feat: 阈值预警/加力钱包/api_key_env/版本跟踪器/一键体检 — bump to v0.7.20

服务端 api.kimi.com/coding/v1/usages 已由绝对 used/limit 行切换为
quota 模型（usages.limit_5h/limit_7d/limit_month_total/limit_month_code，
条目为 used_ratio 0-1 比率 + reset_time），见上游 #3787。

- parse_kimi_coding 优先解析新结构，旧 limits/usage 结构保留为回退
- used_ratio >1 时按已是百分数处理（防御）
- 新 tier id month_code + 前端 planLabel/i18n 中英文案
- 新增 3 个测试：四窗口解析 / 比率防御 / 空窗口回退旧结构

package.json

@@ -1,7 +1,7 @@
 {
   "name": "kimiswitch",
   "private": true,
-  "version": "0.7.18",
+  "version": "0.7.20",
   "type": "module",
   "scripts": {
     "dev": "vite",

release-notes-v0.7.19.md

@@ -0,0 +1,10 @@
+# KimiSwitch v0.7.19
+
+## 适配
+
+- **适配 kimi-code 0.43.0 实验 flag 镜像（7→5）**：移除已转正的 `auto_session_title` 开关（上游 #3749，AI 会话标题不再需要 flag，首轮对话后自动生成、可在重命名处重新生成）。设置页「实验性 flag」、环境变量探测（`get_experimental_env_status`）、中英文案与测试断言同步清理；上游 0.43.0 注册表现剩 5 个 flag： `wait_for` / `tool-select` / `notify_user` / `tower` / `subagent_fork`
+- **兼容性核查结论（无需改动）**：0.43.0 的 wire 持久化重建（#3737）不影响用量统计——`usage.record` / `token_counting.measured` / `context.apply_compaction` / `turn.prompt` 记录全部保留（`turn.prompt` 仅新增可选 `turnId` 字段）；新增配置项 `loop_control.compaction_max_attempts`（#3750）经现有合并式写入验证保存不丢失；MCP `deferred` 字段（#3667）不涉及本应用管理的配置节
+
+## 备注
+
+- 若此前在 `[experimental]` 中显式开启过 `auto_session_title`，残留键会被上游静默忽略，无需手动清理

release-notes-v0.7.20.md

@@ -0,0 +1,23 @@
+# KimiSwitch v0.7.20
+
+## 新增
+
+- **套餐用量阈值预警**：用量配置面板新增「预警阈值 %」（0-100，0 或空 = 关闭）。每次账单查询后，任一百分比用量窗口达到阈值即红色高亮 + ⚠ 标记，低于阈值自动消除；自动查询间隔会周期性重复评估
+- **Kimi For Coding 加力钱包**：0.43.1 起服务端 quota 响应中的 `boosterWallet`（加力钱包）现在解析展示为一行余额（💰 月度用量 / 总额 / 余额，含币种），与四个用量窗口并列
+- **供应商凭证环境变量化（`api_key_env`）**：适配 kimi-code 2.0.0 #3762——供应商编辑页「凭证来源」可切换「直接填写 / 环境变量」，密钥不写进 config.toml；导出时与 `api_key` 互斥；账单查询 / 获取模型列表会实时解析环境变量（变量未设置给出明确错误）；SQLite 新增列（含自动迁移）
+- **Kimi Code 版本跟踪器**：设置页新增「Kimi Code 版本」卡片——本机已装版本（`kimi --version` 检测）、上游最新版本、适配状态徽标（绿已适配 / 黄建议升级 KimiSwitch / 橙部分兼容 / 灰未安装），可一键打开上游 Releases
+- **供应商一键体检**：供应商列表新增「一键体检」按钮——并发探活所有启用供应商的推理端点（`GET /v1/models`）并顺带重新查询账单，每行状态点 + 延迟 tooltip + 顶部汇总。状态判定为三态：账单或探活任一真实通过 = 绿；鉴权失败必红；端点不支持模型列表（404）计中性不误报
+
+## 设置
+
+- **压缩重试次数**：Agent 设置新增 `loop_control.compaction_max_attempts`（上游 0.43.0 #3750，默认 5，0/空不写键）
+- **文件监听开关**：新增 `[watch] enabled` 开关（上游 2.0.1 #3892，2.0.2 起上游默认关闭），并探测 `KIMI_CODE_WATCH` 环境变量
+
+## 适配
+
+- **kimi-code 2.0.2 已适配确认**：wire 记录格式、实验 flag（5 个）、config 模式均无变化；`api_key_env` / `[watch]` / `[database]` 新增配置节经现有合并式写入验证无损往返
+
+## 备注
+
+- 版本跟踪器的「最新版本」查询直连 GitHub，受限网络环境下可能显示 `—`（已装版本与状态判定不受影响）
+- 体检跳过大模型平台类供应商（anthropic / google-genai / vertexai 无 `/v1/models` 路由），显示灰点「未检测」

src-tauri/Cargo.lock

@@ -1978,7 +1978,7 @@ dependencies = [
 
 [[package]]
 name = "kimiswitch"
-version = "0.7.18"
+version = "0.7.20"
 dependencies = [
  "anyhow",
  "chrono",

src-tauri/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "kimiswitch"
-version = "0.7.18"
+version = "0.7.20"
 description = "Kimi Switch - model config manager"
 authors = ["codingplan.site"]
 edition = "2021"

src-tauri/src/db.rs

@@ -81,6 +81,7 @@ pub fn init_db() -> DbResult<Connection> {
             provider_type TEXT NOT NULL,
             base_url TEXT,
             api_key TEXT,
+            api_key_env TEXT,
             env TEXT,
             note TEXT,
             official_url TEXT,
@@ -141,6 +142,9 @@ pub fn init_db() -> DbResult<Connection> {
     for stmt in [
         "ALTER TABLE providers ADD COLUMN icon TEXT",
         "ALTER TABLE providers ADD COLUMN icon_color TEXT",
+        // `api_key_env` (kimi-code 2.0.0+): the name of the environment
+        // variable holding the credential, an alternative to `api_key`.
+        "ALTER TABLE providers ADD COLUMN api_key_env TEXT",
     ] {
         if let Err(e) = conn.execute(stmt, []) {
             let msg = e.to_string();
@@ -162,28 +166,29 @@ pub fn load_config(agent: &Agent) -> DbResult<Config> {
     let mut providers = IndexMap::new();
     {
         let mut stmt = tx.prepare(
-            "SELECT name, provider_type, base_url, api_key, env, note, official_url, managed, enabled, active, icon, icon_color, raw_other
+            "SELECT name, provider_type, base_url, api_key, api_key_env, env, note, official_url, managed, enabled, active, icon, icon_color, raw_other
              FROM providers WHERE agent = ?1 ORDER BY id",
         )?;
         let provider_rows = stmt.query_map(params![agent.as_str()], |row| {
             let provider_type: String = row.get(1)?;
-            let env_json: Option<String> = row.get(4)?;
+            let env_json: Option<String> = row.get(5)?;
-            let raw_json: Option<String> = row.get(12)?;
+            let raw_json: Option<String> = row.get(13)?;
             Ok(Provider {
                 name: row.get(0)?,
                 provider_type: provider_type_for_str(&provider_type),
                 base_url: row.get(2)?,
                 api_key: row.get(3)?,
+                api_key_env: row.get(4)?,
                 env: env_json
                     .and_then(|s| serde_json::from_str(&s).ok())
                     .unwrap_or_default(),
-                note: row.get(5)?,
+                note: row.get(6)?,
-                official_url: row.get(6)?,
+                official_url: row.get(7)?,
-                managed: row.get::<_, i32>(7)? != 0,
+                managed: row.get::<_, i32>(8)? != 0,
-                enabled: row.get::<_, i32>(8)? != 0,
+                enabled: row.get::<_, i32>(9)? != 0,
-                active: row.get::<_, i32>(9)? != 0,
+                active: row.get::<_, i32>(10)? != 0,
-                icon: row.get(10)?,
+                icon: row.get(11)?,
-                icon_color: row.get(11)?,
+                icon_color: row.get(12)?,
                 raw_other: raw_json
                     .and_then(|s| serde_json::from_str(&s).ok())
                     .unwrap_or(Value::Null),
@@ -253,8 +258,8 @@ pub fn save_config(agent: &Agent, config: &Config) -> DbResult<()> {
     {
         let mut insert_provider = tx.prepare(
             "INSERT INTO providers
-             (agent, name, provider_type, base_url, api_key, env, note, official_url, managed, enabled, active, icon, icon_color, raw_other)
+             (agent, name, provider_type, base_url, api_key, api_key_env, env, note, official_url, managed, enabled, active, icon, icon_color, raw_other)
-             VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8, ?9, ?10, ?11, ?12, ?13, ?14)",
+             VALUES (?1, ?2, ?3, ?4, ?5, ?6, ?7, ?8, ?9, ?10, ?11, ?12, ?13, ?14, ?15)",
         )?;
 
         for provider in config.providers.values() {
@@ -264,6 +269,7 @@ pub fn save_config(agent: &Agent, config: &Config) -> DbResult<()> {
                 provider.provider_type.as_str(),
                 provider.base_url,
                 provider.api_key,
+                provider.api_key_env,
                 serde_json::to_string(&provider.env).ok(),
                 provider.note,
                 provider.official_url,

