": {
                         "name": "Kimi Image",
-                        "limit": { "context": 0 },
+                        "limit": { "context": 0, "output": 0 },
                         "modalities": { "input": ["text", "image", "video"] }
                     }
                 }
@@ -309,6 +317,7 @@ mod tests {
         let k3 = &obj["moonshotai/kimi-k3"];
         assert_eq!(k3["name"], "Kimi K3");
         assert_eq!(k3["context"], 262144);
+        assert_eq!(k3["output"], 16384);
         assert_eq!(k3["reasoning"], true);
         assert_eq!(k3["tool_call"], true);
         assert_eq!(k3["structured_output"], true);
@@ -318,9 +327,10 @@ mod tests {
         assert_eq!(k3["cost"]["cache_read"], 0.3);
         assert_eq!(k3["cost"]["cache_write"], 0.0);
 
-        // Context 0 → dropped; video modality → flagged.
+        // Context 0 / output 0 → dropped; video modality → flagged.
         let img = &obj["moonshotai/kimi-image"];
         assert!(img.get("context").is_none());
+        assert!(img.get("output").is_none());
         assert_eq!(img["video"], true);
         assert!(img.get("cost").is_none());
 

src-tauri/src/oauth.rs

@@ -925,8 +925,7 @@ mod tests {
     /// dir). A process-wide mutex serializes env mutation so parallel tests in
     /// this binary can't observe a stale override.
     fn with_kimi_code_home<T>(f: impl FnOnce(&std::path::Path) -> T) -> T {
-        static LOCK: std::sync::OnceLock<std::sync::Mutex<()>> = std::sync::OnceLock::new();
-        let _guard = LOCK.get_or_init(|| std::sync::Mutex::new(())).lock().unwrap();
+        let _guard = crate::test_state::lock();
         let home = tempfile::tempdir().unwrap();
         std::env::set_var("KIMI_CODE_HOME", home.path());
         let out = f(home.path());

src-tauri/tauri.conf.json

@@ -1,6 +1,6 @@
 {
   "productName": "Kimi Switch",
-  "version": "0.7.21",
+  "version": "0.8.3-rc1",
   "identifier": "com.kimiswitch.app",
   "build": {
     "beforeDevCommand": "npm run dev",

src/App.tsx

@@ -20,6 +20,14 @@ import {
   type ProviderPreset,
 } from "./config/providerPresets";
 import { validateProviders } from "./lib/validation";
+import {
+  copyModelsOnProviderDuplicate,
+  rekeyModelsOnProviderRename,
+} from "./lib/providerRename";
+import {
+  purgeSubagentPoolAliases,
+  rekeySubagentPoolAliases,
+} from "./lib/referenceCleanup";
 import type { Agent, Model, Provider, UsageConfig } from "./types";
 
 const AGENT_STORAGE_KEY = "kimi-switch-agent";
@@ -153,9 +161,6 @@ export default function App() {
   //      provider being switched to — only that one needs to be
   //      complete; sibling providers being merely listed in the config
   //      must not block the switch.
-  //   3. Programmatic auto-saves (duplicate) call `save` directly with
-  //      no validation — duplicates inherit completeness from their
-  //      source.
   // Returns `true` when the config was actually persisted, `false` when
   // the user cancelled the validation confirm (callers can react to this
   // to roll back in-memory mutations made before the save).
@@ -287,15 +292,30 @@ export default function App() {
       );
       if (oldEntry && oldEntry[0] !== provider.name) {
         delete providers[oldEntry[0]];
-        const models = { ...cfg.models };
-        for (const key of Object.keys(models)) {
-          if (models[key].provider === editingProvider) {
-            models[key] = { ...models[key], provider: provider.name };
+        // Re-key this provider's model aliases so the `<provider>/<model>`
+        // prefix tracks the new name (see rekeyModelsOnProviderRename).
+        const { models, aliasMap } = rekeyModelsOnProviderRename(
+          cfg.models,
+          editingProvider,
+          provider.name
+        );
+        // The provider's remembered default model and the global default
+        // model are stored as aliases and must follow the re-key.
+        const remembered = getProviderDefaultModel(provider);
+        providers[provider.name] =
+          remembered && aliasMap[remembered]
+            ? setProviderDefaultModel(provider, aliasMap[remembered])
+            : provider;
+        let default_model = cfg.default_model;
+        if (default_model && aliasMap[default_model]) {
+          default_model = aliasMap[default_model];
         }
-        }
-        providers[provider.name] = provider;
+        // The subagent pool references model aliases as its keys and
+        // defaults; they must follow the re-key or they turn into zombie
+        // pool entries.
+        const raw_other = rekeySubagentPoolAliases(cfg.raw_other, aliasMap);
         setEditingProvider(provider.name);
-        return { ...cfg, providers, models };
+        return { ...cfg, providers, models, default_model, raw_other };
       }
       providers[provider.name] = provider;
       return { ...cfg, providers };
@@ -308,12 +328,21 @@ export default function App() {
       const providers = { ...cfg.providers };
       delete providers[name];
       const models = { ...cfg.models };
+      const removedAliases: string[] = [];
       for (const key of Object.keys(models)) {
         if (models[key].provider === name) {
+          removedAliases.push(key);
           delete models[key];
         }
       }
-      return { ...cfg, providers, models };
+      // Everything keyed on the deleted aliases must go with them: the
+      // global default model and the subagent pool entries / pool default.
+      const default_model =
+        cfg.default_model && removedAliases.includes(cfg.default_model)
+          ? null
+          : cfg.default_model;
+      const raw_other = purgeSubagentPoolAliases(cfg.raw_other, removedAliases);
+      return { ...cfg, providers, models, default_model, raw_other };
     });
     if (editingProvider === name) {
       setEditingProvider("");
@@ -356,29 +385,14 @@ export default function App() {
         null
       );
       const providers = { ...cfg.providers, [newName]: copied };
-      // Re-key this provider's models to the new provider name so the copy
-      // has its own independent model set.
-      const models = { ...cfg.models };
-      for (const [alias, m] of Object.entries(cfg.models)) {
-        if (m.provider === name) {
-          // The model belongs to this provider by `m.provider === name`. To
-          // re-key it under the new provider, we want a `newName/...`
-          // alias. The original code assumed `alias` is exactly
-          // `${name}/${modelId}` and used `alias.slice(name.length)`; that
-          // breaks for legacy / non-standard aliases (e.g. `kimi-k3` from
-          // pre-v0.6 data) which would produce `newName-k3` — missing the
-          // `/` separator and pointing at the wrong model. Use an explicit
-          // prefix check and fall back to a full re-prefix.
-          const prefix = `${name}/`;
-          const newAlias = alias.startsWith(prefix)
-            ? newName + alias.slice(name.length)
-            : `${newName}/${alias}`;
-          models[newAlias] = { ...m, alias: newAlias, provider: newName };
-        }
-      }
+      // Clone the provider's models under the duplicate's name — COPY
+      // semantics: the source keeps its models (a rekey would move them).
+      const models = copyModelsOnProviderDuplicate(cfg.models, name, newName);
       return { ...cfg, providers, models };
     });
-    await save();
+    // No immediate save: the copy lives in memory only, like every other
+    // edit. Abandoning (back → discard) must leave the config unchanged;
+    // the copy persists when the user saves (Ctrl+S / footer / switch).
     setSwitchMessage(t("copiedProvider", { name: `${name}-copy` }));
     setTimeout(() => setSwitchMessage(null), 3000);
   };
@@ -529,6 +543,14 @@ export default function App() {
         default_model = null;
       }
 
+      // Pool entries pointing at aliases that disappeared in this apply
+      // (provider renamed, models dropped) must not survive as zombies.
+      const removedAliases = Object.keys(cfg.models).filter(
+        (key) =>
+          cfg.models[key].provider === oldName && !(key in updatedModels)
+      );
+      const raw_other = purgeSubagentPoolAliases(cfg.raw_other, removedAliases);
+
       // Remember this provider's default model preference so it survives
       // provider switches.
       const updatedProvider = providers[provider.name];
@@ -536,7 +558,7 @@ export default function App() {
         providers[provider.name] = setProviderDefaultModel(updatedProvider, default_model);
       }
 
-      return { ...cfg, providers, models: updatedModels, default_model };
+      return { ...cfg, providers, models: updatedModels, default_model, raw_other };
     });
     if (provider.name !== editingProvider) {
       setEditingProvider(provider.name);
@@ -744,16 +766,26 @@ export default function App() {
                 if (default_model === oldKey) default_model = effective.alias;
                 else if (effective !== model && default_model === model.alias)
                   default_model = effective.alias;
-                return { ...cfg, models, default_model };
+                // Pool entries keyed on the discarded alias (placeholder or
+                // replaced old key) must follow the rename or they dangle.
+                const poolAliasMap: Record<string, string> = {};
+                if (effective !== model) poolAliasMap[model.alias] = effective.alias;
+                if (oldKey && oldKey !== effective.alias)
+                  poolAliasMap[oldKey] = effective.alias;
+                const raw_other = rekeySubagentPoolAliases(cfg.raw_other, poolAliasMap);
+                return { ...cfg, models, default_model, raw_other };
               });
             }}
             onModelDelete={(alias) => {
               updateConfig((cfg) => {
                 const models = { ...cfg.models };
                 delete models[alias];
+                // Pool entries keyed on the deleted alias must not survive.
+                const raw_other = purgeSubagentPoolAliases(cfg.raw_other, [alias]);
                 return {
                   ...cfg,
                   models,
+                  raw_other,
                   default_model:
                     cfg.default_model === alias ? null : cfg.default_model,
                 };

src/components/AgentSettingsPanel.tsx

@@ -1,12 +1,20 @@
-import { useEffect, useState } from "react";
+import { useEffect, useReducer, useState } from "react";
 import { invoke } from "@tauri-apps/api/core";
 import { useTranslation } from "../i18n";
 import type { TranslationKey } from "../i18n/zh";
 import { getAgentSettings, setAgentSettings } from "../lib/agent-settings";
+import { getModelRef, modelsDevReady } from "../lib/models-dev";
+import {
+  THINKING_EFFORTS,
+  readSupportEfforts,
+  resolveThinkingEffortSupport,
+  type EffortAvailability,
+} from "../lib/thinking-efforts";
 import type {
   AgentSettings,
   ExperimentalEnvStatus,
   Hook,
+  Model,
   PermissionRule,
 } from "../types";
 import { Card, Checkbox, NumberField, Segmented } from "./ui/controls";
@@ -14,14 +22,18 @@ import { Card, Checkbox, NumberField, Segmented } from "./ui/controls";
 interface AgentSettingsPanelProps {
   rawOther: unknown;
   onChange: (nextRawOther: unknown) => void;
+  /** Configured models (alias → entry) — used to resolve effort support. */
+  models?: Record<string, Model>;
+  /** Alias of the model the effort tier actually applies to. */
+  defaultModel?: string | null;
 }
 
-// Upstream removed the "max" effort tier (auto-migrates to "high").
-const THINKING_LEVELS = ["low", "medium", "high"] as const;
-const THINKING_LABELS: Record<(typeof THINKING_LEVELS)[number], TranslationKey> = {
+const THINKING_LABELS: Record<(typeof THINKING_EFFORTS)[number], TranslationKey> = {
   low: "thinkingLow",
   medium: "thinkingMedium",
   high: "thinkingHigh",
+  max: "thinkingMax",
+  xhigh: "thinkingXHigh",
 };
 
 const PERMISSION_DECISIONS = ["allow", "deny", "ask"] as const;
@@ -44,7 +56,7 @@ const COMMON_EVENTS = [
   "SessionEnd",
 ] as const;
 
-export function AgentSettingsPanel({ rawOther, onChange }: AgentSettingsPanelProps) {
+export function AgentSettingsPanel({ rawOther, onChange, models, defaultModel }: AgentSettingsPanelProps) {
   const { t } = useTranslation();
   const settings = getAgentSettings(rawOther);
   /**
@@ -52,6 +64,9 @@ export function AgentSettingsPanel({ rawOther, onChange }: AgentSettingsPanelPro
    * v1-engine users; the default (v2) writes only max_attempts_per_step.
    */
   const [legacyV1, setLegacyV1] = useState(false);
+  // The models.dev index loads in the background; re-render once it lands so
+  // the reasoning flag can disable the thinking area.
+  const [, forceModelsDevReady] = useReducer((x: number) => x + 1, 0);
 
   useEffect(() => {
     invoke<ExperimentalEnvStatus>("get_experimental_env_status")
@@ -62,6 +77,18 @@ export function AgentSettingsPanel({ rawOther, onChange }: AgentSettingsPanelPro
       .catch(() => setLegacyV1(false));
   }, []);
 