src-tauri/src/kimi_code_io.rs

@@ -153,6 +153,14 @@ pub fn kimi_code_to_config(value: &TomlValue) -> Config {
 
             let base_url = table.get("base_url").and_then(|v| v.as_str()).map(|s| s.to_string());
             let api_key = table.get("api_key").and_then(|v| v.as_str()).map(|s| s.to_string());
+            // kimi-code 2.0.0+: `api_key_env` names the environment variable
+            // holding the credential; it is mutually exclusive with `api_key`.
+            // Captured as its own field (not left in raw_other) so export can
+            // enforce the exclusivity and the UI can edit it.
+            let api_key_env = table
+                .get("api_key_env")
+                .and_then(|v| v.as_str())
+                .map(|s| s.to_string());
             let managed = table.contains_key("oauth")
                 || table.get("managed").and_then(|v| v.as_bool()).unwrap_or(false);
             let enabled = table.get("enabled").and_then(|v| v.as_bool()).unwrap_or(true);
@@ -181,6 +189,7 @@ pub fn kimi_code_to_config(value: &TomlValue) -> Config {
                 rest.remove("type");
                 rest.remove("base_url");
                 rest.remove("api_key");
+                rest.remove("api_key_env");
                 rest.remove("managed");
                 rest.remove("enabled");
                 rest.remove("env");
@@ -200,6 +209,7 @@ pub fn kimi_code_to_config(value: &TomlValue) -> Config {
                     provider_type,
                     base_url: base_url.filter(|s| !s.is_empty()),
                     api_key: api_key.filter(|s| !s.is_empty()),
+                    api_key_env: api_key_env.filter(|s| !s.is_empty()),
                     env,
                     note: None,
                     official_url: None,
@@ -289,11 +299,30 @@ pub fn config_to_kimi_code(config: &Config, existing: Option<&TomlValue>) -> Tom
         if let Some(base_url) = provider.base_url.clone().filter(|s| !s.is_empty()) {
             pt.insert("base_url".to_string(), TomlValue::String(base_url));
         }
+        // `api_key_env` (kimi-code 2.0.0+) and `api_key` are mutually
+        // exclusive upstream: emitting both makes the CLI reject the provider
+        // block, so the env var name alone is written whenever the provider is
+        // configured that way. A legacy row that still carries the key in
+        // raw_other (imports before the field existed) is used as the source
+        // too, so the setting survives the upgrade.
+        let api_key_env = provider
+            .api_key_env
+            .clone()
+            .or_else(|| {
+                provider
+                    .raw_other
+                    .get("api_key_env")
+                    .and_then(|v| v.as_str())
+                    .map(|s| s.to_string())
+            })
+            .filter(|s| !s.is_empty());
         // Managed (OAuth) providers persist an empty api_key line so the block
         // matches the official CLI's provisioned shape (apiKey: ''). A
         // user-set api_key on a managed provider (api_key outranks OAuth in
         // the CLI's credential priority) must survive the round-trip.
-        if provider.managed
+        if let Some(env_key) = api_key_env {
+            pt.insert("api_key_env".to_string(), TomlValue::String(env_key));
+        } else if provider.managed
             && provider.api_key.as_deref().map_or(true, |s| s.is_empty())
         {
             pt.insert("api_key".to_string(), TomlValue::String("".to_string()));
@@ -328,10 +357,12 @@ pub fn config_to_kimi_code(config: &Config, existing: Option<&TomlValue>) -> Tom
                 });
             pt.insert("oauth".to_string(), oauth);
         }
-        // Merge remaining raw fields (oauth and env are handled explicitly above).
+        // Merge remaining raw fields (oauth, env and api_key_env are handled
+        // explicitly above).
         if let TomlValue::Table(mut extra) = json_to_toml(&provider.raw_other).unwrap_or(TomlValue::Table(Table::new())) {
             extra.remove("oauth");
             extra.remove("env");
+            extra.remove("api_key_env");
             // Strip Kimi-Switch-private field: the remembered per-provider
             // default model is stored in raw_other.default_model and must NOT
             // leak into the agent's config.toml.
@@ -562,6 +593,7 @@ api_key = ""
                 provider_type: ProviderType::Anthropic,
                 base_url: Some("https://fast.cdks.work".to_string()),
                 api_key: Some("sk-test".to_string()),
+                api_key_env: None,
                 env: IndexMap::new(),
                 note: None,
                 official_url: None,
@@ -623,6 +655,7 @@ api_key = ""
             provider_type: ProviderType::Openai,
             base_url: Some(base_url.to_string()),
             api_key: Some("sk-test".to_string()),
+            api_key_env: None,
             env: IndexMap::new(),
             note: None,
             official_url: None,
@@ -760,6 +793,7 @@ default_effort = "low"
                 provider_type: ProviderType::Anthropic,
                 base_url: Some("https://example.com".to_string()),
                 api_key: Some("sk-test".to_string()),
+                api_key_env: None,
                 env: IndexMap::new(),
                 note: None,
                 official_url: None,
@@ -868,6 +902,7 @@ max_context_size = 1048576
                 provider_type: ProviderType::Anthropic,
                 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,
@@ -888,6 +923,7 @@ max_context_size = 1048576
                 provider_type: ProviderType::Openai,
                 base_url: Some("https://b.example.com".to_string()),
                 api_key: Some("sk-b".to_string()),
+                api_key_env: None,
                 env: IndexMap::new(),
                 note: None,
                 official_url: None,
@@ -929,6 +965,7 @@ max_context_size = 1048576
                 provider_type: ProviderType::Anthropic,
                 base_url: None,
                 api_key: Some("sk-x".to_string()),
+                api_key_env: None,
                 env: IndexMap::new(),
                 note: None,
                 official_url: None,
@@ -974,6 +1011,7 @@ max_context_size = 1048576
             provider_type: ProviderType::Kimi,
             base_url: Some("https://api.kimi.com/coding/v1".to_string()),
             api_key: api_key.map(String::from),
+            api_key_env: None,
             env: IndexMap::new(),
             note: None,
             official_url: None,
@@ -1153,4 +1191,112 @@ enabled = true
             "user-removed section must be dropped on export"
         );
     }
+
+    #[test]
+    fn kimi_code_import_extracts_api_key_env() {
+        // kimi-code 2.0.0+: `api_key_env` names the environment variable that
+        // holds the credential. It must become its own Provider field and stop
+        // living in raw_other, so export can enforce the api_key exclusivity.
+        let toml_str = r#"
+default_model = "p1/m1"
+
+[providers.p1]
+type = "openai"
+base_url = "https://proxy.example.com/v1"
+api_key_env = "MY_PROXY_KEY"
+
+[models."p1/m1"]
+provider = "p1"
+model = "m1"
+max_context_size = 128000
+"#;
+        let value: TomlValue = toml_str.parse().unwrap();
+        let config = kimi_code_to_config(&value);
+
+        let provider = config.providers.get("p1").unwrap();
+        assert_eq!(provider.api_key_env.as_deref(), Some("MY_PROXY_KEY"));
+        assert_eq!(provider.api_key, None);
+        assert!(
+            provider.raw_other.get("api_key_env").is_none(),
+            "api_key_env must not stay in raw_other"
+        );
+
+        let exported = config_to_kimi_code(&config, Some(&value));
+        let exported_provider = exported
+            .as_table().unwrap()
+            .get("providers").unwrap()
+            .as_table().unwrap()
+            .get("p1").unwrap()
+            .as_table().unwrap();
+        assert_eq!(
+            exported_provider.get("api_key_env").and_then(|v| v.as_str()),
+            Some("MY_PROXY_KEY")
+        );
+        assert!(
+            !exported_provider.contains_key("api_key"),
+            "api_key and api_key_env are mutually exclusive upstream"
+        );
+    }
+
+    #[test]
+    fn kimi_code_export_api_key_env_wins_over_stored_api_key() {
+        // A provider switched to env-var credentials in the UI must export
+        // ONLY `api_key_env`: writing both keys makes the CLI reject the block.
+        // A legacy raw_other copy of the key acts as the source when the field
+        // is empty (rows saved before the field existed).
+        let make_provider = |api_key: Option<&str>, api_key_env: Option<&str>, raw: Value| Provider {
+            name: "p".to_string(),
+            provider_type: ProviderType::Openai,
+            base_url: Some("https://proxy.example.com/v1".to_string()),
+            api_key: api_key.map(String::from),
+            api_key_env: api_key_env.map(String::from),
+            env: IndexMap::new(),
+            note: None,
+            official_url: None,
+            managed: false,
+            enabled: true,
+            active: false,
+            icon: None,
+            icon_color: None,
+            raw_other: raw,
+            usage_kinds: None,
+            usage_config: None,
+        };
+        let export = |p: Provider| {
+            let config = Config {
+                default_model: None,
+                providers: IndexMap::from([("p".to_string(), p)]),
+                models: IndexMap::new(),
+                raw_other: Value::Null,
+                imported_section_keys: Vec::new(),
+            };
+            config_to_kimi_code(&config, None)
+                .as_table().unwrap()
+                .get("providers").unwrap()
+                .as_table().unwrap()
+                .get("p").unwrap()
+                .as_table()
+                .unwrap()
+                .clone()
+        };
+
+        // Both fields set (a DB row saved before the UI cleared the key).
+        let pt = export(make_provider(Some("sk-stale"), Some("MY_PROXY_KEY"), Value::Null));
+        assert_eq!(pt.get("api_key_env").and_then(|v| v.as_str()), Some("MY_PROXY_KEY"));
+        assert!(!pt.contains_key("api_key"), "the stored key must be dropped");
+
+        // Legacy row: only raw_other carries the key.
+        let pt = export(make_provider(
+            Some("sk-stale"),
+            None,
+            serde_json::json!({"api_key_env": "LEGACY_KEY"}),
+        ));
+        assert_eq!(pt.get("api_key_env").and_then(|v| v.as_str()), Some("LEGACY_KEY"));
+        assert!(!pt.contains_key("api_key"));
+
+        // No env reference → the stored key is written as before.
+        let pt = export(make_provider(Some("sk-keep"), None, Value::Null));
+        assert_eq!(pt.get("api_key").and_then(|v| v.as_str()), Some("sk-keep"));
+        assert!(!pt.contains_key("api_key_env"));
+    }
 }

src-tauri/src/lib.rs

@@ -103,10 +103,14 @@ pub fn run() {
             commands::list_provider_models,
             commands::test_connectivity,
             commands::query_provider_usage,
+            commands::health_check_all,
             commands::debug_log,
             commands::get_app_setting,
             commands::set_app_setting,
             commands::check_for_update,
+            commands::get_kimi_code_version,
+            commands::get_kimi_code_latest,
+            commands::get_kimi_code_version_info,
             commands::download_update,
             commands::open_installer,
             commands::open_external_url,

src-tauri/src/models.rs

@@ -107,6 +107,12 @@ pub struct Provider {
     pub provider_type: ProviderType,
     pub base_url: Option<String>,
     pub api_key: Option<String>,
+    /// kimi-code 2.0.0+: name of the process environment variable holding the
+    /// credential, as an alternative to a stored `api_key` (the two are
+    /// mutually exclusive upstream — exporting both makes the CLI reject the
+    /// provider block). Only the variable *name* is stored, never its value.
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub api_key_env: Option<String>,
     #[serde(default)]
     pub env: IndexMap<String, String>,
     #[serde(default, skip_serializing_if = "Option::is_none")]
@@ -149,6 +155,7 @@ impl PartialEq for Provider {
             && self.provider_type == other.provider_type
             && self.base_url == other.base_url
             && self.api_key == other.api_key
+            && self.api_key_env == other.api_key_env
             && self.env == other.env
             && self.note == other.note
             && self.official_url == other.official_url
@@ -172,6 +179,7 @@ impl std::fmt::Debug for Provider {
             .field("provider_type", &self.provider_type)
             .field("base_url", &self.base_url)
             .field("api_key", &"<redacted>")
+            .field("api_key_env", &self.api_key_env)
             .field("env", &"<redacted>")
             .field("managed", &self.managed)
             .field("enabled", &self.enabled)
@@ -285,6 +293,11 @@ pub struct UsageConfig {
     /// Per-request timeout in seconds; 0/None = default (8s).
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub timeout_seconds: Option<u64>,
+    /// 用量预警阈值（百分比，0-100）；None/0 = 关闭。判定在前端
+    /// src/hooks/useUsageQuery.ts，Rust 只负责透传——缺了此字段，serde 会在
+    /// 保存时把前端写下的 threshold 静默丢掉。
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub threshold: Option<u32>,
 }
 
 impl UsageConfig {

src-tauri/src/oauth.rs

@@ -659,6 +659,7 @@ fn provision_managed_provider(region: KimiRegion, oauth_key: &str) -> Result<(),
             provider_type: crate::models::ProviderType::Kimi,
             base_url: None,
             api_key: None,
+            api_key_env: None,
             env: IndexMap::new(),
             note: None,
             official_url: None,

src-tauri/src/pi_io.rs

@@ -332,6 +332,9 @@ pub fn pi_file_to_config(file: &PiModelsFile) -> Config {
             provider_type,
             base_url: pi_provider.base_url.clone().filter(|s| !s.is_empty()),
             api_key: pi_provider.api_key.clone().filter(|s| !s.is_empty()),
+            // Pi's provider file has no env-var credential reference; the
+            // field stays empty so a Kimi-Code-only setting never leaks here.
+            api_key_env: None,
             env: IndexMap::new(),
             note: None,
             official_url: None,

src-tauri/src/services/coding_plan.rs

@@ -9,6 +9,9 @@
 //! "weekly_limit" / "monthly_limit"），`used` = 已用百分比（0-100），`total` = 100，
 //! `remaining` = 剩余百分比，`resets_at` 为 ISO 8601 字符串。
 //!
+//! 例外：Kimi 的加力钱包（`PLAN_BOOSTER_WALLET`）不是百分比窗口而是金额行，
+//! `used` / `total` / `remaining` 均为货币金额（分换算而来），`unit` 为币种。
+//!
 //! 错误通道语义与 balance.rs 一致（Err = 瞬时，Ok(success:false) = 确定性）。
 
 use super::balance::{get_json, get_json_with_ua, AuthStyle, Fetched};
@@ -16,11 +19,17 @@ use super::usage_types::{UsageData, UsageResult};
 use std::time::Duration;
 
 // 套餐类 tier id 的唯一来源：所有套餐供应商（Kimi/智谱/MiniMax/OpenCode Go 及
-// 未来新增）都只用这三个 id。前端 src/lib/usage-display.ts 的 planLabel() 依赖
+// 未来新增）都只用这四个 id。前端 src/lib/usage-display.ts 的 planLabel() 依赖
 // 此约定做本地化映射——新增 tier id 时必须同步加映射。
 const TIER_FIVE_HOUR: &str = "five_hour";
 const TIER_WEEKLY_LIMIT: &str = "weekly_limit";
 const TIER_MONTHLY_LIMIT: &str = "monthly_limit";
+/// Kimi quota 模型的月度代码窗口（kimi-code #3787 后 `limit_month_code`）。
+const TIER_MONTH_CODE: &str = "month_code";
+/// Kimi quota 响应里的加力钱包（boosterWallet）。**不是百分比 tier**：这条是
+/// 金额行，前端 src/lib/usage-display.ts 的 planLabel() 负责本地化，
+/// UsageFooter 按「金额行」而非配额行渲染（used/total 是钱数，不是配额比例）。
+const PLAN_BOOSTER_WALLET: &str = "booster_wallet";
 
 /// 套餐条目的统一构造：按百分比表示用量。
 fn percent_tier(name: &str, used_percent: f64, resets_at: Option<String>) -> UsageData {
@@ -70,7 +79,13 @@ fn parse_f64(value: &serde_json::Value) -> Option<f64> {
 // GET {base_url}/usages
 //   默认 https://api.kimi.com/coding/v1/usages
 //   global: https://api.kimi.ai/coding/v1/usages
-// Response: { limits: [{ detail: { limit, remaining, resetTime } }],
+// Response（kimi-code #3787 / 0.43.1 起服务端切换为 quota 模型）:
+//   { usages: { limit_5h, limit_7d, limit_month_total, limit_month_code:
+//               { used_ratio: 0-1, reset_time?: ISO 8601 } },
+//     boosterWallet: { balanceCents, totalCents, monthlyChargeLimitEnabled,
+//                      monthlyChargeLimitCents, monthlyUsedCents, currency },
+//     goods_version }
+// 旧结构（兼容保留）: { limits: [{ detail: { limit, remaining, resetTime } }],
 //                     usage: { limit, remaining, resetTime } }
 
 /// 由 base_url 拼接 usages 查询 URL；base_url 为空/空白时回退大陆默认。
@@ -95,8 +110,9 @@ pub async fn query_kimi_coding(
         Fetched::Body(body) => {
             let tiers = parse_kimi_coding(&body);
             if tiers.is_empty() {
-                // 响应里没有可解析的套餐档位（limits/usage 缺失或字段变了）。
+                // 响应里没有可解析的套餐档位（quota/usages 与旧 limits/usage
-                // 把原始响应（仅用量数字，无密钥）透出，方便对照接口结构修复。
+                // 均缺失或字段变了）。把原始响应（仅用量数字，无密钥）透出，
+                // 方便对照接口结构修复。
                 let preview = serde_json::to_string(&body)
                     .unwrap_or_else(|_| "<unserializable body>".into());
                 let trimmed: String = preview.chars().take(400).collect();
@@ -111,6 +127,11 @@ pub async fn query_kimi_coding(
 }
 
 fn parse_kimi_coding(body: &serde_json::Value) -> Vec<UsageData> {
+    // 新 quota 模型（kimi-code #3787 / 0.43.1 起服务端下发）：优先解析。
+    if let Some(tiers) = parse_kimi_quota(body) {
+        return tiers;
+    }
+
     let mut tiers = Vec::new();
 
     // 5 小时窗口限额（优先显示）
@@ -130,6 +151,67 @@ fn parse_kimi_coding(body: &serde_json::Value) -> Vec<UsageData> {
     tiers
 }
 
+/// 解析 quota 模型的 `usages` 窗口映射（`limit_5h` / `limit_7d` /
+/// `limit_month_total` / `limit_month_code`）+ 加力钱包。窗口与钱包都缺失或
+/// 无可解析条目时返回 None，让调用方回退旧结构。
+fn parse_kimi_quota(body: &serde_json::Value) -> Option<Vec<UsageData>> {
+    let usages = body.get("usages")?;
+    let mut tiers = Vec::new();
+    for (key, tier) in [
+        ("limit_5h", TIER_FIVE_HOUR),
+        ("limit_7d", TIER_WEEKLY_LIMIT),
+        ("limit_month_total", TIER_MONTHLY_LIMIT),
+        ("limit_month_code", TIER_MONTH_CODE),
+    ] {
+        let Some(entry) = usages.get(key) else { continue };
+        let Some(ratio) = entry.get("used_ratio").and_then(parse_f64) else {
+            continue;
+        };
+        // used_ratio 规范为 0-1 比率；>1 时视为已是百分数，原样使用。
+        let used_percent = if ratio <= 1.0 { ratio * 100.0 } else { ratio };
+        let resets_at = entry.get("reset_time").and_then(extract_reset_time);
+        tiers.push(percent_tier(tier, used_percent, resets_at));
+    }
+    // 加力钱包排在四个窗口之后（钱包是余额，不是滚动窗口）。
+    if let Some(wallet) = body.get("boosterWallet").and_then(booster_wallet_tier) {
+        tiers.push(wallet);
+    }
+    if tiers.is_empty() { None } else { Some(tiers) }
+}
+
+/// 解析加力钱包（boosterWallet）为金额行。字段单位为「分」，统一换算成元/美元。
+///
+/// 只在响应含新版 `*Cents` 字段时产出条目：kimi-code #3787 之前的
+/// `boosterWallet` 只有 `balance`（无币种/精度语义），产出全 0 的伪行只会误导。
+fn booster_wallet_tier(wallet: &serde_json::Value) -> Option<UsageData> {
+    let cents = |key: &str| wallet.get(key).and_then(parse_f64);
+    let (balance, total, used) = (
+        cents("balanceCents"),
+        cents("totalCents"),
+        cents("monthlyUsedCents"),
+    );
+    if balance.is_none() && total.is_none() && used.is_none() {
+        return None;
+    }
+    // 币种缺失时不猜（前端 formatAmount 会退化为纯数字展示）。
+    let unit = wallet
+        .get("currency")
+        .and_then(|v| v.as_str())
+        .map(str::trim)
+        .filter(|s| !s.is_empty())
+        .map(str::to_string);
+    Some(UsageData {
+        plan_name: Some(PLAN_BOOSTER_WALLET.to_string()),
+        remaining: balance.map(|c| c / 100.0),
+        total: total.map(|c| c / 100.0),
+        used: used.map(|c| c / 100.0),
+        unit,
+        is_valid: Some(true),
+        // 钱包没有窗口重置时间。
+        resets_at: None,
+    })
+}
+
 fn kimi_limit_tier(name: &str, detail: &serde_json::Value) -> UsageData {
     let limit = detail.get("limit").and_then(parse_f64).unwrap_or(1.0);
     let remaining = detail.get("remaining").and_then(parse_f64).unwrap_or(0.0);
@@ -459,6 +541,97 @@ mod tests {
         );
     }
 
+    #[test]
+    fn kimi_coding_quota_model_four_windows() {
+        // kimi-code #3787 / 0.43.1 起服务端下发的 quota 模型：used_ratio 为
+        // 0-1 比率，reset_time 为 ISO 8601 字符串。
+        let body = json!({
+            "goods_version": 3,
+            "usages": {
+                "limit_5h": { "used_ratio": 0.6, "reset_time": "2026-09-15T17:00:00Z" },
+                "limit_7d": { "used_ratio": 0.25 },
+                "limit_month_total": { "used_ratio": 0.1, "reset_time": "2026-10-01T00:00:00Z" },
+                "limit_month_code": { "used_ratio": 0.88 }
+            },
+            "boosterWallet": { "balance": 0 }
+        });
+        let tiers = parse_kimi_coding(&body);
+        // 旧版 boosterWallet（只有 balance，无 *Cents）不产出钱包行。
+        assert_eq!(tiers.len(), 4);
+        assert_eq!(tiers[0].plan_name.as_deref(), Some("five_hour"));
+        assert_eq!(tiers[0].used, Some(60.0));
+        assert_eq!(tiers[0].remaining, Some(40.0));
+        assert_eq!(tiers[0].total, Some(100.0));
+        assert_eq!(
+            tiers[0].resets_at.as_deref(),
+            Some("2026-09-15T17:00:00Z")
+        );
+        assert_eq!(tiers[1].plan_name.as_deref(), Some("weekly_limit"));
+        assert_eq!(tiers[1].used, Some(25.0));
+        assert!(tiers[1].resets_at.is_none());
+        assert_eq!(tiers[2].plan_name.as_deref(), Some("monthly_limit"));
+        assert_eq!(tiers[2].used, Some(10.0));
+        assert_eq!(tiers[3].plan_name.as_deref(), Some("month_code"));
+        assert_eq!(tiers[3].used, Some(88.0));
+        assert_eq!(tiers[3].remaining, Some(12.0));
+    }
+
+    #[test]
+    fn kimi_coding_quota_appends_booster_wallet_row() {
+        // 加力钱包是金额行（分 → 元/美元），追加在四个窗口之后。
+        let body = json!({
+            "usages": {
+                "limit_5h": { "used_ratio": 0.6 },
+                "limit_7d": { "used_ratio": 0.25 },
+                "limit_month_total": { "used_ratio": 0.1 },
+                "limit_month_code": { "used_ratio": 0.88 }
+            },
+            "boosterWallet": {
+                "balanceCents": 1234,
+                "totalCents": 5000,
+                "monthlyChargeLimitEnabled": true,
+                "monthlyChargeLimitCents": 2000,
+                "monthlyUsedCents": 800,
+                "currency": "USD"
+            }
+        });
+        let tiers = parse_kimi_coding(&body);
+        assert_eq!(tiers.len(), 5);
+        assert_eq!(tiers[0].plan_name.as_deref(), Some("five_hour"));
+        let wallet = &tiers[4];
+        assert_eq!(wallet.plan_name.as_deref(), Some("booster_wallet"));
+        assert_eq!(wallet.used, Some(8.0));
+        assert_eq!(wallet.total, Some(50.0));
+        assert_eq!(wallet.remaining, Some(12.34));
+        assert_eq!(wallet.unit.as_deref(), Some("USD"));
+        assert_eq!(wallet.is_valid, Some(true));
+        assert!(wallet.resets_at.is_none());
+    }
+
+    #[test]
+    fn kimi_coding_quota_ratio_above_one_is_percent() {
+        // 防御：>1 的 used_ratio 视为已是百分数，不再乘 100。
+        let body = json!({
+            "usages": { "limit_5h": { "used_ratio": 42.0 } }
+        });
+        let tiers = parse_kimi_coding(&body);
+        assert_eq!(tiers.len(), 1);
+        assert_eq!(tiers[0].used, Some(42.0));
+    }
+
+    #[test]
+    fn kimi_coding_quota_missing_windows_falls_back_to_legacy() {
+        // usages 存在但没有任何可解析条目 → 回退旧结构解析。
+        let body = json!({
+            "usages": { "limit_5h": {} },
+            "usage": { "limit": 1000, "remaining": 900, "resetTime": "2026-08-01T00:00:00Z" }
+        });
+        let tiers = parse_kimi_coding(&body);
+        assert_eq!(tiers.len(), 1);
+        assert_eq!(tiers[0].plan_name.as_deref(), Some("weekly_limit"));
+        assert_eq!(tiers[0].used, Some(10.0));
+    }
+
     #[test]
     fn kimi_coding_reset_time_seconds_vs_millis() {
         // 秒级时间戳自动 ×1000

src-tauri/src/services/mod.rs

@@ -10,6 +10,7 @@
 
 pub mod balance;
 pub mod coding_plan;
+pub mod probe;
 pub mod usage_types;
 
 pub use usage_types::{UsageData, UsageResult};
@@ -336,6 +337,7 @@ mod tests {
             user_id: uid.map(str::to_string),
             auto_query_interval_minutes: None,
             timeout_seconds: None,
+            threshold: None,
         };
         // 缺 config / 缺任一项 / 空字符串 → None（确定性配置错误）
         assert_eq!(newapi_creds(None), None);