+  useEffect(() => {
+    // The panel only needs the flag once; App owns the permanent
+    // onModelsDevReady listener for later hot swaps.
+    let alive = true;
+    modelsDevReady().then(() => {
+      if (alive) forceModelsDevReady();
+    });
+    return () => {
+      alive = false;
+    };
+  }, []);
+
   const update = (patch: Partial<AgentSettings>) => {
     onChange(setAgentSettings(rawOther, patch, { legacyV1 }));
   };
@@ -92,6 +119,42 @@ export function AgentSettingsPanel({ rawOther, onChange }: AgentSettingsPanelPro
 
   const thinkingEnabled = settings.thinking?.enabled ?? true;
 
+  // Which tiers the current default model accepts: `[models."<alias>"]
+  // support_efforts` first, then the models.dev `reasoning` flag.
+  const defaultEntry = defaultModel ? models?.[defaultModel] : undefined;
+  const effortSupport = resolveThinkingEffortSupport({
+    supportEfforts: readSupportEfforts(defaultEntry),
+    reasoning: defaultEntry ? getModelRef(defaultEntry.model)?.reasoning : undefined,
+  });
+
+  const effortNote = (availability: EffortAvailability): string | undefined => {
+    const note = availability.note;
+    if (!note) return undefined;
+    if (note.kind === "thinkingUnsupported") return t("thinkingEffortUnsupported");
+    if (note.kind === "tierUnsupported") {
+      return t("thinkingEffortTierUnsupported", { levels: note.supported.join(", ") });
+    }
+    return t("thinkingEffortTierUndeclared");
+  };
+
+  // Upstream takes a free-form string: keep a hand-written tier visible and
+  // selectable instead of dropping it from the control.
+  const effort = settings.thinking?.effort ?? "medium";
+  const effortOptions: {
+    key: string;
+    label: string;
+    disabled?: boolean;
+    title?: string;
+  }[] = THINKING_EFFORTS.map((tier) => ({
+    key: tier,
+    label: t(THINKING_LABELS[tier]),
+    disabled: !effortSupport.levels[tier].enabled,
+    title: effortNote(effortSupport.levels[tier]),
+  }));
+  if (!THINKING_EFFORTS.includes(effort as (typeof THINKING_EFFORTS)[number])) {
+    effortOptions.push({ key: effort, label: effort });
+  }
+
   return (
     <div className="mt-6 space-y-4">
       <h3 className="text-content-muted text-sm font-medium">{t("agentSettings")}</h3>
@@ -105,17 +168,20 @@ export function AgentSettingsPanel({ rawOther, onChange }: AgentSettingsPanelPro
         <div className="flex items-center gap-3 flex-wrap">
           <span className="text-sm text-content-muted">{t("thinkingLevel")}</span>
           <Segmented
-            options={THINKING_LEVELS.map((lvl) => ({
-              key: lvl,
-              label: t(THINKING_LABELS[lvl]),
-            }))}
-            value={settings.thinking?.effort ?? "medium"}
-            onChange={(effort) =>
-              updateThinking({ effort: effort as NonNullable<AgentSettings["thinking"]>["effort"] })
-            }
-            disabled={!thinkingEnabled}
+            options={effortOptions}
+            value={effort}
+            onChange={(next) => updateThinking({ effort: next })}
+            disabled={!thinkingEnabled || !effortSupport.thinkingSupported}
           />
         </div>
+        {!effortSupport.thinkingSupported && (
+          <p className="text-xs text-amber-500 dark:text-amber-400">
+            {t("thinkingEffortUnsupported")}
+          </p>
+        )}
+        <p className="text-xs text-content-muted">
+          {t("thinkingEffortDependsOnDefaultModel")}
+        </p>
         <Checkbox
           label={t("thinkingKeep")}
           checked={settings.thinking?.keep === "all"}
@@ -170,19 +236,37 @@ export function AgentSettingsPanel({ rawOther, onChange }: AgentSettingsPanelPro
       </Card>
 
       <Card title={t("watchSettings")}>
-        {/* Top-level `[watch] enabled` (kimi-code 2.0.1+; off by default since
-            2.0.2). KIMI_CODE_WATCH outranks this config at runtime — the env
-            var is probed by get_experimental_env_status as a non-flag entry
-            but deliberately locks nothing here, the toggle just documents the
-            config value that applies when the env var is unset. */}
+        {/* Top-level `[watch] enabled` (kimi-code 2.0.1+; 2.0.2 flipped the
+            default off, #4015 flipped it back on). KIMI_CODE_WATCH outranks
+            this config at runtime — the env var is probed by
+            get_experimental_env_status as a non-flag entry but deliberately
+            locks nothing here, the toggle just documents the config value
+            that applies when the env var is unset. */}
         <Checkbox
           label={t("watchEnabled")}
-          checked={settings.watch?.enabled ?? false}
+          checked={settings.watch?.enabled ?? true}
           onChange={(checked) => updateWatch({ enabled: checked })}
         />
         <p className="text-xs text-content-muted">{t("watchEnabledDesc")}</p>
       </Card>
 
+      <Card title={t("behaviorSettings")}>
+        {/* Top-level booleans, both default true upstream: false is written
+            explicitly, true removes the key (see setAgentSettings). */}
+        <Checkbox
+          label={t("autoSessionTitle")}
+          checked={settings.auto_session_title ?? true}
+          onChange={(checked) => update({ auto_session_title: checked })}
+        />
+        <p className="text-xs text-content-muted">{t("autoSessionTitleDesc")}</p>
+        <Checkbox
+          label={t("repeatBreaker")}
+          checked={settings.repeat_breaker ?? true}
+          onChange={(checked) => update({ repeat_breaker: checked })}
+        />
+        <p className="text-xs text-content-muted">{t("repeatBreakerDesc")}</p>
+      </Card>
+
       <Card title={t("permissionRules")}>
         {/* kimi-code `[permission] dangerous_command_guard`. The env
             var KIMI_CODE_DANGEROUS_COMMAND_GUARD (literal "true"/"false")

src/components/ProviderEdit.tsx

@@ -1,11 +1,17 @@
-import { useEffect, useId, useState } from "react";
+import { useEffect, useId, useMemo, useReducer, useRef, useState } from "react";
 import { createPortal } from "react-dom";
 import { invoke } from "@tauri-apps/api/core";
 import { useTranslation } from "../i18n";
 import type { TranslationKey } from "../i18n/zh";
 import { findPresetForProvider } from "../config/providerPresets";
 import { getDefaultMaxContextSize } from "../lib/model-defaults";
-import { capabilitiesFromRef, getModelRef } from "../lib/models-dev";
+import {
+  parseMaxOutputInput,
+  readMaxOutputSize,
+  setMaxOutputSize,
+  withMaxOutputSize,
+} from "../lib/model-max-output";
+import { capabilitiesFromRef, getModelRef, modelsDevReady } from "../lib/models-dev";
 import { getIconMetadata } from "../icons/extracted/metadata";
 import { AgentSettingsPanel } from "./AgentSettingsPanel";
 import { KimiOAuthDialog } from "./KimiOAuthDialog";
@@ -33,6 +39,9 @@ const CAPABILITY_LABELS: Record<
   tool_use: "capToolUse",
 };
 
+/** Fallback max-output cap (128K) for models models.dev doesn't know. */
+const FALLBACK_MAX_OUTPUT = 131_072;
+
 const PROVIDER_TYPES: ProviderType[] = [
   "openai",
   "openai_responses",
@@ -142,6 +151,13 @@ export function ProviderEdit({
   const apiKeyEnvSupported = agent === "kimi_code";
   const apiKeyEnvMode = apiKeyEnvSupported && provider.api_key_env != null;
 
+  // Alias → entry lookup for panels that resolve a model by alias
+  // (AgentSettingsPanel reads the default model's support_efforts).
+  const modelsByAlias = useMemo(
+    () => Object.fromEntries(models.map((m) => [m.alias, m])),
+    [models]
+  );
+
   useEffect(() => {
     const def = defaultBaseUrl(agent, provider.provider_type);
     if (!def) return;
@@ -462,6 +478,8 @@ export function ProviderEdit({
               <AgentSettingsPanel
                 rawOther={rawOther}
                 onChange={onRawOtherChange}
+                models={modelsByAlias}
+                defaultModel={defaultModel}
               />
             )}
           </>
@@ -558,6 +576,38 @@ function ModelMapping({
   const [discoverError, setDiscoverError] = useState<string | null>(null);
   const [fetchThinking, setFetchThinking] = useState(true);
 
+  // The models.dev index loads in the background; re-render once it lands so
+  // the max-output "参考" hints appear even when this panel mounted first.
+  const [, forceModelsDevReady] = useReducer((x: number) => x + 1, 0);
+  // One-shot backfill: models with no max_output_size yet get the models.dev
+  // `output` cap, or the 128K fallback when nothing matches. onModelChange
+  // only mutates the in-memory config — the user still reviews and presses
+  // 保存配置.
+  const backfillDone = useRef(false);
+  const modelsRef = useRef(models);
+  modelsRef.current = models;
+  useEffect(() => {
+    let alive = true;
+    modelsDevReady().then(() => {
+      if (!alive) return;
+      forceModelsDevReady();
+      if (backfillDone.current || agent !== "kimi_code") return;
+      backfillDone.current = true;
+      for (const m of modelsRef.current) {
+        if (readMaxOutputSize(m.raw_other) !== undefined) continue;
+        onModelChange(
+          withMaxOutputSize(
+            m,
+            getModelRef(m.model)?.output ?? FALLBACK_MAX_OUTPUT
+          )
+        );
+      }
+    });
+    return () => {
+      alive = false;
+    };
+  }, []);
+
   const handleDiscover = async () => {
     setDiscovering(true);
     setDiscoverError(null);
@@ -627,6 +677,17 @@ function ModelMapping({
           : fetchThinking
             ? ["thinking"]
             : [],
+        // Seed the max_output_size override with the models.dev cap, falling
+        // back to 128K when nothing matches. kimi_code only: Pi's equivalent
+        // is `maxTokens` (not written by this UI).
+        ...(agent === "kimi_code"
+          ? {
+              raw_other: setMaxOutputSize(
+                undefined,
+                ref?.output ?? FALLBACK_MAX_OUTPUT
+              ),
+            }
+          : {}),
       });
     }
     onBulkAdd(toAdd);
@@ -639,7 +700,10 @@ function ModelMapping({
       <div className="flex items-center justify-between">
         <div>
           <h3 className="font-medium text-content-primary">{t("modelMapping")}</h3>
-          <p className="text-xs text-content-muted mt-1">{t("modelMappingDesc")}</p>
+          <p className="text-xs text-content-muted mt-1">
+            {t("modelMappingDesc")}
+            {agent === "kimi_code" ? ` ${t("maxOutputSizeDesc")}` : ""}
+          </p>
         </div>
         <div className="flex items-center gap-2">
           <button
@@ -747,7 +811,9 @@ function ModelMapping({
               <th className="text-left px-4 py-3 font-medium">{t("displayName")}</th>
               <th className="text-left px-4 py-3 font-medium">{t("actualModel")}</th>
               <th className="text-left px-4 py-3 font-medium w-28">{t("contextSize")}</th>
-              <th className="text-center px-4 py-3 font-medium w-28">{t("supports1M")}</th>
+              {agent === "kimi_code" && (
+                <th className="text-left px-4 py-3 font-medium w-32">{t("maxOutputSize")}</th>
+              )}
               {agent === "kimi_code" && (
                 <th className="text-left px-4 py-3 font-medium">{t("capabilities")}</th>
               )}
@@ -788,19 +854,14 @@ function ModelMapping({
                     }}
                   />
                 </td>
-                <td className="px-4 py-2 text-center">
-                  <input
-                    type="checkbox"
-                    checked={m.supports_1m || false}
-                    onChange={(e) =>
-                      onModelChange({
-                        ...m,
-                        supports_1m: e.target.checked,
-                      })
-                    }
-                    className="w-4 h-4 rounded border-border bg-input text-blue-600 focus:ring-blue-500"
+                {agent === "kimi_code" && (
+                  <td className="px-4 py-2">
+                    <MaxOutputCell
+                      model={m}
+                      onChange={(next) => onModelChange(next)}
                     />
                   </td>
+                )}
                 {agent === "kimi_code" && (
                   <td className="px-4 py-2">
                     <CapabilitiesCell
@@ -976,3 +1037,47 @@ function CapabilitiesCell({
     </div>
   );
 }
+
+/**
+ * Max output tokens override (`[models."<alias>"] max_output_size`).
+ *
+ * Not a first-class Model field, so the value lives in `raw_other` (see
+ * lib/model-max-output.ts). Blank input removes the key — "unset" means the
+ * upstream default applies. When models.dev knows the model's cap it is shown
+ * as a click-to-fill reference underneath.
+ */
+function MaxOutputCell({
+  model,
+  onChange,
+}: {
+  model: Model;
+  onChange: (next: Model) => void;
+}) {
+  const { t } = useTranslation();
+  const current = readMaxOutputSize(model.raw_other);
+  const refOutput = getModelRef(model.model)?.output;
+
+  return (
+    <div>
+      <input
+        type="number"
+        min={0}
+        step={1024}
+        className="w-full bg-transparent border border-border rounded px-2 py-1.5 text-sm focus:ring-2 focus:ring-blue-500 focus:outline-none"
+        value={current ?? ""}
+        onChange={(e) =>
+          onChange(withMaxOutputSize(model, parseMaxOutputInput(e.target.value)))
+        }
+      />
+      {refOutput !== undefined && refOutput !== current && (
+        <button
+          type="button"
+          onClick={() => onChange(withMaxOutputSize(model, refOutput))}
+          className="mt-1 text-[11px] text-content-muted hover:text-blue-500"
+        >
+          {t("maxOutputRef", { value: refOutput })}
+        </button>
+      )}
+    </div>
+  );
+}