src-tauri/src/services/probe.rs

@@ -0,0 +1,155 @@
+//! 供应商推理端点探活（「一键体检」的第二步）。
+//!
+//! GET `{base_url}/v1/models`，带 `Authorization: Bearer <api_key>`。
+//! 2xx 且响应体是合法 JSON 即算通过（不校验内容结构，只验证端点能应答）；
+//! 其余状态码 / 网络错误归类成确定性错误文案。
+//!
+//! 错误文案与 balance.rs 一致用英文短句，前端 localizeHealthError 负责
+//! 双语展示（见 src/hooks/useHealthCheck.ts）。
+
+use std::time::Duration;
+
+/// 探活请求超时。体检要快速出结果、不卡住 UI，比账单查询的 8s 更短。
+pub const PROBE_TIMEOUT: Duration = Duration::from_secs(5);
+
+/// 由 base_url 拼出 `/v1/models` 探测地址。
+///
+/// 供应商配置里的 base_url 多数已经带 `/v1` 后缀（如
+/// `https://api.kimi.com/coding/v1`），直接再拼 `/v1/models` 会得到
+/// `/v1/v1/models`，所以先剥掉重复后缀；结尾斜杠与首尾空白一并归一。
+/// 空/纯空白返回 None（调用方直接判失败，不发请求）。纯函数，便于单测。
+fn models_url(base_url: &str) -> Option<String> {
+    let base = base_url.trim().trim_end_matches('/');
+    if base.is_empty() {
+        return None;
+    }
+    let root = base.strip_suffix("/v1").unwrap_or(base);
+    Some(format!("{root}/v1/models"))
+}
+
+/// 探活前置校验：返回最终请求 URL。base_url 缺失、api_key 为空都是确定性
+/// 配置错误，不发请求即失败（没有 key 时必然 401，直接给出更可操作的结论）。
+fn validate_probe_input(base_url: &str, api_key: &str) -> Result<String, String> {
+    let Some(url) = models_url(base_url) else {
+        return Err("missing base_url".to_string());
+    };
+    if api_key.trim().is_empty() {
+        return Err("no API key configured".to_string());
+    }
+    Ok(url)
+}
+
+/// 探活一个供应商的推理端点。`Ok(())` = 端点正常应答且返回合法 JSON。
+pub async fn probe_provider(
+    base_url: &str,
+    api_key: &str,
+    timeout: Duration,
+) -> Result<(), String> {
+    let url = validate_probe_input(base_url, api_key)?;
+
+    let client = reqwest::Client::builder()
+        .timeout(timeout)
+        .build()
+        .map_err(|e| format!("failed to build HTTP client: {e}"))?;
+
+    // 注意：api_key 只允许进请求头，严禁拼进 URL / 日志 / 错误信息。
+    let resp = client
+        .get(&url)
+        .header("Accept", "application/json")
+        .header("Authorization", format!("Bearer {api_key}"))
+        .send()
+        .await
+        .map_err(|e| {
+            if e.is_timeout() {
+                "request timed out".to_string()
+            } else if e.is_connect() {
+                "connection failed (DNS or refused)".to_string()
+            } else {
+                format!("network error: {e}")
+            }
+        })?;
+
+    let status = resp.status();
+    if status.is_success() {
+        // 先取完整响应体再解析：读体超时/中断是瞬时错误，与「返回的不是
+        // JSON」区分开（reqwest 的 .json() 会把两者都包成 decode）。
+        let raw = resp
+            .bytes()
+            .await
+            .map_err(|e| format!("failed to read response: {e}"))?;
+        return match serde_json::from_slice::<serde_json::Value>(&raw) {
+            Ok(_) => Ok(()),
+            Err(_) => Err("response is not valid JSON".to_string()),
+        };
+    }
+
+    match status.as_u16() {
+        401 | 403 => Err(format!("Authentication failed (HTTP {status})")),
+        404 => Err(format!("endpoint not found (HTTP {status})")),
+        code => Err(format!("API error (HTTP {code})")),
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn models_url_dedupes_v1_suffix() {
+        assert_eq!(
+            models_url("https://api.kimi.com/coding/v1").as_deref(),
+            Some("https://api.kimi.com/coding/v1/models")
+        );
+        // 不带 /v1 的裸域补上 /v1
+        assert_eq!(
+            models_url("https://api.deepseek.com").as_deref(),
+            Some("https://api.deepseek.com/v1/models")
+        );
+    }
+
+    #[test]
+    fn models_url_normalizes_slashes_and_whitespace() {
+        // 结尾斜杠（含多个）与首尾空白都要归一，且不得拼出 /v1/v1
+        assert_eq!(
+            models_url("https://api.kimi.com/coding/v1/").as_deref(),
+            Some("https://api.kimi.com/coding/v1/models")
+        );
+        assert_eq!(
+            models_url("  https://api.deepseek.com/v1///  ").as_deref(),
+            Some("https://api.deepseek.com/v1/models")
+        );
+        assert_eq!(
+            models_url("https://api.deepseek.com/").as_deref(),
+            Some("https://api.deepseek.com/v1/models")
+        );
+        // 非 /v1 的路径后缀原样保留（只有 /v1 需要去重）
+        assert_eq!(
+            models_url("https://api.novita.ai/v3").as_deref(),
+            Some("https://api.novita.ai/v3/v1/models")
+        );
+    }
+
+    #[test]
+    fn models_url_rejects_empty_base() {
+        assert_eq!(models_url(""), None);
+        assert_eq!(models_url("   "), None);
+        assert_eq!(models_url("/"), None);
+    }
+
+    #[test]
+    fn probe_input_validation_rejects_empty_base_and_missing_key() {
+        // 两者都是确定性配置错误，不发请求即返回。
+        assert_eq!(
+            validate_probe_input("  ", "sk-x").unwrap_err(),
+            "missing base_url"
+        );
+        assert_eq!(
+            validate_probe_input("https://example.com", "   ").unwrap_err(),
+            "no API key configured"
+        );
+        assert_eq!(
+            validate_probe_input("https://example.com/v1", "sk-x").unwrap(),
+            "https://example.com/v1/models"
+        );
+    }
+}

src-tauri/tauri.conf.json

@@ -1,6 +1,6 @@
 {
   "productName": "Kimi Switch",
-  "version": "0.7.18",
+  "version": "0.7.20",
   "identifier": "com.kimiswitch.app",
   "build": {
     "beforeDevCommand": "npm run dev",

src/components/AgentSettingsPanel.tsx

@@ -78,6 +78,10 @@ export function AgentSettingsPanel({ rawOther, onChange }: AgentSettingsPanelPro
     update({ background: { ...settings.background, ...patch } });
   };
 
+  const updateWatch = (patch: Partial<AgentSettings["watch"]>) => {
+    update({ watch: { ...settings.watch, ...patch } });
+  };
+
   const setRules = (rules: PermissionRule[]) => {
     update({ permission: { rules } });
   };
@@ -141,6 +145,15 @@ export function AgentSettingsPanel({ rawOther, onChange }: AgentSettingsPanelPro
           value={settings.loop_control?.reserved_context_size ?? 50000}
           onChange={(v) => updateLoopControl({ reserved_context_size: v })}
         />
+        {/* kimi-code 0.43.0+ `loop_control.compaction_max_attempts`. The
+            upstream default is 5; an empty field or 0 removes the key instead
+            of writing a literal 0 (see setAgentSettings). */}
+        <NumberField
+          label={t("compactionMaxAttempts")}
+          value={settings.loop_control?.compaction_max_attempts ?? 5}
+          onChange={(v) => updateLoopControl({ compaction_max_attempts: v })}
+        />
+        <p className="text-xs text-content-muted">{t("compactionMaxAttemptsHint")}</p>
       </Card>
 
       <Card title={t("backgroundSettings")}>
@@ -156,6 +169,20 @@ 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. */}
+        <Checkbox
+          label={t("watchEnabled")}
+          checked={settings.watch?.enabled ?? false}
+          onChange={(checked) => updateWatch({ enabled: checked })}
+        />
+        <p className="text-xs text-content-muted">{t("watchEnabledDesc")}</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

@@ -11,6 +11,7 @@ import { AgentSettingsPanel } from "./AgentSettingsPanel";
 import { KimiOAuthDialog } from "./KimiOAuthDialog";
 import { ProviderIcon } from "./ProviderIcon";
 import { IconPicker } from "./IconPicker";
+import { Segmented } from "./ui/controls";
 import type { Agent, DiscoveredModel, Model, Provider, ProviderType } from "../types";
 
 const KNOWN_CAPABILITIES = [
@@ -127,6 +128,19 @@ export function ProviderEdit({
   // Preset this provider was created from (if any) — provides the
   // "Get API Key" / referral links shown under the key input.
   const preset = findPresetForProvider(provider);
+  /**
+   * Credential source. `api_key_env` (kimi-code 2.0.0+) names an environment
+   * variable instead of storing the key in config.toml; the two are mutually
+   * exclusive upstream, so the UI is a mode switch and each mode clears the
+   * other field. Derived from the provider itself (no local state) so the
+   * switch follows provider switches and reloads; an empty-but-present value
+   * ("env mode picked, name not typed yet") still reads as env mode.
+   * Only Kimi Code understands `api_key_env` — Pi's provider file has no
+   * equivalent, so the switch and the env input are both gated on the agent
+   * (a Pi provider that somehow carries the key keeps its plain key input).
+   */
+  const apiKeyEnvSupported = agent === "kimi_code";
+  const apiKeyEnvMode = apiKeyEnvSupported && provider.api_key_env != null;
 
   useEffect(() => {
     const def = defaultBaseUrl(agent, provider.provider_type);
@@ -322,9 +336,49 @@ export function ProviderEdit({
 
                 {!provider.managed && (
                   <div>
+                    {apiKeyEnvSupported && (
+                      <div className="flex items-center gap-3 mb-2">
+                        <span className="text-sm text-content-muted">
+                          {t("apiKeySource")}
+                        </span>
+                        <Segmented
+                          options={[
+                            { key: "direct", label: t("apiKeySourceDirect") },
+                            { key: "env", label: t("apiKeySourceEnv") },
+                          ]}
+                          value={apiKeyEnvMode ? "env" : "direct"}
+                          onChange={(mode) => {
+                            // Mutual exclusion: picking a source clears the
+                            // other one so export never writes both keys.
+                            if (mode === "env") {
+                              onChange({
+                                ...provider,
+                                api_key: null,
+                                api_key_env: provider.api_key_env ?? "",
+                              });
+                            } else {
+                              onChange({ ...provider, api_key_env: null });
+                            }
+                          }}
+                        />
+                      </div>
+                    )}
                     <label htmlFor={apiKeyId} className="block text-sm text-content-muted mb-1.5">
-                      {t("apiKey")} <span className="text-red-500">*</span>
+                      {apiKeyEnvMode ? t("apiKeyEnvName") : t("apiKey")}{" "}
+                      <span className="text-red-500">*</span>
                     </label>
+                    {apiKeyEnvMode ? (
+                      <input
+                        id={apiKeyId}
+                        type="text"
+                        className="w-full bg-input border border-border rounded-lg px-3 py-2 text-sm font-mono focus:ring-2 focus:ring-blue-500 focus:outline-none"
+                        value={provider.api_key_env ?? ""}
+                        placeholder="KIMI_API_KEY"
+                        onChange={(e) =>
+                          onChange({ ...provider, api_key_env: e.target.value })
+                        }
+                      />
+                    ) : (
                       <div className="relative">
                         <input
                           id={apiKeyId}
@@ -343,6 +397,12 @@ export function ProviderEdit({
                           {showApiKey ? t("hide") : t("show")}
                         </button>
                       </div>
+                    )}
+                    {apiKeyEnvMode && (
+                      <p className="mt-1.5 text-xs text-content-muted">
+                        {t("apiKeyEnvHint")}
+                      </p>
+                    )}
                     {(() => {
                       const linkUrl = preset?.apiKeyUrl ?? provider.official_url ?? null;
                       if (!linkUrl) return null;

src/components/ProviderList.tsx

@@ -1,10 +1,26 @@
 import { useState } from "react";
 import { invoke } from "@tauri-apps/api/core";
-import { Pencil, Copy, Activity, Loader2, Trash2, BarChart3 } from "lucide-react";
+import {
+  Pencil,
+  Copy,
+  Activity,
+  Loader2,
+  Trash2,
+  BarChart3,
+  Stethoscope,
+} from "lucide-react";
 import { useTranslation } from "../i18n";
 import { ProviderIcon } from "./ProviderIcon";
 import { UsageFooter } from "./UsageFooter";
 import { useUsageQuery } from "../hooks/useUsageQuery";
+import {
+  useHealthCheck,
+  localizeHealthError,
+  probeLine,
+  usageLine,
+  healthDotClass,
+  type HealthResult,
+} from "../hooks/useHealthCheck";
 import type { Agent, Model, Provider } from "../types";
 
 interface ConnectivityResult {
@@ -40,6 +56,12 @@ interface ProviderCardProps {
   models: Record<string, Model>;
   /** Inline connectivity-test state for this provider, or undefined when idle. */
   ts: TestState | undefined;
+  /** 一键体检结论；undefined = 本轮未出结果（检测中或未检测）。 */
+  health: HealthResult | undefined;
+  /** 体检请求进行中。 */
+  healthRunning: boolean;
+  /** 体检至少跑过一轮（决定是否渲染灰色「未检测」点）。 */
+  healthRan: boolean;
   onEdit: (name: string) => void;
   onDelete: (name: string) => void;
   onDuplicate: (name: string) => void;
@@ -54,6 +76,9 @@ function ProviderCard({
   defaultModel,
   models,
   ts,
+  health,
+  healthRunning,
+  healthRan,
   onEdit,
   onDelete,
   onDuplicate,
@@ -70,7 +95,8 @@ function ProviderCard({
     provider.name,
     provider.usageKinds,
     provider.usageConfig?.autoQueryIntervalMinutes,
-    provider.usageConfig?.enabled === false
+    provider.usageConfig?.enabled === false,
+    provider.usageConfig?.threshold
   );
   const providerModels = Object.values(models).filter(
     (m) => m.provider === provider.name
@@ -80,6 +106,17 @@ function ProviderCard({
     : null;
   const isActive = provider.active === true;
 