src/components/ProviderList.tsx

@@ -96,7 +96,9 @@ function ProviderCard({
     provider.usageKinds,
     provider.usageConfig?.autoQueryIntervalMinutes,
     provider.usageConfig?.enabled === false,
-    provider.usageConfig?.threshold
+    provider.usageConfig?.threshold,
+    provider.usageConfig?.templateType === "newapi" ||
+      provider.usageConfig?.templateType === "sub2api"
   );
   const providerModels = Object.values(models).filter(
     (m) => m.provider === provider.name
@@ -221,7 +223,7 @@ function ProviderCard({
         )}
         {/* compact usage summary — sits left of the switch button,
             mirroring cc-switch's card layout (usage → action buttons) */}
-        {(provider.usageKinds?.length ?? 0) > 0 && (
+        {usage.supported && (
           <UsageFooter usage={usage} variant="compact" />
         )}
         <button

src/components/SubagentSettingsPage.tsx

@@ -43,12 +43,15 @@ interface SubagentSettingsPageProps {
   onBack: () => void;
 }
 
-// Upstream removed the "max" effort tier (auto-migrates to "high").
-const EFFORTS = ["low", "medium", "high"] as const;
+// Effort tiers the CLI accepts; the value is forwarded verbatim upstream, and
+// per-model support comes from `[models."<alias>"] support_efforts`.
+const EFFORTS = ["low", "medium", "high", "xhigh", "max"] as const;
 const EFFORT_LABELS: Record<(typeof EFFORTS)[number], TranslationKey> = {
   low: "thinkingLow",
   medium: "thinkingMedium",
   high: "thinkingHigh",
+  max: "thinkingMax",
+  xhigh: "thinkingXHigh",
 };
 
 const FLAG_LABELS: Record<string, { name: TranslationKey; desc: TranslationKey }> = {
@@ -281,9 +284,8 @@ export function SubagentSettingsPage({
         ? `${Math.round(n / 1000)}K`
         : String(n);
   const effortLabel = (e: string): string => {
-    // Stored "max" tiers from old configs are shown as "high" (upstream
-    // removed the tier; it auto-migrates to "high").
-    const key = EFFORT_LABELS[e === "max" ? "high" : (e as (typeof EFFORTS)[number])];
+    // Unknown values (hand-written config) are shown verbatim.
+    const key = EFFORT_LABELS[e as (typeof EFFORTS)[number]];
     return key ? t(key) : e;
   };
 

src/components/dashboard/DashboardPage.tsx

@@ -179,7 +179,7 @@ export function DashboardPage() {
           </div>
           {loadStats && !loading && (
             <span className="hidden text-[10px] text-content-muted sm:inline">
-              {t("loadStats", { ms: loadStats.ms, kb: loadStats.kb })}
+              {t("loadStats", { ms: loadStats.ms })}
             </span>
           )}
         </div>

src/components/ui/controls.tsx

@@ -106,7 +106,7 @@ export function Segmented({
   onChange,
   disabled,
 }: {
-  options: { key: string; label: string }[];
+  options: { key: string; label: string; disabled?: boolean; title?: string }[];
   value: string;
   onChange: (value: string) => void;
   disabled?: boolean;
@@ -117,21 +117,27 @@ export function Segmented({
         disabled ? "opacity-50" : ""
       }`}
     >
-      {options.map((opt) => (
+      {options.map((opt) => {
+        const optDisabled = disabled || opt.disabled === true;
+        return (
+          // The title lives on the wrapper: a disabled button swallows pointer
+          // events in Chromium, so its own tooltip would never show.
+          <span key={opt.key} title={opt.title} className="inline-flex">
             <button
-          key={opt.key}
               type="button"
-          disabled={disabled}
+              disabled={optDisabled}
               onClick={() => onChange(opt.key)}
               className={`px-3 py-1 text-sm ${
                 value === opt.key
                   ? "bg-blue-600 text-white"
                   : "bg-input text-content-muted hover:bg-hover-2"
-          }`}
+              } ${optDisabled ? "cursor-not-allowed opacity-40 hover:bg-input" : ""}`}
             >
               {opt.label}
             </button>
-      ))}
+          </span>
+        );
+      })}
     </div>
   );
 }

src/hooks/useDashboard.ts

@@ -1,32 +1,68 @@
-import { useCallback, useEffect, useState } from "react";
+import { useCallback, useEffect, useRef, useState } from "react";
 import { invoke } from "@tauri-apps/api/core";
+import { listen } from "@tauri-apps/api/event";
 import type { SummaryResult } from "../types/dashboard";
 
 export type DashboardRange = "today" | "yesterday" | "7d" | "30d" | "all";
 
 const RANGE_STORAGE_KEY = "kimi-switch-dashboard-range";
 
-export function useDashboard() {
-  const [range, setRange] = useState<DashboardRange>(() => {
+/**
+ * Emitted by the backend when the background verification walk found records the
+ * fast (disk-cache) path did not have. The payload is empty — the hook only
+ * needs to know that a re-fetch is worth making.
+ */
+const RECORDS_UPDATED_EVENT = "dashboard://records-updated";
+
+const RANGES: readonly DashboardRange[] = ["today", "yesterday", "7d", "30d", "all"];
+
+function readStoredRange(): DashboardRange {
   try {
     const stored = localStorage.getItem(RANGE_STORAGE_KEY) as DashboardRange | null;
-      if (stored === "today" || stored === "yesterday" || stored === "7d" || stored === "30d" || stored === "all") {
+    if (stored && (RANGES as readonly string[]).includes(stored)) {
       return stored;
     }
   } catch {
     // ignore
   }
   return "30d";
-  });
+}
 
-  const [data, setData] = useState<SummaryResult | null>(null);
+/**
+ * Last successful payload per range, kept across mounts so a range switch (and
+ * reopening the page) paints instantly from memory and revalidates in the
+ * background instead of showing a spinner over data that is already known.
+ * Bounded by the five ranges the dashboard offers.
+ */
+const summaryCache = new Map<DashboardRange, SummaryResult>();
+
+export function useDashboard() {
+  const [range, setRange] = useState<DashboardRange>(readStoredRange);
+
+  const [data, setData] = useState<SummaryResult | null>(
+    () => summaryCache.get(readStoredRange()) ?? null,
+  );
   const [loading, setLoading] = useState(false);
   const [error, setError] = useState<string | null>(null);
   /** Round-trip timing for the last get_summary call (user-perceived lag). */
-  const [loadStats, setLoadStats] = useState<{ ms: number; kb: number } | null>(null);
+  const [loadStats, setLoadStats] = useState<{ ms: number } | null>(null);
+  // Generation counter: stale responses (superseded range / unmounted) are dropped.
+  const genRef = useRef(0);
+  // The event handler must call the *current* refresh, not the one captured when
+  // the listener was installed, so it reads the callback through a ref.
+  const refreshRef = useRef<(force?: boolean) => Promise<void>>(async () => {});
 
   const refresh = useCallback(async (force = false) => {
-    setLoading(true);
+    const gen = ++genRef.current;
+    // Stale-while-revalidate: hand the cached payload for this range to the
+    // view before the request goes out, so a range switch is instant and the
+    // response only refreshes numbers that are already on screen. `loading` is
+    // raised only when there is nothing to show (first load of a range) or when
+    // the user asked for a refresh — the button keeps its busy state, the silent
+    // revalidation behind a cached range does not.
+    const cached = summaryCache.get(range);
+    if (cached) setData(cached);
+    setLoading(force || !cached);
     setError(null);
     const start = performance.now();
     try {
@@ -34,23 +70,59 @@ export function useDashboard() {
         range,
         refresh: force,
       });
+      if (genRef.current !== gen) return;
+      summaryCache.set(range, result);
       setData(result);
-      setLoadStats({
-        ms: Math.round(performance.now() - start),
-        kb: Math.round(JSON.stringify(result).length / 1024),
-      });
+      setLoadStats({ ms: Math.round(performance.now() - start) });
     } catch (err) {
+      if (genRef.current !== gen) return;
       const msg = err instanceof Error ? err.message : String(err);
       setError(msg);
     } finally {
-      setLoading(false);
+      if (genRef.current === gen) setLoading(false);
     }
   }, [range]);
 
+  refreshRef.current = refresh;
+
   useEffect(() => {
     refresh();
   }, [refresh]);
 
+  // The first load of a session answers from the on-disk snapshot, which can be
+  // behind a session written since the app last ran. The backend verifies in the
+  // background and signals here when the numbers moved; re-fetching silently
+  // (never raising `loading`) keeps the already-drawn dashboard on screen and
+  // just corrects it. A silent call still takes a fresh generation, so it cannot
+  // be interleaved with a range switch in flight.
+  useEffect(() => {
+    let unlisten: (() => void) | undefined;
+    let cancelled = false;
+    listen(RECORDS_UPDATED_EVENT, () => {
+      refreshRef.current();
+    })
+      .then((fn) => {
+        // The effect may have been torn down before `listen` resolved.
+        if (cancelled) fn();
+        else unlisten = fn;
+      })
+      .catch(() => {
+        // No event bridge (e.g. a browser build): the dashboard still works,
+        // it just will not be corrected mid-session.
+      });
+    return () => {
+      cancelled = true;
+      unlisten?.();
+    };
+  }, []);
+
+  // Ignore late responses after unmount.
+  useEffect(() => {
+    return () => {
+      genRef.current += 1;
+    };
+  }, []);
+
   const changeRange = useCallback((next: DashboardRange) => {
     setRange(next);
     try {

src/hooks/useUsageQuery.test.ts

@@ -1,5 +1,5 @@
 import { describe, expect, it } from "vitest";
-import { isAlertTier, type UsageData } from "./useUsageQuery";
+import { isAlertTier, usageQuerySupported, type UsageData } from "./useUsageQuery";
 
 /** 百分比窗口行（five_hour 等）；套餐类 tier 的 unit 恒为 "%"。 */
 function percentTier(used: number): UsageData {
@@ -42,3 +42,22 @@ describe("isAlertTier", () => {
     expect(isAlertTier({ planName: "five_hour", unit: "%", remaining: 20 }, 50)).toBe(false);
   });
 });
+
+describe("usageQuerySupported", () => {
+  it("有识别类型即可查询", () => {
+    expect(usageQuerySupported(["balance:deepseek"])).toBe(true);
+    expect(usageQuerySupported([], true)).toBe(true);
+  });
+
+  it("显式模板（newapi/sub2api）绕过 usage_kinds——自定义供应商也该查", () => {
+    expect(usageQuerySupported(undefined, true)).toBe(true);
+  });
+
+  it("无识别类型且非显式模板时不查询", () => {
+    expect(usageQuerySupported(undefined)).toBe(false);
+    expect(usageQuerySupported([])).toBe(false);
+    // auto 模板没有识别到类型 = 不支持（与弹窗短路口径一致）
+    expect(usageQuerySupported([], false)).toBe(false);
+    expect(usageQuerySupported(undefined, undefined)).toBe(false);
+  });
+});

src/hooks/useUsageQuery.ts

@@ -75,15 +75,30 @@ export function isAlertTier(
   return d.unit === "%" && d.used != null && d.used >= threshold;
 }
 