+  // 一键体检状态点 tooltip：两项检测各自如实展示——探活可能是「不支持模型
+  // 列表」（中性），账单可能是「失败: xxx」，都不掩盖。
+  const healthTip = health
+    ? [
+        health.probe ? probeLine(health.probe, health.latencyMs, t) : null,
+        health.usage ? usageLine(health.usage, t) : null,
+      ]
+        .filter((s): s is string => s !== null)
+        .join(" · ")
+    : "";
+
   return (
     <div
       className={`group relative flex flex-col gap-3 p-4 rounded-xl border transition-colors cursor-pointer w-full ${
@@ -102,6 +139,34 @@ function ProviderCard({
           <h3 className="font-semibold text-content-primary truncate">
             {provider.name}
           </h3>
+          {/* 一键体检状态点：绿=正常、红=失败、灰=未判定（未检测 / 检测中） */}
+          {health ? (
+            <span
+              role="img"
+              className={`w-2.5 h-2.5 shrink-0 rounded-full ${healthDotClass(
+                health.verdict
+              )}`}
+              title={healthTip}
+              aria-label={healthTip}
+            />
+          ) : healthRunning ? (
+            <span
+              className="shrink-0 flex items-center"
+              title={t("healthCheckRunning")}
+              aria-label={t("healthCheckRunning")}
+            >
+              <Loader2 className="w-3.5 h-3.5 animate-spin text-content-muted" />
+            </span>
+          ) : healthRan ? (
+            <span
+              role="img"
+              className={`w-2.5 h-2.5 shrink-0 rounded-full ${healthDotClass(
+                "neutral"
+              )}`}
+              title={t("healthNotChecked")}
+              aria-label={t("healthNotChecked")}
+            />
+          ) : null}
           {isActive && (
             <span className="text-xs px-2 py-0.5 rounded-full bg-green-100 dark:bg-green-900/30 text-green-600 dark:text-green-400 border border-green-300 dark:border-green-500/30">
               {t("inUse")}
@@ -273,6 +338,7 @@ export function ProviderList({
 }: ProviderListProps) {
   const { t } = useTranslation();
   const [testState, setTestState] = useState<Record<string, TestState>>({});
+  const health = useHealthCheck(agent);
 
   // managed（OAuth 登录托管）供应商排在最前；其余保持原有顺序（稳定排序）。
   const sortedProviders = [...providers].sort(
@@ -312,8 +378,48 @@ export function ProviderList({
 
   return (
     <div className="h-full flex flex-col">
-      <div className="flex items-center justify-between px-4 py-3 border-b border-border">
+      <div className="flex items-center justify-between gap-3 px-4 py-3 border-b border-border">
-        <h2 className="font-medium text-content-primary">{t("providers")}</h2>
+        <div className="flex items-center gap-3 min-w-0">
+          <h2 className="font-medium text-content-primary shrink-0">
+            {t("providers")}
+          </h2>
+          {/* 体检汇总：检测中 / 请求失败 / x 正常 / y 失败 */}
+          {(health.running || health.ran || health.error) && (
+            <span
+              className={`text-xs tabular-nums truncate ${
+                health.error ? "text-red-500 dark:text-red-400" : "text-content-muted"
+              }`}
+              title={health.error ? localizeHealthError(health.error, t) : undefined}
+            >
+              {health.error
+                ? t("healthRequestFailed", {
+                    error: localizeHealthError(health.error, t),
+                  })
+                : health.running
+                  ? t("healthCheckRunning")
+                  : t("healthSummary", {
+                      ok: health.okCount,
+                      fail: health.failCount,
+                    })}
+            </span>
+          )}
+        </div>
+        <div className="flex items-center gap-2 shrink-0">
+          <button
+            type="button"
+            onClick={health.run}
+            disabled={health.running}
+            title={t("healthCheckHint")}
+            aria-label={t("healthCheck")}
+            className="flex items-center gap-1.5 h-9 px-3 rounded-full border border-border text-sm text-content-muted hover:text-content-primary hover:bg-hover-2 disabled:opacity-50 focus:ring-2 focus:ring-blue-500 focus:outline-none transition-colors whitespace-nowrap"
+          >
+            {health.running ? (
+              <Loader2 className="w-4 h-4 animate-spin" />
+            ) : (
+              <Stethoscope className="w-4 h-4" />
+            )}
+            {t("healthCheck")}
+          </button>
           <button
             type="button"
             onClick={onAdd}
@@ -335,6 +441,7 @@ export function ProviderList({
             </svg>
           </button>
         </div>
+      </div>
 
       <div className="flex-1 overflow-auto p-4">
         {providers.length === 0 ? (
@@ -371,6 +478,9 @@ export function ProviderList({
                 defaultModel={defaultModel}
                 models={models}
                 ts={testState[provider.name]}
+                health={health.results[provider.name]}
+                healthRunning={health.running}
+                healthRan={health.ran}
                 onEdit={onEdit}
                 onDelete={onDelete}
                 onDuplicate={onDuplicate}

src/components/SettingsModal.tsx

@@ -5,12 +5,20 @@ import { listen } from "@tauri-apps/api/event";
 import { useTranslation, type Language } from "../i18n";
 import { useTheme, type Theme } from "../hooks/useTheme";
 import type { UpdateInfo } from "../hooks/useUpdateCheck";
+import {
+  useKimiCodeVersion,
+  isNewerVersion,
+  type KimiCodeStatus,
+} from "../hooks/useKimiCodeVersion";
 
 function formatVersion(version: string): string {
   const v = version.trim();
   return v.startsWith("v") || v.startsWith("V") ? v : `v${v}`;
 }
 
+/** 上游 kimi-code releases 页面。 */
+const KIMI_CODE_RELEASES_URL = "https://github.com/MoonshotAI/kimi-code/releases";
+
 interface SettingsModalProps {
   open: boolean;
   onClose: () => void;
@@ -30,6 +38,11 @@ export function SettingsModal({
 }: SettingsModalProps) {
   const { t, lang, setLang } = useTranslation();
   const { theme, setTheme } = useTheme();
+  const {
+    info: kimiCodeInfo,
+    checking: kimiCodeChecking,
+    checkNow: checkKimiCode,
+  } = useKimiCodeVersion();
   const [appVersion, setAppVersion] = useState("");
   const [downloadProgress, setDownloadProgress] = useState<number | null>(null);
   const [downloadedPath, setDownloadedPath] = useState<string | null>(null);
@@ -158,6 +171,47 @@ export function SettingsModal({
     return d.toLocaleString(lang === "zh" ? "zh-CN" : "en-US");
   };
 
+  // 状态徽标：绿=已适配 / 黄=建议升级 KimiSwitch / 橙=部分兼容 / 灰=未知或未安装
+  const kimiCodeBadge = (status: KimiCodeStatus) => {
+    switch (status) {
+      case "current":
+        return {
+          label: t("kimiCodeStatusCurrent"),
+          className:
+            "bg-green-100 dark:bg-green-900/30 text-green-600 dark:text-green-400 border-green-300 dark:border-green-500/30",
+        };
+      case "outdated":
+        return {
+          label: t("kimiCodeStatusOutdated"),
+          className:
+            "bg-yellow-100 dark:bg-yellow-900/30 text-yellow-700 dark:text-yellow-400 border-yellow-300 dark:border-yellow-500/30",
+        };
+      case "partial":
+        return {
+          label: t("kimiCodeStatusPartial"),
+          className:
+            "bg-orange-100 dark:bg-orange-900/30 text-orange-600 dark:text-orange-400 border-orange-300 dark:border-orange-500/30",
+        };
+      case "unknown":
+        return {
+          label: t("kimiCodeStatusUnknown"),
+          className: "bg-hover-2 text-content-muted border-border",
+        };
+      default:
+        return {
+          label: t("kimiCodeStatusNotInstalled"),
+          className: "bg-hover-2 text-content-muted border-border",
+        };
+    }
+  };
+
+  const kimiCodeUpdateAvailable =
+    !!kimiCodeInfo?.installed &&
+    !!kimiCodeInfo?.latest &&
+    isNewerVersion(kimiCodeInfo.latest, kimiCodeInfo.installed);
+
+  const kimiCodeBadgeInfo = kimiCodeInfo ? kimiCodeBadge(kimiCodeInfo.status) : null;
+
   return createPortal(
     <div
       className="fixed inset-0 z-[9999] flex items-center justify-center bg-black/60 backdrop-blur-sm p-4"
@@ -362,6 +416,72 @@ export function SettingsModal({
             </div>
           </section>
 
+          {/* Kimi Code CLI version & compatibility */}
+          <section>
+            <h3 className="mb-3 text-[11px] uppercase tracking-wider text-content-muted">
+              {t("kimiCodeSection")}
+            </h3>
+            <div className="space-y-2 rounded-lg border border-border bg-input p-3">
+              <div className="flex items-center justify-between text-sm">
+                <span className="text-content-muted">{t("kimiCodeInstalled")}</span>
+                <span className="font-medium text-content-primary tabular-nums">
+                  {kimiCodeInfo?.installed ? formatVersion(kimiCodeInfo.installed) : "—"}
+                </span>
+              </div>
+              <div className="flex items-center justify-between text-sm">
+                <span className="text-content-muted">{t("kimiCodeLatest")}</span>
+                <span className="font-medium text-content-primary tabular-nums">
+                  {kimiCodeInfo?.latest ? formatVersion(kimiCodeInfo.latest) : "—"}
+                </span>
+              </div>
+              <div className="flex items-center justify-between text-sm">
+                <span className="text-content-muted">{t("kimiCodeStatusLabel")}</span>
+                {kimiCodeBadgeInfo ? (
+                  <span
+                    className={`text-xs px-2 py-0.5 rounded-full border ${kimiCodeBadgeInfo.className}`}
+                  >
+                    {kimiCodeBadgeInfo.label}
+                  </span>
+                ) : (
+                  <span className="text-xs text-content-muted">—</span>
+                )}
+              </div>
+              <div className="pt-2 flex items-center gap-2">
+                <button
+                  type="button"
+                  onClick={checkKimiCode}
+                  disabled={kimiCodeChecking}
+                  className="px-3 py-1.5 text-sm rounded border border-border hover:bg-hover-2 disabled:opacity-50 transition-colors flex items-center gap-1.5"
+                >
+                  <svg
+                    width="14"
+                    height="14"
+                    viewBox="0 0 24 24"
+                    fill="none"
+                    stroke="currentColor"
+                    strokeWidth="2"
+                    strokeLinecap="round"
+                    strokeLinejoin="round"
+                    className={kimiCodeChecking ? "animate-spin" : ""}
+                  >
+                    <path d="M21 12a9 9 0 1 1-2.64-6.36" />
+                    <polyline points="21 3 21 9 15 9" />
+                  </svg>
+                  {kimiCodeChecking ? t("checking") : t("kimiCodeCheck")}
+                </button>
+                {kimiCodeUpdateAvailable && (
+                  <button
+                    type="button"
+                    onClick={() => openExternal(KIMI_CODE_RELEASES_URL)}
+                    className="px-3 py-1.5 text-sm rounded border border-border hover:bg-hover-2 transition-colors"
+                  >
+                    {t("kimiCodeOpenReleases")}
+                  </button>
+                )}
+              </div>
+            </div>
+          </section>
+
           {/* About */}
           <section>
             <h3 className="mb-3 text-[11px] uppercase tracking-wider text-content-muted">

src/components/SubagentSettingsPage.tsx

@@ -48,10 +48,6 @@ const FLAG_LABELS: Record<string, { name: TranslationKey; desc: TranslationKey }
   notify_user: { name: "flagNotifyUser", desc: "flagNotifyUserDesc" },
   tower: { name: "flagTower", desc: "flagTowerDesc" },
   subagent_fork: { name: "flagSubagentFork", desc: "flagSubagentForkDesc" },
-  auto_session_title: {
-    name: "flagAutoSessionTitle",
-    desc: "flagAutoSessionTitleDesc",
-  },
 };
 
 /** Validation-error i18n key per engine rule, for the pre-write self-check. */

src/components/UsageConfigModal.tsx

@@ -364,7 +364,7 @@ export function UsageConfigModal({ open, agent, provider, onClose, onSave }: Usa
                 )}
               </div>
 
-              {/* timeout + auto query interval */}
+              {/* timeout + auto query interval + alert threshold */}
               <div className="rounded-lg border border-border px-4 py-4 grid grid-cols-2 gap-6">
                 <div>
                   <label className="block text-sm text-content-muted mb-1.5">{t("usageTimeout")}</label>
@@ -394,6 +394,31 @@ export function UsageConfigModal({ open, agent, provider, onClose, onSave }: Usa
                   />
                   <p className="mt-1 text-xs text-content-muted">{t("usageAutoIntervalHint")}</p>
                 </div>
+                <div>
+                  <label className="block text-sm text-content-muted mb-1.5">{t("usageThreshold")} %</label>
+                  <input
+                    type="number"
+                    min={0}
+                    max={100}
+                    className="w-32 bg-input border border-border rounded-lg px-3 py-2 text-sm focus:ring-2 focus:ring-blue-500 focus:outline-none"
+                    value={cfg.threshold ?? ""}
+                    onChange={(e) => {
+                      const raw = e.target.value.trim();
+                      // 清空输入框写回 undefined（关闭），而不是 0：两者判定等价，
+                      // 但 undefined 会被 serde 的 skip_serializing_if 从 SQLite
+                      // JSON 里省掉，不留冗余字段。
+                      if (raw === "") {
+                        setCfg({ ...cfg, threshold: undefined });
+                        return;
+                      }
+                      const n = parseInt(raw, 10);
+                      // 非数字（如 "1e" / "-"）不落库，保留上一次的有效值。
+                      if (Number.isNaN(n)) return;
+                      setCfg({ ...cfg, threshold: Math.min(100, Math.max(0, n)) });
+                    }}
+                  />
+                  <p className="mt-1 text-xs text-content-muted">{t("usageThresholdHint")}</p>
+                </div>
               </div>
             </>
           )}

src/components/UsageFooter.tsx

@@ -1,7 +1,13 @@
 import { useEffect, useState } from "react";
 import { Loader2, RefreshCw } from "lucide-react";
 import { useTranslation } from "../i18n";
-import { formatAmount, localizeUsageError, planLabel } from "../lib/usage-display";
+import {
+  BOOSTER_WALLET_PLAN,
+  formatAmount,
+  localizeUsageError,
+  planLabel,
+} from "../lib/usage-display";
+import { isAlertTier } from "../hooks/useUsageQuery";
 import type { UsageData, UsageQueryState } from "../hooks/useUsageQuery";
 
 interface UsageFooterProps {
@@ -14,6 +20,20 @@ interface UsageFooterProps {
   variant?: "detail" | "compact";
 }
 
+/**
+ * 金额行 vs 配额行：NewAPI 与 Kimi 加力钱包（booster_wallet）只报金额，其
+ * used/total 是钱数而不是配额比例，必须走 formatAmount 渲染路径。判定只能看
+ * planName —— sub2api 的 USD 行 unit 同样不是 "%"，但语义仍是「额度用了百分之
+ * 多少」，按 unit 判定会把它误渲染成余额。
+ */
+function isMoneyRow(d: UsageData): boolean {
+  return d.planName === "NewAPI" || d.planName === BOOSTER_WALLET_PLAN;
+}
+
+function isPlanRow(d: UsageData): boolean {
+  return !!d.planName && !isMoneyRow(d) && (d.total != null || d.used != null);
+}
+
 export function UsageFooter({ usage, variant = "detail" }: UsageFooterProps) {
   const { t } = useTranslation();
   const { status, data, error, updatedAt, supported, refresh } = usage;
@@ -85,10 +105,7 @@ export function UsageFooter({ usage, variant = "detail" }: UsageFooterProps) {
     // A "plan" is a quota tier (used/total, e.g. 5h / weekly limit). Balance
     // entries carry only remaining + unit — their plan_name is a currency /
     // brand label and must render as an amount, not a percentage.
-    const isPlan =
+    const isPlan = isPlanRow(d);
-      !!d.planName &&
-      d.planName !== "NewAPI" &&
-      (d.total != null || d.used != null);
     if (d.isValid === false) {
       return { icon: "⚡", text: t("usageInvalidKey"), color: "text-red-500 dark:text-red-400" };
     }
@@ -207,13 +224,14 @@ export function UsageFooter({ usage, variant = "detail" }: UsageFooterProps) {
       {data.map((d, i) => {
         // Same plan-vs-balance rule as compactSummary: balance rows have no
         // quota fields (total/used), so they render as an amount.
-        const isPlan =
+        const isPlan = isPlanRow(d);
-          !!d.planName &&
-          d.planName !== "NewAPI" &&
-          (d.total != null || d.used != null);
         const pct = isPlan ? percentOf(d) : null;
-        const color =
+        // 预警高亮：usage.alert 已按「查询成功」门控，失败时保留的上次数据
-          pct == null
+        // 不会被误标红（此处是唯一的高亮来源，随每次查询结果重算）。
+        const rowAlert = usage.alert && isAlertTier(d, usage.threshold);
+        const color = rowAlert
+          ? "text-red-500 dark:text-red-400"
+          : pct == null
             ? "text-content-muted"
             : pct < 70
               ? "text-green-600 dark:text-green-400"
@@ -248,6 +266,15 @@ export function UsageFooter({ usage, variant = "detail" }: UsageFooterProps) {
             >
               {main}
             </span>
+            {rowAlert && (
+              <span
+                className="shrink-0 text-red-500 dark:text-red-400"
+                title={t("usageThresholdAlert")}
+                aria-label={t("usageThresholdAlert")}
+              >
+                ⚠
+              </span>
+            )}
             {resetText && (
               <span className="text-content-muted shrink-0">{resetText}</span>
             )}

src/hooks/useHealthCheck.test.ts

@@ -0,0 +1,125 @@
+import { describe, expect, it } from "vitest";
+import {
+  localizeHealthError,
+  probeLine,
+  usageLine,
+  healthDotClass,
+  type CheckResult,
+} from "./useHealthCheck";
+
+/**
+ * 只回显 key，便于断言选中的是哪条文案（真实插值由 i18n Provider 负责）。
+ */
+const t = ((key: string) => `<${key}>`) as unknown as Parameters<
+  typeof localizeHealthError
+>[1];
+
+/** 带插值的假 t：`<key|{a}=1>`，用于断言参数透传。 */
+const tWithVars = ((key: string, vars?: Record<string, string | number>) =>
+  vars
+    ? `<${key}|${Object.entries(vars)
+        .map(([k, v]) => `${k}=${v}`)
+        .join(",")}>`
+    : `<${key}>`) as unknown as Parameters<typeof localizeHealthError>[1];
+
+describe("localizeHealthError", () => {
+  it("探活错误按前缀映射到对应文案", () => {
+    expect(localizeHealthError("missing base_url", t)).toBe(
+      "<healthErrNoBaseUrl>"
+    );
+    expect(localizeHealthError("no API key configured", t)).toBe(
+      "<healthErrNoKey>"
+    );
+    expect(localizeHealthError("Authentication failed (HTTP 401)", t)).toBe(
+      "<healthErrAuth>"
+    );
+    expect(localizeHealthError("endpoint not found (HTTP 404)", t)).toBe(
+      "<healthErrNotFound>"
+    );
+    expect(localizeHealthError("request timed out", t)).toBe(
+      "<healthErrTimeout>"
+    );
+    expect(
+      localizeHealthError("connection failed (DNS or refused)", t)
+    ).toBe("<healthErrUnreachable>");
+    expect(localizeHealthError("response is not valid JSON", t)).toBe(
+      "<healthErrBadJson>"
+    );
+  });
+
+  it("未知错误原样透出（HTTP 状态码等诊断细节）", () => {
+    expect(localizeHealthError("API error (HTTP 502)", t)).toBe(
+      "API error (HTTP 502)"
+    );
+  });
+
+  it("探活 + 账单两段拼接时逐段映射", () => {
+    expect(
+      localizeHealthError(
+        "endpoint not found (HTTP 404); usage query disabled in config panel",
+        t
+      )
+    ).toBe("<healthErrNotFound>; <usageErrDisabled>");
+  });
+
+  it("账单链路的报错复用用量错误映射", () => {
+    // commands.rs 的 kinds 循环会把单条错误包成 `{kind}: {msg}`。
+    expect(
+      localizeHealthError(
+        "Authentication failed (HTTP 401); balance:deepseek: Authentication failed (HTTP 401)",
+        t
+      )
+    ).toBe("<healthErrAuth>; <usageInvalidKey>");
+  });
+});
+
+describe("probeLine", () => {
+  it("通过时带延迟", () => {
+    expect(probeLine({ state: "pass" }, 123, tWithVars)).toBe(
+      "<healthProbeOk|latency=123>"
+    );
+  });
+
+  it("中性（404）说明端点不支持模型列表，而不是报失败", () => {
+    expect(
+      probeLine(
+        { state: "neutral", error: "endpoint not found (HTTP 404)" },
+        0,
+        tWithVars
+      )
+    ).toBe("<healthProbeUnsupported>");
+  });
+
+  it("失败时透出本地化后的原因", () => {
+    expect(
+      probeLine({ state: "fail", error: "request timed out" }, 0, tWithVars)
+    ).toBe("<healthProbeFail|error=<healthErrTimeout>>");
+  });
+});
+
+describe("usageLine", () => {
+  it("通过 / 失败 / 兜底三态", () => {
+    const pass: CheckResult = { state: "pass" };
+    expect(usageLine(pass, tWithVars)).toBe("<healthUsageOk>");
+
+    expect(
+      usageLine(
+        { state: "fail", error: "Authentication failed (HTTP 401)" },
+        tWithVars
+      )
+    ).toBe("<healthUsageFail|error=<healthErrAuth>>");
+
+    // 账单链路不会产生中性结论，兜底按「未判定」展示
+    expect(usageLine({ state: "neutral" }, tWithVars)).toBe(
+      "<healthNotChecked>"
+    );
+  });
+});
+
+describe("healthDotClass", () => {
+  it("绿=通过、红=失败、灰=未判定（中性）", () => {
+    expect(healthDotClass("pass")).toBe("bg-green-500");
+    expect(healthDotClass("fail")).toBe("bg-red-500");
+    expect(healthDotClass("neutral")).toBe("bg-gray-400 dark:bg-gray-500");
+  });
+});

src/hooks/useHealthCheck.ts

@@ -0,0 +1,175 @@
+import { useCallback, useEffect, useRef, useState } from "react";
+import { invoke } from "@tauri-apps/api/core";
+import type { Agent } from "../types";
+import type { TranslationKey } from "../i18n/zh";
+import { localizeUsageError } from "../lib/usage-display";
+
+type TranslateFn = (
+  key: TranslationKey,
+  vars?: Record<string, string | number>
+) => string;
+
+/** 单项检测的三态结论（mirrors Rust `HealthCheck`）。 */
+export type HealthCheck = "pass" | "fail" | "neutral";
+
+/** 单项检测结论（探活 / 账单各一份，mirrors Rust `CheckResult`）。 */
+export interface CheckResult {
+  state: HealthCheck;
+  /** 失败 / 中性原因的原始文案；通过为 null。 */
+  error?: string | null;
+}
+
+/** Mirrors the Rust `HealthResult` (camelCase via serde). */
+export interface HealthResult {
+  providerName: string;
+  /** 总体判定（Rust 侧是唯一事实来源）：pass=绿 / fail=红 / neutral=灰。 */
+  verdict: HealthCheck;
+  /** verdict === "pass" 的便捷布尔。 */
+  ok: boolean;
+  /** 失败原因汇总（仅 fail 项，`; ` 分隔）；无失败为 null。 */
+  error?: string | null;
+  /** 账单/用量结论；null/undefined = 未执行。 */
+  usage?: CheckResult | null;
+  /** 推理端点探活结论；探活是必做项，恒为 Some。 */
+  probe?: CheckResult | null;
+  /** 探活耗时（毫秒）。 */
+  latencyMs: number;
+}
+
+/**
+ * 单条 Rust 错误文案 → 本地化文案。未知错误原样透出（HTTP 状态码等诊断
+ * 细节）。前缀与 src-tauri/src/services/probe.rs、commands.rs 保持一致。
+ */
+function localizeHealthPart(error: string, t: TranslateFn): string {
+  if (error.startsWith("missing base_url")) return t("healthErrNoBaseUrl");
+  if (error.startsWith("no API key configured")) return t("healthErrNoKey");
+  if (error.startsWith("Authentication failed")) return t("healthErrAuth");
+  if (error.startsWith("endpoint not found")) return t("healthErrNotFound");
+  if (error.startsWith("request timed out")) return t("healthErrTimeout");
+  if (error.startsWith("connection failed")) return t("healthErrUnreachable");
+  if (error.startsWith("response is not valid JSON")) return t("healthErrBadJson");
+  // 账单链路的报错复用已有的用量错误映射（禁用 / OAuth 过期 / NewAPI 凭据…）。
+  return localizeUsageError(error, t);
+}
+
+/**
+ * 体检结论里的错误文案本地化。探活与账单两条链路的错误由 Rust 用 `; `
+ * 拼进同一个字段，逐段映射后再拼回，保证两段都能翻译。
+ */
+export function localizeHealthError(error: string, t: TranslateFn): string {
+  return error
+    .split("; ")
+    .map((part) => localizeHealthPart(part, t))
+    .join("; ");
+}
+
+/** 探活一项的 tooltip 文案。 */
+export function probeLine(
+  probe: CheckResult,
+  latencyMs: number,
+  t: TranslateFn
+): string {
+  if (probe.state === "pass") return t("healthProbeOk", { latency: latencyMs });
+  // 中性 = 端点未实现 /v1/models（404），如实说明「不支持」而不是报失败。
+  if (probe.state === "neutral") return t("healthProbeUnsupported");
+  return t("healthProbeFail", {
+    error: probe.error ? localizeHealthError(probe.error, t) : "",
+  });
+}
+
+/** 账单一项的 tooltip 文案。 */
+export function usageLine(usage: CheckResult, t: TranslateFn): string {
+  if (usage.state === "pass") return t("healthUsageOk");
+  if (usage.state === "fail") {
+    return t("healthUsageFail", {
+      error: usage.error ? localizeHealthError(usage.error, t) : "",
+    });
+  }
+  // 账单链路不产生中性结论（只有探活 404 会），兜底按「未判定」展示。
+  return t("healthNotChecked");
+}
+
+/**
+ * 状态点颜色：绿=正常、红=失败、灰=未判定（未检测 / 检测中 / 全部中性）。
+ * 判定本身由 Rust 的 overall_verdict 给出，这里只做映射。
+ */
+export function healthDotClass(verdict: HealthCheck): string {
+  switch (verdict) {
+    case "pass":
+      return "bg-green-500";
+    case "fail":
+      return "bg-red-500";
+    default:
+      return "bg-gray-400 dark:bg-gray-500";
+  }
+}
+
+export interface HealthCheckState {
+  /** 体检进行中：按钮转圈、尚未出结果的行显示灰色检测中指示。 */
+  running: boolean;
+  /** 至少跑过一轮（决定是否渲染状态点）。 */
+  ran: boolean;
+  /** provider name → 结果。 */
+  results: Record<string, HealthResult>;
+  okCount: number;
+  failCount: number;
+  /** 整体请求失败（配置读取失败 / 总超时）的文案；null = 无错误。 */
+  error: string | null;
+  run: () => void;
+}
+
+/**
+ * 一键体检：invoke `health_check_all`，把结果按 provider 名索引。
+ * 结果保留到下次点击或组件卸载（页面刷新），不在后台轮询。
+ */
+export function useHealthCheck(agent: Agent): HealthCheckState {
+  const [running, setRunning] = useState(false);
+  const [ran, setRan] = useState(false);
+  const [results, setResults] = useState<Record<string, HealthResult>>({});
+  const [error, setError] = useState<string | null>(null);
+  // Generation counter: 重跑/卸载后到达的旧响应直接丢弃，不写回状态。
+  const genRef = useRef(0);
+
+  useEffect(
+    () => () => {
+      genRef.current += 1;
+    },
+    []
+  );
+
+  const run = useCallback(async () => {
+    const gen = ++genRef.current;
+    setRunning(true);
+    setError(null);
+    // 清空上一轮结果：检测中的行显示灰色转圈，而不是留着旧结论误导。
+    setResults({});
+    try {
+      const list = await invoke<HealthResult[]>("health_check_all", { agent });
+      if (genRef.current !== gen) return;
+      const map: Record<string, HealthResult> = {};
+      for (const r of list) map[r.providerName] = r;
+      setResults(map);
+    } catch (e) {
+      if (genRef.current !== gen) return;
+      // invoke 拒绝时 Rust 侧 Err(String) 是字符串，不是 Error 实例。
+      setError(typeof e === "string" ? e : e instanceof Error ? e.message : String(e));
+    } finally {
+      if (genRef.current === gen) {
+        setRunning(false);
+        setRan(true);
+      }
+    }
+  }, [agent]);
+
+  const values = Object.values(results);
+  return {
+    running,
+    ran,
+    results,
+    // 中性（灰）既不算正常也不算失败，不计入汇总。
+    okCount: values.filter((r) => r.verdict === "pass").length,
+    failCount: values.filter((r) => r.verdict === "fail").length,
+    error,
+    run: () => void run(),
+  };
+}