+/**
+ * 是否值得为该供应商发起用量查询：有预设识别类型，或后端按显式模板
+ * （newapi / sub2api，二者绕过 usage_kinds）查询。
+ */
+export function usageQuerySupported(
+  usageKinds?: string[],
+  explicitTemplate?: boolean
+): boolean {
+  return (usageKinds?.length ?? 0) > 0 || explicitTemplate === true;
+}
+
 export function useUsageQuery(
   agent: Agent,
   providerName: string,
   usageKinds?: string[],
   autoIntervalMinutes?: number,
   disabled?: boolean,
-  threshold?: number
+  threshold?: number,
+  /** Backend resolves the query from an explicit template (newapi / sub2api)
+   * regardless of usageKinds — those bypass kinds entirely, so the row is
+   * supported even when the provider has no preset-detected kinds. */
+  explicitTemplate?: boolean
 ): UsageQueryState {
-  const supported = (usageKinds?.length ?? 0) > 0;
+  const supported = usageQuerySupported(usageKinds, explicitTemplate);
   // Cache key includes the agent so a Kimi Code provider and a Pi provider
   // with the same name do not clobber each other's cached result.
   const cacheKey = `${agent}:${providerName}`;

src/i18n/en.ts

@@ -111,7 +111,8 @@ export const enTranslations: Record<TranslationKey, string> = {
   baseUrlHint: "Enter a {type} API-compatible endpoint without a trailing slash.",
   envPairs: "Env pairs",
   addEnv: "+ Add",
-  modelMappingDesc: "Display name only affects the /model menu; 1M declares context capability for Kimi Code.",
+  modelMappingDesc: "Display name only affects the /model menu.",
+  maxOutputSizeDesc: "Max output writes max_output_size — leave blank to send no output cap upstream.",
   oneClickSetup: "One-click setup",
   fetchModels: "Fetch models",
   fetchingModels: "Fetching...",
@@ -121,7 +122,8 @@ export const enTranslations: Record<TranslationKey, string> = {
   displayName: "Display name",
   actualModel: "Actual model",
   contextSize: "Context size",
-  supports1M: "Supports 1M",
+  maxOutputSize: "Max output",
+  maxOutputRef: "Ref {value}",
   default: "Default",
   operation: "Operation",
   isDefault: "Default",
@@ -164,6 +166,15 @@ export const enTranslations: Record<TranslationKey, string> = {
   thinkingLow: "Low",
   thinkingMedium: "Medium",
   thinkingHigh: "High",
+  thinkingMax: "Max",
+  thinkingXHigh: "XHigh",
+  thinkingEffortUnsupported: "The current default model does not support thinking.",
+  thinkingEffortTierUnsupported:
+    "The current default model does not support this tier (supported: {levels}).",
+  thinkingEffortTierUndeclared:
+    "This model declares no thinking effort tiers; the upstream may reject this one.",
+  thinkingEffortDependsOnDefaultModel:
+    "The tier that actually applies depends on the current default model.",
   thinkingContextHint: "Thinking uses more context. Ensure the model context length and reserved size are sufficient.",
   loopControlSettings: "Loop Control",
   maxAttemptsPerStep: "Max attempts per step",
@@ -177,7 +188,14 @@ export const enTranslations: Record<TranslationKey, string> = {
   watchSettings: "File Watching",
   watchEnabled: "Enable file watching",
   watchEnabledDesc:
-    "Watch config files and the workspace for changes (configurable since kimi-code 2.0.1, off by default since 2.0.2). The KIMI_CODE_WATCH environment variable overrides this setting.",
+    "Watch config files and the workspace for changes (configurable since kimi-code 2.0.1, on by default). The KIMI_CODE_WATCH environment variable overrides this setting.",
+  behaviorSettings: "Behavior",
+  autoSessionTitle: "Auto session title",
+  autoSessionTitleDesc:
+    "Let clients auto-generate session titles (on by default; turning it off writes false).",
+  repeatBreaker: "Repeat breaker",
+  repeatBreakerDesc:
+    "Remind and force-stop on consecutively repeated identical tool calls (on by default). The KIMI_CODE_REPEAT_BREAKER environment variable overrides this setting.",
   permissionRules: "Permission Rules",
   permissionDecision: "Decision",
   permissionPattern: "Pattern",
@@ -239,7 +257,7 @@ export const enTranslations: Record<TranslationKey, string> = {
   scanning: "Scanning…",
   refresh: "Refresh",
   refreshing: "Refreshing…",
-  loadStats: "Loaded in {ms} ms · {kb} KB payload",
+  loadStats: "Loaded in {ms} ms",
   refreshed: "Refreshed",
   cancel: "Cancel",
   confirmDeleteTitle: "Delete session forever?",

src/i18n/zh.ts

@@ -108,7 +108,8 @@ export const zhTranslations = {
   baseUrlHint: "填写兼容 {type} API 的服务端点地址，不要以斜杠结尾",
   envPairs: "Env 键值对",
   addEnv: "+ 添加",
-  modelMappingDesc: "显示名称只影响 /model 菜单；1M 只是给 Kimi Code 的上下文能力声明。",
+  modelMappingDesc: "显示名称只影响 /model 菜单。",
+  maxOutputSizeDesc: "「最大输出」写入 max_output_size，留空则不发送输出上限。",
   oneClickSetup: "一键设置",
   fetchModels: "获取模型列表",
   fetchingModels: "获取中...",
@@ -118,7 +119,8 @@ export const zhTranslations = {
   displayName: "显示名称",
   actualModel: "实际请求模型",
   contextSize: "上下文长度",
-  supports1M: "声明支持 1M",
+  maxOutputSize: "最大输出",
+  maxOutputRef: "参考 {value}",
   default: "默认",
   operation: "操作",
   isDefault: "已默认",
@@ -158,9 +160,15 @@ export const zhTranslations = {
   enableThinking: "启用思考",
   thinkingLevel: "思考等级",
   thinkingKeep: "保留思考内容",
-  thinkingLow: "低",
-  thinkingMedium: "中",
-  thinkingHigh: "高",
+  thinkingLow: "Low",
+  thinkingMedium: "Medium",
+  thinkingHigh: "High",
+  thinkingMax: "Max",
+  thinkingXHigh: "XHigh",
+  thinkingEffortUnsupported: "当前默认模型不支持思考。",
+  thinkingEffortTierUnsupported: "当前默认模型不支持该档位（支持：{levels}）。",
+  thinkingEffortTierUndeclared: "该模型未声明思考等级，上游可能拒绝此档位。",
+  thinkingEffortDependsOnDefaultModel: "实际生效的档位取决于当前默认模型。",
   thinkingContextHint: "启用思考会占用更多上下文，请确保模型上下文长度和预留空间足够。",
   loopControlSettings: "循环控制",
   maxAttemptsPerStep: "单步最大尝试次数",
@@ -173,7 +181,14 @@ export const zhTranslations = {
   watchSettings: "文件监听",
   watchEnabled: "启用文件监听",
   watchEnabledDesc:
-    "监听配置文件与工作目录变化（kimi-code 2.0.1 起可配置，2.0.2 起默认关闭）。环境变量 KIMI_CODE_WATCH 会覆盖本配置。",
+    "监听配置文件与工作目录变化（kimi-code 2.0.1 起可配置，默认开启）。环境变量 KIMI_CODE_WATCH 会覆盖本配置。",
+  behaviorSettings: "行为开关",
+  autoSessionTitle: "自动生成会话标题",
+  autoSessionTitleDesc:
+    "允许客户端自动生成会话标题（默认开启；关闭后写入 false）。",
+  repeatBreaker: "重复调用拦截",
+  repeatBreakerDesc:
+    "连续重复相同工具调用时提醒并强制停止（默认开启）。环境变量 KIMI_CODE_REPEAT_BREAKER 会覆盖本配置。",
   permissionRules: "权限规则",
   permissionDecision: "处置",
   permissionPattern: "模式",
@@ -234,7 +249,7 @@ export const zhTranslations = {
   scanning: "扫描中…",
   refresh: "刷新",
   refreshing: "刷新中…",
-  loadStats: "加载 {ms} ms · 数据 {kb} KB",
+  loadStats: "加载 {ms} ms",
   refreshed: "已刷新",
   cancel: "取消",
   confirmDeleteTitle: "永久删除会话？",

src/lib/agent-settings.test.ts

@@ -20,7 +20,7 @@ function loopOf(raw: unknown): Record<string, unknown> {
 }
 
 // ---------------------------------------------------------------------------
-// Reading — legacy off values normalized to "off", "max" → "high"
+// Reading — legacy off values normalized to "off"; effort tiers pass through
 // ---------------------------------------------------------------------------
 
 describe("getAgentSettings — thinking.keep normalization", () => {
@@ -52,12 +52,29 @@ describe("getAgentSettings — thinking.keep normalization", () => {
   });
 });
 
-describe("getAgentSettings — effort \"max\" read mapping", () => {
-  it("normalizes a stored \"max\" to \"high\" on read", () => {
+describe("getAgentSettings — thinking.effort read mapping", () => {
+  it("keeps a stored \"max\" as-is (the tier is valid upstream)", () => {
     expect(getAgentSettings({ thinking: { effort: "max" } }).thinking?.effort).toBe(
-      "high"
+      "max"
     );
   });
+
+  it("passes an \"xhigh\" tier through untouched", () => {
+    expect(getAgentSettings({ thinking: { effort: "xhigh" } }).thinking?.effort).toBe(
+      "xhigh"
+    );
+  });
+
+  it("keeps the default \"medium\" when the key is absent", () => {
+    expect(getAgentSettings({}).thinking?.effort).toBe("medium");
+  });
+
+  it("round-trips an arbitrary hand-written tier on save", () => {
+    const next = setAgentSettings({ thinking: { effort: "ultra" } }, {
+      thinking: { effort: "ultra" },
+    });
+    expect(thinkingOf(next).effort).toBe("ultra");
+  });
 });
 
 // ---------------------------------------------------------------------------
@@ -184,8 +201,8 @@ describe("loop_control.compaction_max_attempts", () => {
 });
 
 // ---------------------------------------------------------------------------
-// [watch] enabled (kimi-code 2.0.1+) — top-level section, off by default since
-// 2.0.2; mirrors the [background] handling
+// [watch] enabled (kimi-code 2.0.1+) — top-level section; 2.0.2 flipped the
+// default off, #4015 flipped it back on. Mirrors the [background] handling.
 // ---------------------------------------------------------------------------
 
 function watchOf(raw: unknown): Record<string, unknown> {
@@ -196,9 +213,9 @@ function watchOf(raw: unknown): Record<string, unknown> {
 }
 
 describe("[watch] enabled", () => {
-  it("reads as false when the section or key is absent (2.0.2 default)", () => {
-    expect(getAgentSettings({}).watch?.enabled).toBe(false);
-    expect(getAgentSettings({ watch: {} }).watch?.enabled).toBe(false);
+  it("reads as true when the section or key is absent (#4015 default)", () => {
+    expect(getAgentSettings({}).watch?.enabled).toBe(true);
+    expect(getAgentSettings({ watch: {} }).watch?.enabled).toBe(true);
   });
 
   it("reads an explicit value", () => {
@@ -231,6 +248,49 @@ describe("[watch] enabled", () => {
   });
 });
 