src/hooks/useKimiCodeVersion.test.ts

@@ -0,0 +1,28 @@
+import { describe, expect, it } from "vitest";
+import { isNewerVersion } from "./useKimiCodeVersion";
+
+describe("isNewerVersion", () => {
+  it("同版本不算有更新", () => {
+    expect(isNewerVersion("2.0.2", "2.0.2")).toBe(false);
+    expect(isNewerVersion("v2.0.2", "2.0.2")).toBe(false);
+  });
+
+  it("按 major/minor/patch 逐段比较", () => {
+    expect(isNewerVersion("2.0.2", "2.0.1")).toBe(true);
+    expect(isNewerVersion("2.1.0", "2.0.9")).toBe(true);
+    expect(isNewerVersion("3.0.0", "2.99.99")).toBe(true);
+    expect(isNewerVersion("2.0.1", "2.0.2")).toBe(false);
+    expect(isNewerVersion("1.9.9", "2.0.0")).toBe(false);
+  });
+
+  it("忽略 v 前缀与预发布后缀", () => {
+    expect(isNewerVersion("v0.44.0-beta.1", "0.43.1")).toBe(true);
+    expect(isNewerVersion("0.43.1", "0.43.1-rc.2")).toBe(false);
+  });
+
+  it("缺位补 0，无法解析时不显示更新", () => {
+    expect(isNewerVersion("2.1", "2.0.9")).toBe(true);
+    expect(isNewerVersion("abc", "2.0.1")).toBe(false);
+    expect(isNewerVersion("2.0.2", "abc")).toBe(false);
+  });
+});

src/hooks/useKimiCodeVersion.ts

@@ -0,0 +1,101 @@
+import { useCallback, useEffect, useRef, useState } from "react";
+import { invoke } from "@tauri-apps/api/core";
+
+/** kimi-code 与本应用的适配状态（Rust 侧基线表决定）。 */
+export type KimiCodeStatus =
+  | "current"
+  | "outdated"
+  | "partial"
+  | "unknown"
+  | "not_installed";
+
+export interface KimiCodeVersionInfo {
+  /** 本机安装的 kimi-code 版本；未安装 / 检测失败为 null。 */
+  installed: string | null;
+  /** 上游最新 release tag（已去掉 v 前缀）；查询失败为 null。 */
+  latest: string | null;
+  status: KimiCodeStatus;
+}
+
+const THROTTLE_KEY = "kimi-switch-last-kc-version-check";
+// 与 useUpdateCheck 一致：节流窗口与定时复查同为 8h，长时间挂着的窗口也能
+// 自动拿到新版本，同时避免短时间反复触发后端调用。
+const CHECK_INTERVAL_MS = 8 * 60 * 60 * 1000; // 8 hours
+
+/** 解析 major.minor.patch 三元组；缺位补 0，无法解析返回 null。 */
+function versionTriple(version: string): [number, number, number] | null {
+  const core = version.trim().replace(/^[vV]/, "").split(/[-+]/)[0];
+  const nums = core.split(".").slice(0, 3).map((p) => Number.parseInt(p.trim(), 10));
+  if (nums.length === 0 || nums.some((n) => !Number.isFinite(n) || n < 0)) return null;
+  while (nums.length < 3) nums.push(0);
+  return [nums[0], nums[1], nums[2]];
+}
+
+/** latest 是否严格高于 installed（用于决定是否显示 releases 链接）。 */
+export function isNewerVersion(latest: string, installed: string): boolean {
+  const l = versionTriple(latest);
+  const i = versionTriple(installed);
+  if (!l || !i) return false;
+  for (let k = 0; k < 3; k += 1) {
+    if (l[k] !== i[k]) return l[k] > i[k];
+  }
+  return false;
+}
+
+/**
+ * 读取「本机 kimi-code 版本 / 上游最新版 / 适配状态」，模式与 useUpdateCheck
+ * 相同：挂载时按 8h 节流自动查一次，之后定时复查，checkNow 强制刷新。
+ */
+export function useKimiCodeVersion() {
+  const [info, setInfo] = useState<KimiCodeVersionInfo | null>(null);
+  const [checking, setChecking] = useState(false);
+  const [lastChecked, setLastChecked] = useState<number | null>(() => {
+    try {
+      const v = localStorage.getItem(THROTTLE_KEY);
+      return v ? Number(v) : null;
+    } catch {
+      return null;
+    }
+  });
+
+  const doCheck = useCallback(
+    async (force: boolean) => {
+      if (!force) {
+        if (lastChecked && Date.now() - lastChecked < CHECK_INTERVAL_MS) return;
+      }
+      setChecking(true);
+      try {
+        const next = await invoke<KimiCodeVersionInfo>("get_kimi_code_version_info");
+        setInfo(next);
+        const now = Date.now();
+        setLastChecked(now);
+        try {
+          localStorage.setItem(THROTTLE_KEY, String(now));
+        } catch {
+          // ignore
+        }
+      } catch {
+        // 静默失败：检测不到 kimi / 网络不通都不该打扰用户，手动检查仍可用。
+      } finally {
+        setChecking(false);
+      }
+    },
+    [lastChecked]
+  );
+
+  // 定时器不随 lastChecked 重建，所以用 ref 持有最新的 doCheck。
+  const doCheckRef = useRef(doCheck);
+  useEffect(() => {
+    doCheckRef.current = doCheck;
+  }, [doCheck]);
+
+  useEffect(() => {
+    doCheckRef.current(false);
+    const interval = setInterval(() => doCheckRef.current(false), CHECK_INTERVAL_MS);
+    return () => clearInterval(interval);
+  }, []);
+
+  const checkNow = useCallback(() => doCheck(true), [doCheck]);
+
+  return { info, checking, checkNow, lastChecked };
+}

src/hooks/useUsageQuery.test.ts

@@ -0,0 +1,44 @@
+import { describe, expect, it } from "vitest";
+import { isAlertTier, 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 {
+  return { planName: "five_hour", used, total: 100, remaining: 100 - used, unit: "%" };
+}
+
+describe("isAlertTier", () => {
+  it("阈值缺省或 0（关闭）时不告警", () => {
+    expect(isAlertTier(percentTier(100), undefined)).toBe(false);
+    expect(isAlertTier(percentTier(100), null)).toBe(false);
+    expect(isAlertTier(percentTier(100), 0)).toBe(false);
+  });
+
+  it("低于阈值不告警", () => {
+    expect(isAlertTier(percentTier(79), 80)).toBe(false);
+    expect(isAlertTier(percentTier(79.9), 80)).toBe(false);
+  });
+
+  it("等于或超过阈值告警", () => {
+    expect(isAlertTier(percentTier(80), 80)).toBe(true);
+    expect(isAlertTier(percentTier(80.5), 80)).toBe(true);
+    expect(isAlertTier(percentTier(100), 80)).toBe(true);
+  });
+
+  it("非 % unit 的行不告警（金额行的 used 是钱数）", () => {
+    // Kimi 加力钱包：used = monthlyUsedCents/100，与百分比阈值不可比。
+    expect(
+      isAlertTier(
+        { planName: "booster_wallet", used: 8, total: 50, remaining: 12.34, unit: "USD" },
+        5
+      )
+    ).toBe(false);
+    // sub2api 的 USD 配额行同理，不参与百分比预警。
+    expect(
+      isAlertTier({ planName: "daily_limit", used: 90, total: 100, unit: "USD" }, 50)
+    ).toBe(false);
+  });
+
+  it("used 缺失时不告警", () => {
+    expect(isAlertTier({ planName: "five_hour", unit: "%", remaining: 20 }, 50)).toBe(false);
+  });
+});

src/hooks/useUsageQuery.ts

@@ -56,6 +56,23 @@ export interface UsageQueryState {
   updatedAt: number | null;
   refresh: (force: boolean) => void;
   supported: boolean;
+  /** 生效的预警阈值（0/缺失已归一为 undefined）；footer 用它判定单行高亮。 */
+  threshold?: number;
+  /** 查询成功且任一百分比行 used ≥ threshold。失败/加载中恒为 false， */
+  /** 所以失败时保留的上次数据不会被误高亮。 */
+  alert: boolean;
+}
+
+/**
+ * 单行是否触发预警。只判定百分比行（unit === "%"）：余额/钱包等金额行的
+ * used 是钱数，与百分比阈值不可比。threshold 缺失或 ≤0（关闭）时恒为 false。
+ */
+export function isAlertTier(
+  d: UsageData,
+  threshold?: number | null
+): boolean {
+  if (threshold == null || threshold <= 0) return false;
+  return d.unit === "%" && d.used != null && d.used >= threshold;
 }
 
 export function useUsageQuery(
@@ -63,7 +80,8 @@ export function useUsageQuery(
   providerName: string,
   usageKinds?: string[],
   autoIntervalMinutes?: number,
-  disabled?: boolean
+  disabled?: boolean,
+  threshold?: number
 ): UsageQueryState {
   const supported = (usageKinds?.length ?? 0) > 0;
   // Cache key includes the agent so a Kimi Code provider and a Pi provider
@@ -175,6 +193,15 @@ export function useUsageQuery(
     return () => clearInterval(id);
   }, [supported, autoIntervalMinutes, runQuery, disabled]);
 