+// ---------------------------------------------------------------------------
+// Top-level booleans (auto_session_title #3962 / repeat_breaker #3995):
+// upstream default true — false is written explicitly, true removes the key.
+// ---------------------------------------------------------------------------
+
+describe("top-level boolean switches", () => {
+  it("reads as undefined when the key is absent (default true)", () => {
+    const s = getAgentSettings({});
+    expect(s.auto_session_title).toBeUndefined();
+    expect(s.repeat_breaker).toBeUndefined();
+  });
+
+  it("reads explicit false values", () => {
+    const s = getAgentSettings({
+      auto_session_title: false,
+      repeat_breaker: false,
+    });
+    expect(s.auto_session_title).toBe(false);
+    expect(s.repeat_breaker).toBe(false);
+  });
+
+  it("writes false explicitly and removes the key on true", () => {
+    const off = setAgentSettings({}, { repeat_breaker: false }) as Record<
+      string,
+      unknown
+    >;
+    expect(off.repeat_breaker).toBe(false);
+    const on = setAgentSettings(
+      { repeat_breaker: false },
+      { repeat_breaker: true }
+    ) as Record<string, unknown>;
+    expect("repeat_breaker" in on).toBe(false);
+  });
+
+  it("keeps an explicit false on saves that do not touch it", () => {
+    const next = setAgentSettings(
+      { auto_session_title: false },
+      { thinking: { effort: "high" } }
+    ) as Record<string, unknown>;
+    expect(next.auto_session_title).toBe(false);
+  });
+});
+
 
 describe("permission.dangerous_command_guard", () => {
   it("reads as undefined when the key is absent (default on)", () => {

src/lib/agent-settings.ts

@@ -13,10 +13,11 @@ const DEFAULT_SETTINGS: AgentSettings = {
   background: {
     keep_alive_on_exit: false,
   },
-  // `[watch] enabled` — kimi-code 2.0.1 added the key, 2.0.2 flipped the
-  // default to off, so an absent key reads as "off".
+  // `[watch] enabled` — kimi-code 2.0.1 added the key; 2.0.2 flipped the
+  // default to off, then #4015 flipped it back on, so an absent key reads
+  // as "on".
   watch: {
-    enabled: false,
+    enabled: true,
   },
   permission: { rules: [] },
   hooks: [],
@@ -37,6 +38,7 @@ function getSection<T>(rawOther: unknown, key: string): T | undefined {
 }
 
 export function getAgentSettings(rawOther: unknown): AgentSettings {
+  const root = asRecord(rawOther);
   const sectionLoop = getSection<AgentSettings["loop_control"]>(
     rawOther,
     "loop_control"
@@ -65,11 +67,6 @@ export function getAgentSettings(rawOther: unknown): AgentSettings {
   if (thinking.keep !== undefined && thinking.keep !== "all") {
     thinking.keep = "off";
   }
-  // The "max" effort tier was removed upstream (auto-migrates to "high");
-  // old configs still carrying it are shown as "high" (not rewritten on read).
-  if (thinking.effort === "max") {
-    thinking.effort = "high";
-  }
   const sectionPermission = getSection<AgentSettings["permission"]>(
     rawOther,
     "permission"
@@ -97,6 +94,16 @@ export function getAgentSettings(rawOther: unknown): AgentSettings {
         : {}),
     },
     hooks: getSection<AgentSettings["hooks"]>(rawOther, "hooks") ?? [],
+    // Top-level booleans (upstream default true): keep only explicit values,
+    // an absent key means "on" — the UI shows `?? true`.
+    auto_session_title:
+      typeof root.auto_session_title === "boolean"
+        ? root.auto_session_title
+        : undefined,
+    repeat_breaker:
+      typeof root.repeat_breaker === "boolean"
+        ? root.repeat_breaker
+        : undefined,
   };
 }
 
@@ -171,5 +178,17 @@ export function setAgentSettings(
     delete root.hooks;
   }
 
+  // Top-level booleans with an upstream default of true: an explicit false
+  // is written; true/undefined removes the key so the config tracks the
+  // upstream default instead of pinning it.
+  for (const key of ["auto_session_title", "repeat_breaker"] as const) {
+    const value = patch[key] ?? current[key];
+    if (value === false) {
+      root[key] = false;
+    } else {
+      delete root[key];
+    }
+  }
+
   return root;
 }

src/lib/model-max-output.test.ts

@@ -0,0 +1,83 @@
+import { describe, expect, it } from "vitest";
+import {
+  parseMaxOutputInput,
+  readMaxOutputSize,
+  setMaxOutputSize,
+  withMaxOutputSize,
+} from "./model-max-output";
+
+const model = (raw_other: unknown) =>
+  ({
+    alias: "a",
+    provider: "p",
+    model: "m",
+    max_context_size: 0,
+    display_name: null,
+    raw_other,
+  }) as const;
+
+describe("readMaxOutputSize", () => {
+  it("reads the key from a model entry's preserved config fields", () => {
+    expect(readMaxOutputSize({ max_output_size: 32768 })).toBe(32768);
+  });
+
+  it("treats absent / malformed / non-positive values as unset", () => {
+    expect(readMaxOutputSize(undefined)).toBeUndefined();
+    expect(readMaxOutputSize(null)).toBeUndefined();
+    expect(readMaxOutputSize({})).toBeUndefined();
+    expect(readMaxOutputSize("32768")).toBeUndefined();
+    expect(readMaxOutputSize({ max_output_size: 0 })).toBeUndefined();
+    expect(readMaxOutputSize({ max_output_size: -1 })).toBeUndefined();
+    expect(readMaxOutputSize({ max_output_size: NaN })).toBeUndefined();
+    expect(readMaxOutputSize([1, 2])).toBeUndefined();
+  });
+});
+
+describe("parseMaxOutputInput", () => {
+  it("keeps positive integers", () => {
+    expect(parseMaxOutputInput("32768")).toBe(32768);
+  });
+
+  it("maps blank / zero / garbage to unset", () => {
+    expect(parseMaxOutputInput("")).toBeUndefined();
+    expect(parseMaxOutputInput("0")).toBeUndefined();
+    expect(parseMaxOutputInput("-5")).toBeUndefined();
+    expect(parseMaxOutputInput("abc")).toBeUndefined();
+  });
+});
+
+describe("setMaxOutputSize", () => {
+  it("sets the key while preserving other preserved fields", () => {
+    expect(setMaxOutputSize({ support_efforts: ["low"] }, 16384)).toEqual({
+      support_efforts: ["low"],
+      max_output_size: 16384,
+    });
+  });
+
+  it("drops the key when the value is unset", () => {
+    expect(setMaxOutputSize({ max_output_size: 16384, force: true }, undefined)).toEqual({
+      force: true,
+    });
+  });
+
+  it("starts from an empty object for null / non-object raw_other", () => {
+    expect(setMaxOutputSize(null, 1024)).toEqual({ max_output_size: 1024 });
+    expect(setMaxOutputSize("junk", 1024)).toEqual({ max_output_size: 1024 });
+    expect(setMaxOutputSize({ max_output_size: 1 }, undefined)).toEqual({});
+  });
+});
+
+describe("withMaxOutputSize", () => {
+  it("returns a new model with the override applied", () => {
+    const before = model({ max_output_size: 4096 });
+    const after = withMaxOutputSize(before, 8192);
+    expect(readMaxOutputSize(after.raw_other)).toBe(8192);
+    expect(readMaxOutputSize(before.raw_other)).toBe(4096);
+    expect(after).not.toBe(before);
+  });
+
+  it("clears the override without touching other fields", () => {
+    const after = withMaxOutputSize(model({ max_output_size: 4096 }), undefined);
+    expect(after.raw_other).toEqual({});
+  });
+});

src/lib/model-max-output.ts

@@ -0,0 +1,59 @@
+import type { Model } from "../types";
+
+/**
+ * Helpers for the `max_output_size` model field.
+ *
+ * kimi-code keeps it in `[models."<alias>"] max_output_size`; it is not a
+ * first-class field of Kimi Switch's Rust `Model` struct, so it round-trips
+ * through `raw_other` (the pass-through bucket for unknown keys). Absent key =
+ * unset = the upstream default applies, so clearing the field removes the key
+ * rather than writing 0.
+ */
+
+function asRecord(value: unknown): Record<string, unknown> {
+  if (value && typeof value === "object" && !Array.isArray(value)) {
+    return value as Record<string, unknown>;
+  }
+  return {};
+}
+
+/** Read the override; undefined when absent or malformed (treated as unset). */
+export function readMaxOutputSize(rawOther: unknown): number | undefined {
+  const value = asRecord(rawOther).max_output_size;
+  if (typeof value !== "number" || !Number.isFinite(value) || value <= 0) {
+    return undefined;
+  }
+  return value;
+}
+
+/** Parse the number input: blank / 0 / NaN mean "unset". */
+export function parseMaxOutputInput(text: string): number | undefined {
+  const value = parseInt(text, 10);
+  if (isNaN(value) || value <= 0) return undefined;
+  return value;
+}
+
+/**
+ * Set (positive value) or drop (undefined) `max_output_size` on a raw_other
+ * blob; every other preserved key is kept.
+ */
+export function setMaxOutputSize(
+  rawOther: unknown,
+  value: number | undefined,
+): Record<string, unknown> {
+  const next = { ...asRecord(rawOther) };
+  if (value === undefined) {
+    delete next.max_output_size;
+  } else {
+    next.max_output_size = Math.floor(value);
+  }
+  return next;
+}
+
+/** Immutable update of a model entry's max_output_size override. */
+export function withMaxOutputSize(
+  model: Model,
+  value: number | undefined,
+): Model {
+  return { ...model, raw_other: setMaxOutputSize(model.raw_other, value) };
+}

src/lib/models-dev.ts

@@ -25,6 +25,8 @@ export interface ModelCost {
 export interface ModelRef {
   name?: string;
   context?: number;
+  /** Upper bound on generated tokens (`limit.output` on models.dev). */
+  output?: number;
   reasoning?: boolean;
   tool_call?: boolean;
   structured_output?: boolean;

src/lib/providerRename.test.ts

@@ -0,0 +1,120 @@
+import { describe, expect, it } from "vitest";
+import {
+  copyModelsOnProviderDuplicate,
+  rekeyModelsOnProviderRename,
+} from "./providerRename";
+import type { Model } from "../types";
+
+function model(alias: string, provider: string): Model {
+  return {
+    alias,
+    provider,
+    model: alias.slice(provider.length + 1),
+    max_context_size: 512000,
+    display_name: null,
+  };
+}
+
+describe("rekeyModelsOnProviderRename", () => {
+  it("re-keys aliases and provider when a provider is renamed (opencode-go-copy scenario)", () => {
+    const models = {
+      "opencode-go-copy/deepseek-v4-flash": model("opencode-go-copy/deepseek-v4-flash", "opencode-go-copy"),
+      "opencode-go-copy/deepseek-v4.1-flash": model("opencode-go-copy/deepseek-v4.1-flash", "opencode-go-copy"),
+      "opencode-go/longcat-2.5-preview-free": model("opencode-go/longcat-2.5-preview-free", "opencode-go"),
+      "other/m1": model("other/m1", "other"),
+    };
+
+    const { models: next, aliasMap } = rekeyModelsOnProviderRename(models, "opencode-go-copy", "opencode-go");
+
+    expect(Object.keys(next).sort()).toEqual([
+      "opencode-go/deepseek-v4-flash",
+      "opencode-go/deepseek-v4.1-flash",
+      "opencode-go/longcat-2.5-preview-free",
+      "other/m1",
+    ]);
+    expect(next["opencode-go/deepseek-v4-flash"].provider).toBe("opencode-go");
+    expect(next["opencode-go/deepseek-v4-flash"].model).toBe("deepseek-v4-flash");
+    expect(aliasMap).toEqual({
+      "opencode-go-copy/deepseek-v4-flash": "opencode-go/deepseek-v4-flash",
+      "opencode-go-copy/deepseek-v4.1-flash": "opencode-go/deepseek-v4.1-flash",
+    });
+    // Untouched providers stay identical.
+    expect(next["opencode-go/longcat-2.5-preview-free"]).toBe(models["opencode-go/longcat-2.5-preview-free"]);
+    expect(next["other/m1"]).toBe(models["other/m1"]);
+  });
+
+  it("fully re-prefixes non-standard aliases missing the old `<name>/` prefix", () => {
+    const models = { "legacy-k3": model("legacy-k3", "kimi-code") };
+    const { models: next } = rekeyModelsOnProviderRename(models, "kimi-code", "renamed");
+    expect(Object.keys(next)).toEqual(["renamed/legacy-k3"]);
+    expect(next["renamed/legacy-k3"].provider).toBe("renamed");
+  });
+
+  it("leaves other providers' models alone when renaming", () => {
+    const models = { "a/m1": model("a/m1", "a"), "b/m1": model("b/m1", "b") };
+    const { models: next, aliasMap } = rekeyModelsOnProviderRename(models, "a", "c");
+    expect(Object.keys(next).sort()).toEqual(["b/m1", "c/m1"]);
+    expect(aliasMap).toEqual({ "a/m1": "c/m1" });
+  });
+
+  it("suffixes instead of overwriting when the renamed alias collides (rename into an occupied name)", () => {
+    // Renaming provider `a` to `b` while `b` already has b/m1 — the re-keyed
+    // entry must not silently swallow b's model.
+    const models = {
+      "a/m1": model("a/m1", "a"),
+      "b/m1": model("b/m1", "b"),
+    };
+    const { models: next, aliasMap } = rekeyModelsOnProviderRename(models, "a", "b");
+    expect(next["b/m1"].model).toBe("m1"); // b's original survives
+    expect(next["b/m1-2"].provider).toBe("b"); // a's model re-keyed with suffix
+    expect(aliasMap).toEqual({ "a/m1": "b/m1-2" });
+    expect(Object.keys(next).sort()).toEqual(["b/m1", "b/m1-2"]);
+  });
+
+  it("suffixes when a legacy bare alias collides with the standard alias after rename", () => {
+    // Pre-v0.6 data: bare alias `k3` (no provider prefix) and `kimi/k3` both
+    // belong to `kimi`; both re-key to `x/k3`.
+    const models = {
+      k3: { ...model("k3", "kimi"), model: "k3" },
+      "kimi/k3": model("kimi/k3", "kimi"),
+    };
+    const { models: next, aliasMap } = rekeyModelsOnProviderRename(models, "kimi", "x");
+    expect(Object.keys(next).sort()).toEqual(["x/k3", "x/k3-2"]);
+    expect(next["x/k3"]).toBeDefined();
+    expect(next["x/k3-2"]).toBeDefined();
+    // Both old keys are mapped; the pool re-key consumes the map, so every
+    // entry lands on a distinct target.
+    expect(Object.values(aliasMap).sort()).toEqual(["x/k3", "x/k3-2"]);
+  });
+});
+
+describe("copyModelsOnProviderDuplicate", () => {
+  it("clones the provider's models and KEEPS the source (copy, not move)", () => {
+    const models = {
+      "opencode-go/longcat": model("opencode-go/longcat", "opencode-go"),
+      "other/m1": model("other/m1", "other"),
+    };
+    const next = copyModelsOnProviderDuplicate(models, "opencode-go", "opencode-go-copy");
+    expect(Object.keys(next).sort()).toEqual([
+      "opencode-go-copy/longcat",
+      "opencode-go/longcat",
+      "other/m1",
+    ]);
+    expect(next["opencode-go/longcat"].provider).toBe("opencode-go");
+    expect(next["opencode-go-copy/longcat"].provider).toBe("opencode-go-copy");
+  });
+
+  it("fully re-prefixes non-standard aliases and suffixes collisions", () => {
+    const models = {
+      "legacy-k3": { ...model("legacy-k3", "kimi"), model: "k3" },
+      "dupe/legacy-k3": model("dupe/legacy-k3", "other"),
+    };
+    const next = copyModelsOnProviderDuplicate(models, "kimi", "dupe");
+    // `dupe/legacy-k3` is taken → the clone gets a suffix.
+    expect(Object.keys(next).sort()).toEqual([
+      "dupe/legacy-k3",
+      "dupe/legacy-k3-2",
+      "legacy-k3",
+    ]);
+  });
+});

src/lib/providerRename.ts

@@ -0,0 +1,79 @@
+import type { Model } from "../types";
+
+export interface ProviderRenameResult {
+  models: Record<string, Model>;
+  /** old alias → new alias (collision-deduped), for callers fixing alias references. */
+  aliasMap: Record<string, string>;
+}
+
+function rekeyedAlias(alias: string, oldName: string, newName: string): string {
+  const prefix = `${oldName}/`;
+  return alias.startsWith(prefix)
+    ? newName + alias.slice(oldName.length)
+    : `${newName}/${alias}`;
+}
+
+/**
+ * Re-key a provider's model aliases after the provider is renamed so the
+ * `<provider>/<model>` prefix tracks the new name. Only patching
+ * `m.provider` (the old behaviour) left stale prefixes like
+ * `opencode-go-copy/...` pointing at the renamed provider — zombie aliases
+ * that the UI then re-saved to config.toml and SQLite forever.
+ *
+ * Collision-safe: a target alias that is already taken (renaming into an
+ * existing provider's name, or a legacy bare alias colliding with a standard
+ * `<name>/<model>` one — both map to the same new key) gets a `-2`/`-3`
+ * suffix instead of silently overwriting the other entry. Non-standard
+ * aliases without the `<oldName>/` prefix fall back to a full re-prefix.
+ */
+export function rekeyModelsOnProviderRename(
+  models: Record<string, Model>,
+  oldName: string,
+  newName: string
+): ProviderRenameResult {
+  const next: Record<string, Model> = {};
+  const toRename: Array<[string, Model]> = [];
+  for (const [key, m] of Object.entries(models)) {
+    if (m.provider === oldName) toRename.push([key, m]);
+    else next[key] = m;
+  }
+  const aliasMap: Record<string, string> = {};
+  for (const [key, m] of toRename) {
+    const base = rekeyedAlias(key, oldName, newName);
+    let target = base;
+    let n = 2;
+    while (target in next) {
+      target = `${base}-${n}`;
+      n++;
+    }
+    next[target] = { ...m, alias: target, provider: newName };
+    aliasMap[key] = target;
+  }
+  return { models: next, aliasMap };
+}
+
+/**
+ * Clone a provider's models under the duplicate's name — COPY semantics:
+ * the source provider keeps its models (unlike rekeyModelsOnProviderRename,
+ * which moves them). Collisions with any existing alias get a `-2`/`-3`
+ * suffix.
+ */
+export function copyModelsOnProviderDuplicate(
+  models: Record<string, Model>,
+  oldName: string,
+  newName: string
+): Record<string, Model> {
+  const next = { ...models };
+  for (const [key, m] of Object.entries(models)) {
+    if (m.provider !== oldName) continue;
+    const base = rekeyedAlias(key, oldName, newName);
+    let target = base;
+    let n = 2;
+    while (target in next) {
+      target = `${base}-${n}`;
+      n++;
+    }
+    next[target] = { ...m, alias: target, provider: newName };
+  }
+  return next;
+}

src/lib/referenceCleanup.test.ts

@@ -0,0 +1,124 @@
+import { describe, expect, it } from "vitest";
+import { purgeSubagentPoolAliases, rekeySubagentPoolAliases } from "./referenceCleanup";
+
+const rawWithPool = {
+  thinking: { enabled: true },
+  secondary_model: {
+    default_model: "opencode-go/longcat-2.5-preview-free",
+    model: "opencode-go/longcat-2.5-preview-free",
+    models: {
+      "TokenHub/deepseek-v4.1-flash": "",
+      "opencode-go/longcat-2.5-preview-free": "",
+      "opencode-go/mimo-v2.6-flash": "",
+      "opencode-go-copy/deepseek-v4-flash": "route to deepseek",
+    },
+    force: false,
+  },
+  services: { moonshot_fetch: { api_key: "" } },
+};
+
+describe("rekeySubagentPoolAliases", () => {
+  it("re-keys pool keys and defaults when the provider is renamed", () => {
+    const next = rekeySubagentPoolAliases(rawWithPool, {
+      "opencode-go-copy/deepseek-v4-flash": "opencode-go/deepseek-v4-flash",
+    }) as typeof rawWithPool;
+
+    const section = next.secondary_model as Record<string, any>;
+    expect(Object.keys(section.models).sort()).toEqual([
+      "TokenHub/deepseek-v4.1-flash",
+      "opencode-go/deepseek-v4-flash",
+      "opencode-go/longcat-2.5-preview-free",
+      "opencode-go/mimo-v2.6-flash",
+    ]);
+    // Descriptions survive the re-key.
+    expect(section.models["opencode-go/deepseek-v4-flash"]).toBe("route to deepseek");
+    // Untouched defaults stay put.
+    expect(section.default_model).toBe("opencode-go/longcat-2.5-preview-free");
+    // Unrelated sections are preserved.
+    expect(next.thinking).toEqual({ enabled: true });
+  });
+
+  it("follows the default when it referenced a renamed alias", () => {
+    const raw = {
+      secondary_model: {
+        default_model: "old/m1",
+        model: "old/m1",
+        models: { "old/m1": "" },
+      },
+    };
+    const next = rekeySubagentPoolAliases(raw, { "old/m1": "new/m1" }) as any;
+    expect(next.secondary_model.default_model).toBe("new/m1");
+    expect(next.secondary_model.model).toBe("new/m1");
+    expect(Object.keys(next.secondary_model.models)).toEqual(["new/m1"]);
+  });
+
+  it("returns the same reference when nothing matches", () => {
+    expect(rekeySubagentPoolAliases(rawWithPool, { "a/b": "c/d" })).toBe(rawWithPool);
+  });
+
+  it("never injects an empty models table into the implicit single-model form", () => {
+    // `{ model: "old/m1" }` with no pool table: re-keying the default must
+    // not add `[secondary_model.models]` — an empty table is invalid for the
+    // engine (its default has to be a key of the table).
+    const raw = { secondary_model: { model: "old/m1" } };
+    const next = rekeySubagentPoolAliases(raw, { "old/m1": "new/m1" }) as any;
+    expect(next.secondary_model).toEqual({ model: "new/m1" });
+    expect(next.secondary_model.models).toBeUndefined();
+  });
+
+  it("never injects a models table into the force form", () => {
+    const raw = { secondary_model: { force: true, default_model: "old/m1", model: "old/m1" } };
+    const next = rekeySubagentPoolAliases(raw, { "old/m1": "new/m1" }) as any;
+    expect(next.secondary_model).toEqual({
+      force: true,
+      default_model: "new/m1",
+      model: "new/m1",
+    });
+    expect(next.secondary_model.models).toBeUndefined();
+  });
+});
+
+describe("purgeSubagentPoolAliases", () => {
+  it("removes deleted aliases and keeps the rest of the section intact", () => {
+    const next = purgeSubagentPoolAliases(rawWithPool, [
+      "opencode-go-copy/deepseek-v4-flash",
+    ]) as typeof rawWithPool;
+    const section = next.secondary_model as Record<string, any>;
+    expect(Object.keys(section.models).sort()).toEqual([
+      "TokenHub/deepseek-v4.1-flash",
+      "opencode-go/longcat-2.5-preview-free",
+      "opencode-go/mimo-v2.6-flash",
+    ]);
+    expect(section.default_model).toBe("opencode-go/longcat-2.5-preview-free");
+    expect(section.force).toBe(false);
+    expect(next.thinking).toEqual({ enabled: true });
+  });
+
+  it("falls back to the first remaining key when the default was purged", () => {
+    const next = purgeSubagentPoolAliases(rawWithPool, [
+      "opencode-go/longcat-2.5-preview-free",
+    ]) as any;
+    const section = next.secondary_model;
+    expect(section.models).not.toHaveProperty("opencode-go/longcat-2.5-preview-free");
+    expect(section.default_model).toBe("TokenHub/deepseek-v4.1-flash");
+    expect(section.model).toBe("TokenHub/deepseek-v4.1-flash");
+  });
+
+  it("drops the whole section when the pool becomes empty", () => {
+    const raw = {
+      thinking: { enabled: true },
+      secondary_model: {
+        default_model: "p/m",
+        model: "p/m",
+        models: { "p/m": "" },
+      },
+    };
+    const next = purgeSubagentPoolAliases(raw, ["p/m"]) as any;
+    expect(next).not.toHaveProperty("secondary_model");
+    expect(next.thinking).toEqual({ enabled: true });
+  });
+
+  it("returns the same reference when no pool entry is affected", () => {
+    expect(purgeSubagentPoolAliases(rawWithPool, ["other/m9"])).toBe(rawWithPool);
+  });
+});

src/lib/referenceCleanup.ts

@@ -0,0 +1,117 @@
+/**
+ * Keep `[secondary_model]` (the subagent model pool in `raw_other`) free of
+ * dangling model-alias references.
+ *
+ * Model aliases are deleted or re-keyed in several places (provider delete,
+ * single-model delete, provider rename). The pool stores aliases as its
+ * `models` table keys and as the `default_model` / `model` values — none of
+ * those may outlive the alias they point at, or the engine resolves the pool
+ * entry to nothing and the zombie key lingers in config.toml forever.
+ */
+
+function asRecord(value: unknown): Record<string, unknown> {
+  return value && typeof value === "object" && !Array.isArray(value)
+    ? (value as Record<string, unknown>)
+    : {};
+}
+
+/**
+ * Re-key pool aliases after a provider rename: `aliasMap` maps old alias →
+ * new alias. Pool keys keep their descriptions; `default_model` / `model`
+ * follow when they referenced a renamed alias. Returns the input reference
+ * when nothing changed.
+ */
+export function rekeySubagentPoolAliases(
+  rawOther: unknown,
+  aliasMap: Record<string, string>
+): unknown {
+  if (Object.keys(aliasMap).length === 0) return rawOther;
+  const root = asRecord(rawOther);
+  const section = asRecord(root.secondary_model);
+  if (Object.keys(section).length === 0) return rawOther;
+
+  const models = asRecord(section.models);
+  let changed = false;
+  const nextModels: Record<string, unknown> = {};
+  for (const [k, v] of Object.entries(models)) {
+    const mapped = aliasMap[k];
+    if (mapped && mapped !== k) {
+      nextModels[mapped] = v;
+      changed = true;
+    } else {
+      nextModels[k] = v;
+    }
+  }
+
+  let nextDefault = section.default_model;
+  let nextModel = section.model;
+  if (typeof nextDefault === "string" && aliasMap[nextDefault]) {
+    nextDefault = aliasMap[nextDefault];
+    changed = true;
+  }
+  if (typeof nextModel === "string" && aliasMap[nextModel]) {
+    nextModel = aliasMap[nextModel];
+    changed = true;
+  }
+  if (!changed) return rawOther;
+
+  // Only write a `models` table when the section had one — the implicit
+  // single-model form (`{ model: "..." }`) and the force form must not gain
+  // an empty `[secondary_model.models]` table, which the engine can never
+  // validate (its default must be a key of the table).
+  const hadModelsTable = section.models !== undefined;
+  const nextSection: Record<string, unknown> = { ...section };
+  if (hadModelsTable || Object.keys(nextModels).length > 0) {
+    nextSection.models = nextModels;
+  }
+  if (nextDefault !== section.default_model) nextSection.default_model = nextDefault;
+  if (nextModel !== section.model) nextSection.model = nextModel;
+  return { ...root, secondary_model: nextSection };
+}
+
+/**
+ * Remove pool entries pointing at deleted aliases. When the pool's default
+ * referenced a deleted alias it falls back to the first remaining key; when
+ * the pool becomes empty the whole `[secondary_model]` section is dropped
+ * (an empty pool is invalid for the engine — subagents then inherit the
+ * primary model). Returns the input reference when nothing changed.
+ */
+export function purgeSubagentPoolAliases(
+  rawOther: unknown,
+  removedAliases: Iterable<string>
+): unknown {
+  const removed = new Set(removedAliases);
+  if (removed.size === 0) return rawOther;
+  const root = asRecord(rawOther);
+  const section = asRecord(root.secondary_model);
+  if (Object.keys(section).length === 0) return rawOther;
+
+  const models = asRecord(section.models);
+  const kept: Record<string, unknown> = {};
+  let removedAny = false;
+  for (const [k, v] of Object.entries(models)) {
+    if (removed.has(k)) {
+      removedAny = true;
+    } else {
+      kept[k] = v;
+    }
+  }
+  const defaultGone =
+    (typeof section.default_model === "string" && removed.has(section.default_model)) ||
+    (typeof section.model === "string" && removed.has(section.model));
+  if (!removedAny && !defaultGone) return rawOther;
+
+  if (Object.keys(kept).length === 0) {
+    const nextRoot: Record<string, unknown> = { ...root };
+    delete nextRoot.secondary_model;
+    return nextRoot;
+  }
+
+  const nextSection: Record<string, unknown> = { ...section, models: kept };
+  if (defaultGone) {
+    const fallback = Object.keys(kept)[0];
+    nextSection.default_model = fallback;
+    nextSection.model = fallback;
+  }
+  return { ...root, secondary_model: nextSection };
+}

src/lib/thinking-efforts.test.ts

@@ -0,0 +1,139 @@
+import { describe, expect, it } from "vitest";
+import {
+  THINKING_EFFORTS,
+  normalizeSupportEfforts,
+  readSupportEfforts,
+  resolveThinkingEffortSupport,
+} from "./thinking-efforts";
+
+// ---------------------------------------------------------------------------
+// support_efforts normalization — an absent / malformed key means "unknown",
+// never "no tiers supported"
+// ---------------------------------------------------------------------------
+
+describe("normalizeSupportEfforts", () => {
+  it("keeps a non-empty list of strings", () => {
+    expect(normalizeSupportEfforts(["low", "high"])).toEqual(["low", "high"]);
+  });
+
+  it("drops blanks and non-strings but keeps the order", () => {
+    expect(normalizeSupportEfforts(["low", "", 3, null, "max"])).toEqual([
+      "low",
+      "max",
+    ]);
+  });
+
+  it("returns null for absent / empty / non-array values", () => {
+    expect(normalizeSupportEfforts(undefined)).toBeNull();
+    expect(normalizeSupportEfforts([])).toBeNull();
+    expect(normalizeSupportEfforts(["  "])).toBeNull();
+    expect(normalizeSupportEfforts("low")).toBeNull();
+  });
+});
+
+describe("readSupportEfforts", () => {
+  const model = (raw_other: unknown) =>
+    ({ alias: "a", provider: "p", model: "m", max_context_size: 0, display_name: null, raw_other }) as const;
+
+  it("reads the key from a model entry's preserved config fields", () => {
+    expect(readSupportEfforts(model({ support_efforts: ["low", "max"] }))).toEqual([
+      "low",
+      "max",
+    ]);
+  });
+
+  it("returns null when the entry or the key is absent", () => {
+    expect(readSupportEfforts(undefined)).toBeNull();
+    expect(readSupportEfforts(model(undefined))).toBeNull();
+    expect(readSupportEfforts(model({ other: 1 }))).toBeNull();
+  });
+});
+
+// ---------------------------------------------------------------------------
+// Tier resolution — three layers, in priority order
+// ---------------------------------------------------------------------------
+
+describe("resolveThinkingEffortSupport — declared support_efforts wins", () => {
+  const support = resolveThinkingEffortSupport({
+    supportEfforts: ["low", "high", "max"],
+    reasoning: true,
+  });
+
+  it("exposes the declared list and keeps thinking enabled", () => {
+    expect(support.declared).toEqual(["low", "high", "max"]);
+    expect(support.thinkingSupported).toBe(true);
+  });
+
+  it("enables exactly the declared tiers", () => {
+    for (const tier of THINKING_EFFORTS) {
+      expect(support.levels[tier].enabled).toBe(tier === "low" || tier === "high" || tier === "max");
+    }
+  });
+
+  it("explains a disabled tier with the supported list", () => {
+    expect(support.levels.medium).toEqual({
+      enabled: false,
+      note: { kind: "tierUnsupported", supported: ["low", "high", "max"] },
+    });
+    expect(support.levels.xhigh.enabled).toBe(false);
+  });
+
+  it("does not disable thinking even when models.dev says reasoning: false", () => {
+    const conflicting = resolveThinkingEffortSupport({
+      supportEfforts: ["high"],
+      reasoning: false,
+    });
+    expect(conflicting.thinkingSupported).toBe(true);
+    expect(conflicting.levels.low.enabled).toBe(false);
+    expect(conflicting.levels.high.enabled).toBe(true);
+  });
+
+  it("matches tiers case-insensitively", () => {
+    const upper = resolveThinkingEffortSupport({ supportEfforts: ["HIGH"] });
+    expect(upper.levels.high.enabled).toBe(true);
+  });
+});
+
+describe("resolveThinkingEffortSupport — model without thinking", () => {
+  const support = resolveThinkingEffortSupport({ reasoning: false });
+
+  it("disables the whole thinking area", () => {
+    expect(support.thinkingSupported).toBe(false);
+    expect(support.declared).toBeNull();
+    for (const tier of THINKING_EFFORTS) {
+      expect(support.levels[tier]).toEqual({
+        enabled: false,
+        note: { kind: "thinkingUnsupported" },
+      });
+    }
+  });
+});
+
+describe("resolveThinkingEffortSupport — nothing known about the model", () => {
+  it("offers low/medium/high plain and flags max/xhigh as undeclared", () => {
+    const support = resolveThinkingEffortSupport({});
+    expect(support.thinkingSupported).toBe(true);
+    expect(support.declared).toBeNull();
+    for (const tier of ["low", "medium", "high"] as const) {
+      expect(support.levels[tier]).toEqual({ enabled: true });
+    }
+    for (const tier of ["max", "xhigh"] as const) {
+      expect(support.levels[tier]).toEqual({
+        enabled: true,
+        note: { kind: "tierUndeclared" },
+      });
+    }
+  });
+
+  it("treats an undefined reasoning flag as unknown, not as unsupported", () => {
+    expect(
+      resolveThinkingEffortSupport({ reasoning: undefined }).thinkingSupported
+    ).toBe(true);
+  });
+
+  it("ignores a malformed support_efforts value and falls back to unknown", () => {
+    const support = resolveThinkingEffortSupport({ supportEfforts: "low" });
+    expect(support.declared).toBeNull();
+    expect(support.levels.low).toEqual({ enabled: true });
+  });
+});

src/lib/thinking-efforts.ts

@@ -0,0 +1,127 @@
+/**
+ * Thinking-effort capability resolution.
+ *
+ * kimi-code's `[thinking] effort` is a free-form string (low/medium/high/
+ * xhigh/max in practice) that the CLI forwards verbatim to OpenAI-compatible
+ * upstreams as `reasoning_effort`. Which tiers a given model accepts is
+ * declared per model in config.toml (`[models."<alias>"] support_efforts`);
+ * models.dev only carries a boolean `reasoning` flag with no tier list. The
+ * two sources are merged here so the UI can grey out tiers the current default
+ * model would reject, without ever rejecting a stored value on read.
+ */
+
+import type { Model } from "../types";
+
+export const THINKING_EFFORTS = [
+  "low",
+  "medium",
+  "high",
+  "xhigh",
+  "max",
+] as const;
+
+export type ThinkingEffort = (typeof THINKING_EFFORTS)[number];
+
+/** Why a tier is blocked, or (for undeclared models) merely unverified. */
+export type EffortNote =
+  | { kind: "thinkingUnsupported" }
+  | { kind: "tierUnsupported"; supported: string[] }
+  | { kind: "tierUndeclared" };
+
+export interface EffortAvailability {
+  /** Whether the tier can be picked. */
+  enabled: boolean;
+  /** Advisory message — set for blocked tiers and for unverified ones. */
+  note?: EffortNote;
+}
+
+export interface ThinkingEffortSupport {
+  /** False only when the model is known not to support thinking at all. */
+  thinkingSupported: boolean;
+  /** Declared tiers, or null when nothing is known about the model. */
+  declared: string[] | null;
+  levels: Record<ThinkingEffort, EffortAvailability>;
+}
+
+/** Tiers offered unverified when a model declares nothing at all. */
+const ALWAYS_OFFERED = new Set<string>(["low", "medium", "high"]);
+
+/**
+ * Normalize a `support_efforts` value into a comparable list, or null when the
+ * key is absent / malformed (treating it as "unknown", not "none supported").
+ */
+export function normalizeSupportEfforts(value: unknown): string[] | null {
+  if (!Array.isArray(value)) return null;
+  const tiers = value
+    .filter((v): v is string => typeof v === "string")
+    .map((v) => v.trim())
+    .filter((v) => v.length > 0);
+  return tiers.length > 0 ? tiers : null;
+}
+
+/** Read `support_efforts` from a model entry's preserved config.toml fields. */
+export function readSupportEfforts(model: Model | undefined): string[] | null {
+  if (!model) return null;
+  const raw = model.raw_other;
+  if (!raw || typeof raw !== "object" || Array.isArray(raw)) return null;
+  return normalizeSupportEfforts(
+    (raw as Record<string, unknown>).support_efforts
+  );
+}
+
+function buildLevels(
+  resolve: (effort: ThinkingEffort) => EffortAvailability
+): Record<ThinkingEffort, EffortAvailability> {
+  const levels = {} as Record<ThinkingEffort, EffortAvailability>;
+  for (const effort of THINKING_EFFORTS) levels[effort] = resolve(effort);
+  return levels;
+}
+
+/**
+ * Resolve per-tier availability from the two capability sources, in priority
+ * order:
+ *  1. `supportEfforts` declared → tiers follow the list exactly.
+ *  2. models.dev `reasoning === false` → the whole thinking area is off.
+ *  3. Nothing known → low/medium/high are offered plain, max/xhigh stay
+ *     selectable but carry a "not declared" note (upstream may reject them).
+ */
+export function resolveThinkingEffortSupport(input: {
+  supportEfforts?: unknown;
+  reasoning?: boolean;
+}): ThinkingEffortSupport {
+  const declared = normalizeSupportEfforts(input.supportEfforts);
+  if (declared) {
+    const supported = new Set(declared.map((t) => t.toLowerCase()));
+    return {
+      thinkingSupported: true,
+      declared,
+      levels: buildLevels((effort) =>
+        supported.has(effort)
+          ? { enabled: true }
+          : {
+              enabled: false,
+              note: { kind: "tierUnsupported", supported: declared },
+            }
+      ),
+    };
+  }
+  if (input.reasoning === false) {
+    return {
+      thinkingSupported: false,
+      declared: null,
+      levels: buildLevels(() => ({
+        enabled: false,
+        note: { kind: "thinkingUnsupported" },
+      })),
+    };
+  }
+  return {
+    thinkingSupported: true,
+    declared: null,
+    levels: buildLevels((effort) =>
+      ALWAYS_OFFERED.has(effort)
+        ? { enabled: true }
+        : { enabled: true, note: { kind: "tierUndeclared" } }
+    ),
+  };
+}

src/types/index.ts

@@ -111,11 +111,12 @@ export interface DiscoveredModel {
 export interface ThinkingConfig {
   enabled?: boolean;
   /**
-   * Effort tier. `max` is read-compatible only — upstream removed the tier
-   * (old configs auto-migrate to `high`); the UI normalizes it and the
-   * serialization path never writes it.
+   * Effort tier, forwarded verbatim to OpenAI-compatible upstreams as
+   * `reasoning_effort`. Upstream accepts a free-form string; the UI offers
+   * low/medium/high/max/xhigh and derives per-model support from
+   * `[models."<alias>"] support_efforts`.
    */
-  effort?: "low" | "medium" | "high" | "max";
+  effort?: string;
   /**
    * Keep thinking content. The legacy off values (`false`, `0`, "no", "none",
    * `null`) are read-compatible — old configs may carry them; the UI
@@ -171,6 +172,13 @@ export interface AgentSettings {
   background?: BackgroundConfig;
   /** Top-level `[watch]` section (kimi-code 2.0.1+). */
   watch?: WatchConfig;
+  /** Top-level config.toml key (kimi-code #3962): let clients auto-generate
+   *  session titles. Default true; only an explicit false disables. */
+  auto_session_title?: boolean;
+  /** Top-level config.toml key (kimi-code #3995): repeat-breaker for
+   *  consecutively repeated identical tool calls. Default true; env
+   *  KIMI_CODE_REPEAT_BREAKER outranks this config. */
+  repeat_breaker?: boolean;
   permission?: {
     rules?: PermissionRule[];
     /** kimi-code `[permission]` key. Default true when the key is

website/index.html

@@ -50,8 +50,8 @@
             "url": "https://billowliu2.github.io/KimiSwitch/",
             "applicationCategory": "DeveloperApplication",
             "operatingSystem": "Windows 10, Windows 11, macOS, Linux",
-            "softwareVersion": "0.7.12",
-            "dateModified": "2026-09-03",
+            "softwareVersion": "0.8.2",
+            "dateModified": "2026-10-09",
             "description": "桌面端 LLM 供应商配置管理器：22 个预设选好即写入 config.toml，用量与账单直读各厂商官方接口。Tauri + Rust，开源 MIT。",
             "downloadUrl": "https://github.com/billowliu2/KimiSwitch/releases",
             "softwareHelp": "https://github.com/billowliu2/KimiSwitch",

website/src/i18n.tsx

@@ -16,10 +16,10 @@ const zh = {
     performance: "性能",
     changelog: "更新日志",
     download: "下载",
-    downloadBtn: "下载 v0.7.12",
+    downloadBtn: "下载 v0.8.2",
   },
   hero: {
-    badge: "v0.7.12 · 开源 MIT · Windows / macOS / Linux",
+    badge: "v0.8.2 · 开源 MIT · Windows / macOS / Linux",
     titleBefore: "统一管理你的",
     titleAccent: "AI 供应商",
     titleAfter: "",
@@ -29,7 +29,7 @@ const zh = {
     source: "查看源码",
     stats: [
       { value: "22", label: "预设供应商" },
-      { value: "7495", label: "模型价格库" },
+      { value: "8467", label: "模型价格库" },
       { value: "MIT", label: "开源协议" },
     ],
   },
@@ -76,7 +76,7 @@ const zh = {
       {
         id: "model-discovery",
         title: "模型自动发现",
-        desc: "可用模型直接问供应商 API；显示名 / 上下文 / 能力取自 models.dev 快照（当前 7495 个模型、212 家供应商）。",
+        desc: "可用模型直接问供应商 API；显示名 / 上下文 / 能力取自 models.dev 快照（当前 8,467 个模型、225 家供应商）。",
         img: "screenshots/kimi-cli-select-model.png",
       },
       {
@@ -89,7 +89,7 @@ const zh = {
   },
   showcase: {
     title: "界面演示",
-    subtitle: "以下截图取自 v0.7.12 实机运行。",
+    subtitle: "以下截图取自 v0.8.2 实机运行。",
     items: [
       {
         src: "screenshots/usage-config-page.png",
@@ -140,6 +140,42 @@ const zh = {
     syncedNote: "数据已同步 GitHub Releases",
     fallbackNote: "内置版本记录",
     entries: [
+      {
+        version: "v0.8.2",
+        date: "2026-10-09",
+        items: [
+          "仪表盘重启后首次打开从约 25 秒降到 1 秒内：用量解析快照持久化到本地，启动直读并在后台校验，有变化自动无感更新（6.4 万条记录实测）",
+          "切换时间范围几乎瞬时：热力图 / 模型总览 / 各范围汇总按记录集缓存；去除原「全时段」聚合的整表无用遍历",
+          "界面不再闪加载态：切范围与重新进入页面先呈现上次数据再后台静默刷新；供应商卡片墙由 N 次配置全量加载降为 1 次",
+          "models.dev 快照同步至 2026-10-09（8,467 模型 / 225 供应商）",
+        ],
+      },
+      {
+        version: "v0.8.1",
+        date: "2026-10-04",
+        items: [
+          "全局配置新增两个顶层开关：自动生成会话标题（auto_session_title）与重复调用拦截（repeat_breaker），默认开启",
+          "最大输出未匹配 models.dev 时回填默认值 131072（128K 兜底）",
+          "修复 [watch] enabled 默认值显示（上游 #4015 翻回默认开启）；「最大输出」列描述对齐上游 #4091",
+        ],
+      },
+      {
+        version: "v0.8.0",
+        date: "2026-10-03",
+        items: [
+          "模型映射新增「最大输出 Token」列：逐模型设置 max_output_size（留空用上游默认），models.dev 有 output 上限时显示可点击参考值，打开编辑页自动补全空值",
+          "models.dev 快照新增 output 上限字段，同步至 2026-10-03（8,385 模型 / 226 供应商，output 覆盖 97.4%）",
+          "移除无效的「声明支持 1M」复选框（勾选从不生效），由「最大输出 Token」列取代",
+        ],
+      },
+      {
+        version: "v0.7.23",
+        date: "2026-09-30",
+        items: [
+          "思考等级五档 + 能力感知：Low / Medium / High / XHigh / Max，按当前默认模型声明的 support_efforts 自动置灰不支持档位并提示原因",
+          "models.dev 标记不支持思考的模型整体禁用思考区；未声明档位的模型保留可选并提示上游可能拒绝",
+        ],
+      },
       {
         version: "v0.7.12",
         date: "2026-09-03",
@@ -322,7 +358,7 @@ const zh = {
   },
   download: {
     title: "下载 Kimi Switch",
-    subtitle: "当前版本 v0.7.12 · Windows / macOS / Linux 三平台已发布",
+    subtitle: "当前版本 v0.8.2 · Windows / macOS / Linux 三平台已发布",
     autoUpdate: "更新检查跑在启动时和每 8 小时的周期任务里，设置页也可以手动触发。",
     ready: "已发布",
     wip: "开发中",
@@ -391,10 +427,10 @@ const en: Dict = {
     performance: "Performance",
     changelog: "Changelog",
     download: "Download",
-    downloadBtn: "Download v0.7.12",
+    downloadBtn: "Download v0.8.2",
   },
   hero: {
-    badge: "v0.7.12 · Open source MIT · Windows / macOS / Linux",
+    badge: "v0.8.2 · Open source MIT · Windows / macOS / Linux",
     titleBefore: "One app for all your ",
     titleAccent: "AI providers",
     titleAfter: "",
@@ -404,7 +440,7 @@ const en: Dict = {
     source: "View source",
     stats: [
       { value: "22", label: "Provider presets" },
-      { value: "7495", label: "Model price DB" },
+      { value: "8467", label: "Model price DB" },
       { value: "MIT", label: "License" },
     ],
   },
@@ -451,7 +487,7 @@ const en: Dict = {
       {
         id: "model-discovery",
         title: "Model discovery",
-        desc: "Model lists are fetched from the provider API itself; display names, context sizes and capabilities come from a build-time models.dev snapshot (7495 models, 212 providers).",
+        desc: "Model lists are fetched from the provider API itself; display names, context sizes and capabilities come from a build-time models.dev snapshot (8,467 models, 225 providers).",
         img: "screenshots/kimi-cli-select-model.png",
       },
       {
@@ -464,7 +500,7 @@ const en: Dict = {
   },
   showcase: {
     title: "Screenshots",
-    subtitle: "All screenshots below are taken from the running v0.7.12 build.",
+    subtitle: "All screenshots below are taken from the running v0.8.2 build.",
     items: [
       {
         src: "screenshots/usage-config-page.png",
@@ -515,6 +551,42 @@ const en: Dict = {
     syncedNote: "Synced from GitHub Releases",
     fallbackNote: "Built-in release notes",
     entries: [
+      {
+        version: "v0.8.2",
+        date: "2026-10-09",
+        items: [
+          "First dashboard open after a restart drops from ~25s to under 1s: the usage snapshot is persisted locally and read straight at startup, then verified in the background with silent updates when things changed (measured at 64k records)",
+          "Switching the time range is near-instant: heatmap / model overview / per-range totals are cached per record set, and the wasteful full-table pass behind the 'all time' aggregate is gone",
+          "No more loading flicker: switching ranges or reopening the page shows the previous data first and revalidates silently; the provider card wall loads config once instead of once per card",
+          "models.dev snapshot synced to 2026-10-09 (8,467 models / 225 providers)",
+        ],
+      },
+      {
+        version: "v0.8.1",
+        date: "2026-10-04",
+        items: [
+          "Two new top-level switches in Global Settings: auto session title (auto_session_title) and repeat breaker (repeat_breaker), on by default",
+          "Max output backfills a 131072 (128K) default when models.dev has no match",
+          "Fixed the [watch] enabled default display (upstream #4015 flipped it back on); Max output column wording aligned with upstream #4091",
+        ],
+      },
+      {
+        version: "v0.8.0",
+        date: "2026-10-03",
+        items: [
+          "New per-model Max Output Token column: sets max_output_size in config.toml (blank keeps the upstream default), shows a click-to-fill reference when models.dev knows the cap, and backfills empty rows automatically",
+          "models.dev snapshot gains the output-limit field, synced to 2026-10-03 (8,385 models / 226 providers, 97.4% output coverage)",
+          "Removed the ineffective 'Supports 1M' checkbox (it was never persisted), replaced by the Max Output Token column",
+        ],
+      },
+      {
+        version: "v0.7.23",
+        date: "2026-09-30",
+        items: [
+          "Thinking levels expanded to five tiers with capability awareness: Low / Medium / High / XHigh / Max, greying out tiers the current default model's support_efforts doesn't declare, with an explanation",
+          "Models flagged non-reasoning by models.dev disable the whole thinking area; undeclared tiers stay selectable with an upstream-may-reject hint",
+        ],
+      },
       {
         version: "v0.7.12",
         date: "2026-09-03",
@@ -696,7 +768,7 @@ const en: Dict = {
   },
   download: {
     title: "Download Kimi Switch",
-    subtitle: "Current version v0.7.12 · Windows, macOS and Linux now released",
+    subtitle: "Current version v0.8.2 · Windows, macOS and Linux now released",
     autoUpdate: "Update checks run at startup and every 8 hours; Settings also has a manual check.",
     ready: "Available",
     wip: "In development",

website/src/sections/Download.tsx

@@ -8,13 +8,13 @@ const icons: Record<string, Icon> = {
   linux: LinuxLogo,
 };
 
-const VERSION = "0.7.12";
+const VERSION = "0.8.2";
 const GITHUB = "https://github.com/billowliu2/KimiSwitch";
 const MIRROR_RELEASES = "https://git.codingplan.site/admin/KimiCodeSwitch/releases";
 const dl = (file: string) => `${GITHUB}/releases/download/v${VERSION}/${file}`;
 
 /** Per-platform download assets for the current release (names match the CI
- *  release workflow exactly — verified against the published v0.7.12 assets:
+ *  release workflow exactly — verified against the published v0.8.2 assets:
  *  tauri-bundler on CI produces `Kimi.Switch_…` (dots) on every platform,
  *  except the rpm which uses `Kimi.Switch-<version>-1` (hyphens).
  *  Filenames interpolate VERSION, so a release bump only changes the