+  // 阈值 0/负数/非有限值一律视为关闭（与配置面板「0 或空 = 关闭」一致）。
+  // 每次查询结果写入 data 后此处重新求值，因此刷新后预警自动出现/消除。
+  const alertThreshold =
+    typeof threshold === "number" && Number.isFinite(threshold) && threshold > 0
+      ? threshold
+      : undefined;
+  const alert =
+    status === "success" && data.some((d) => isAlertTier(d, alertThreshold));
+
   return {
     status,
     data,
@@ -182,5 +209,7 @@ export function useUsageQuery(
     updatedAt,
     refresh: runQuery,
     supported,
+    threshold: alertThreshold,
+    alert,
   };
 }

src/i18n/en.ts

@@ -43,6 +43,26 @@ export const enTranslations: Record<TranslationKey, string> = {
   connectivityOk: "{name} reachable ({latency}ms)",
   connectivitySlow: "{name} reachable but slow ({latency}ms)",
   connectivityFail: "{name} unreachable: {error}",
+  // Health check (ProviderList toolbar button + per-row status dot)
+  healthCheck: "Health check",
+  healthCheckHint:
+    "Probe every enabled provider's inference endpoint and billing query in parallel",
+  healthCheckRunning: "Checking…",
+  healthSummary: "{ok} OK / {fail} failed",
+  healthProbeOk: "Endpoint OK ({latency}ms)",
+  healthProbeUnsupported: "Endpoint exposes no model list (404)",
+  healthProbeFail: "Endpoint failed: {error}",
+  healthUsageOk: "Billing OK",
+  healthUsageFail: "Billing failed: {error}",
+  healthNotChecked: "Not checked",
+  healthRequestFailed: "Health check request failed: {error}",
+  healthErrNoBaseUrl: "Missing base URL",
+  healthErrNoKey: "No API key configured",
+  healthErrAuth: "Authentication failed",
+  healthErrNotFound: "Endpoint not found",
+  healthErrTimeout: "Connection timed out",
+  healthErrUnreachable: "Cannot connect (DNS failure or connection refused)",
+  healthErrBadJson: "Response is not valid JSON",
   copiedProvider: "Duplicated as {name}",
 
   // Provider edit
@@ -78,6 +98,12 @@ export const enTranslations: Record<TranslationKey, string> = {
   authField: "Auth Field",
   apiKey: "API Key",
   getApiKeyLink: "Get API Key",
+  apiKeySource: "Credential source",
+  apiKeySourceDirect: "Enter key",
+  apiKeySourceEnv: "Environment variable",
+  apiKeyEnvName: "Variable name",
+  apiKeyEnvHint:
+    "Only the variable name (api_key_env) is written; Kimi Code reads the secret from its environment. The two sources are mutually exclusive, so switching clears the other.",
   show: "Show",
   hide: "Hide",
   requestUrl: "Request URL",
@@ -142,9 +168,16 @@ export const enTranslations: Record<TranslationKey, string> = {
   loopControlSettings: "Loop Control",
   maxAttemptsPerStep: "Max attempts per step",
   reservedContextSize: "Reserved context size",
+  compactionMaxAttempts: "Max compaction attempts",
+  compactionMaxAttemptsHint:
+    "Retry limit after a failed context compaction. Empty or 0 leaves the key out of the config (upstream default 5).",
   backgroundSettings: "Background Tasks",
   maxRunningTasks: "Max running tasks",
   keepAliveOnExit: "Keep alive on exit",
+  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.",
   permissionRules: "Permission Rules",
   permissionDecision: "Decision",
   permissionPattern: "Pattern",
@@ -283,6 +316,18 @@ export const enTranslations: Record<TranslationKey, string> = {
   install: "Install",
   updateCheckFailed: "Update check failed",
   lastChecked: "Last checked",
+  // Kimi Code version tracking
+  kimiCodeSection: "Kimi Code version",
+  kimiCodeInstalled: "Installed",
+  kimiCodeLatest: "Latest",
+  kimiCodeStatusLabel: "Compatibility",
+  kimiCodeStatusCurrent: "Supported",
+  kimiCodeStatusOutdated: "Update KimiSwitch recommended",
+  kimiCodeStatusPartial: "Partially compatible",
+  kimiCodeStatusUnknown: "Unknown version",
+  kimiCodeStatusNotInstalled: "Not installed",
+  kimiCodeCheck: "Check",
+  kimiCodeOpenReleases: "Open releases",
   appDescription: "Windows desktop LLM provider config manager",
   settingsWebsite: "Official website",
   kimiEyesSite: "Kimi Eyes plugin site",
@@ -449,6 +494,11 @@ export const enTranslations: Record<TranslationKey, string> = {
   usageTierDaily: "Daily",
   usageTierWeekly: "Weekly",
   usageTierMonthly: "Monthly",
+  usageTierMonthCode: "Monthly (code)",
+  usageTierBoosterWallet: "Booster",
+  usageThreshold: "Alert threshold",
+  usageThresholdHint: "Highlight any window that reaches this percentage; 0 = off",
+  usageThresholdAlert: "Threshold reached",
   usageErrNoKey: "No API key configured",
   usageErrDisabled: "Usage query is disabled in the config panel",
   usageErrLoginExpired: "Kimi Code login expired and auto-refresh failed; run `kimi login` again",
@@ -490,8 +540,6 @@ export const enTranslations: Record<TranslationKey, string> = {
   flagSubagentForkDesc: "Agent/AgentSwarm tools can start a subagent with a snapshot of the calling agent's history (fork parameter)",
   flagWaitFor: "WaitFor Tool",
   flagWaitForDesc: "The agent can wait for background tasks to finish within the current turn",
-  flagAutoSessionTitle: "Auto Session Title",
-  flagAutoSessionTitleDesc: "Generate session titles automatically via the managed chat_title tool",
   flagNotifyUser: "NotifyUser tool",
   flagNotifyUserDesc: "Experimental Updates panel: paginated progress messages from the main agent and subagents (off by default)",
   flagDefaultOn: "On by default",

src/i18n/zh.ts

@@ -41,6 +41,25 @@ export const zhTranslations = {
   connectivityOk: "{name} 连通正常（{latency}ms）",
   connectivitySlow: "{name} 连通较慢（{latency}ms）",
   connectivityFail: "{name} 无法连通：{error}",
+  // 一键体检（ProviderList 工具栏按钮 + 行内状态点）
+  healthCheck: "一键体检",
+  healthCheckHint: "并行检测所有启用供应商的推理端点与账单查询",
+  healthCheckRunning: "体检中…",
+  healthSummary: "{ok} 正常 / {fail} 失败",
+  healthProbeOk: "端点正常（{latency}ms）",
+  healthProbeUnsupported: "端点不支持模型列表（404）",
+  healthProbeFail: "端点失败：{error}",
+  healthUsageOk: "账单正常",
+  healthUsageFail: "账单失败：{error}",
+  healthNotChecked: "未检测",
+  healthRequestFailed: "体检请求失败：{error}",
+  healthErrNoBaseUrl: "缺少 base_url",
+  healthErrNoKey: "未配置 API Key",
+  healthErrAuth: "鉴权失败",
+  healthErrNotFound: "端点不存在",
+  healthErrTimeout: "连接超时",
+  healthErrUnreachable: "无法连接（DNS 解析失败或连接被拒绝）",
+  healthErrBadJson: "响应不是合法 JSON",
   copiedProvider: "已复制为 {name}",
 
   // Provider edit
@@ -76,6 +95,12 @@ export const zhTranslations = {
   authField: "认证字段",
   apiKey: "API Key",
   getApiKeyLink: "获取 API Key",
+  apiKeySource: "凭证来源",
+  apiKeySourceDirect: "直接填写",
+  apiKeySourceEnv: "环境变量",
+  apiKeyEnvName: "环境变量名",
+  apiKeyEnvHint:
+    "只写入变量名（api_key_env），密钥由 Kimi Code 从运行环境读取；两者互斥，切换后会清空另一种。",
   show: "显示",
   hide: "隐藏",
   requestUrl: "请求地址",
@@ -140,9 +165,15 @@ export const zhTranslations = {
   loopControlSettings: "循环控制",
   maxAttemptsPerStep: "单步最大尝试次数",
   reservedContextSize: "上下文预留大小",
+  compactionMaxAttempts: "压缩最大尝试次数",
+  compactionMaxAttemptsHint: "上下文压缩失败后的重试上限，留空或 0 表示不写入该项（上游默认 5）。",
   backgroundSettings: "后台任务",
   maxRunningTasks: "最大并发数",
   keepAliveOnExit: "退出时保持运行",
+  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。",
   permissionRules: "权限规则",
   permissionDecision: "处置",
   permissionPattern: "模式",
@@ -277,6 +308,18 @@ export const zhTranslations = {
   install: "安装",
   updateCheckFailed: "检查更新失败",
   lastChecked: "上次检查",
+  // Kimi Code 版本跟踪
+  kimiCodeSection: "Kimi Code 版本",
+  kimiCodeInstalled: "已安装版本",
+  kimiCodeLatest: "最新版本",
+  kimiCodeStatusLabel: "适配状态",
+  kimiCodeStatusCurrent: "已适配",
+  kimiCodeStatusOutdated: "建议升级 KimiSwitch",
+  kimiCodeStatusPartial: "部分兼容",
+  kimiCodeStatusUnknown: "未知版本",
+  kimiCodeStatusNotInstalled: "未安装",
+  kimiCodeCheck: "检查",
+  kimiCodeOpenReleases: "打开 Releases",
   appDescription: "Windows 桌面端 LLM 供应商配置管理器",
   settingsWebsite: "官方网站",
   kimiEyesSite: "Kimi Eyes 插件站",
@@ -443,6 +486,11 @@ export const zhTranslations = {
   usageTierDaily: "每日",
   usageTierWeekly: "7天",
   usageTierMonthly: "30天",
+  usageTierMonthCode: "30天（代码）",
+  usageTierBoosterWallet: "加力钱包",
+  usageThreshold: "预警阈值",
+  usageThresholdHint: "任一用量窗口达到该百分比时高亮提醒，0 为关闭",
+  usageThresholdAlert: "已达预警阈值",
   usageErrNoKey: "未配置 API Key",
   usageErrDisabled: "用量查询已在配置面板中停用",
   usageErrLoginExpired: "Kimi Code 登录已过期且自动续期失败，请重新运行 `kimi login`",
@@ -484,8 +532,6 @@ export const zhTranslations = {
   flagSubagentForkDesc: "Agent/AgentSwarm 工具可携带主代理会话快照启动子代理（fork 参数）",
   flagWaitFor: "WaitFor 工具",
   flagWaitForDesc: "代理可在当前回合内等待后台任务完成，而无需结束回合再被唤醒",
-  flagAutoSessionTitle: "自动会话标题",
-  flagAutoSessionTitleDesc: "通过托管 chat_title 工具自动生成会话标题",
   flagNotifyUser: "NotifyUser 工具",
   flagNotifyUserDesc: "实验性 Updates 面板：分页展示主代理与子代理的进度消息（默认关闭）",
   flagDefaultOn: "默认开启",

src/lib/agent-settings.test.ts

@@ -134,10 +134,104 @@ describe("loop_control — v1/v2 key handling", () => {
 });
 
 // ---------------------------------------------------------------------------
-// [permission] dangerous_command_guard (kimi-code 0.40.1) — absent key means
+// loop_control.compaction_max_attempts (kimi-code 0.43.0) — read via the loop
-// upstream default ON; an explicit false must survive even with no rules
+// spread, written only when set, and 0/empty removes the key (upstream
+// default 5 applies)
 // ---------------------------------------------------------------------------
 
+describe("loop_control.compaction_max_attempts", () => {
+  it("reads an existing value through the loop spread", () => {
+    expect(
+      getAgentSettings({ loop_control: { compaction_max_attempts: 7 } })
+        .loop_control?.compaction_max_attempts
+    ).toBe(7);
+  });
+
+  it("reads as undefined when the key is absent (UI falls back to 5)", () => {
+    expect(
+      getAgentSettings({ loop_control: {} }).loop_control?.compaction_max_attempts
+    ).toBeUndefined();
+  });
+
+  it("writes the value when the patch sets it", () => {
+    const next = setAgentSettings({}, {
+      loop_control: { compaction_max_attempts: 8 },
+    });
+    expect(loopOf(next).compaction_max_attempts).toBe(8);
+  });
+
+  it("drops the key for 0 and for a cleared (undefined) field", () => {
+    const raw = { loop_control: { compaction_max_attempts: 6 } };
+    expect(loopOf(setAgentSettings(raw, {
+      loop_control: { compaction_max_attempts: 0 },
+    }))).not.toHaveProperty("compaction_max_attempts");
+    expect(loopOf(setAgentSettings(raw, {
+      loop_control: { compaction_max_attempts: undefined },
+    }))).not.toHaveProperty("compaction_max_attempts");
+  });
+
+  it("does not materialize the upstream default on an unrelated save", () => {
+    // An absent key must stay absent — a plain save must not pin `5`.
+    const next = setAgentSettings({ thinking: { effort: "high" } }, {});
+    expect(loopOf(next)).not.toHaveProperty("compaction_max_attempts");
+    // An existing value survives a save that does not touch it.
+    const kept = setAgentSettings(
+      { loop_control: { compaction_max_attempts: 4 } },
+      { thinking: { effort: "low" } }
+    );
+    expect(loopOf(kept).compaction_max_attempts).toBe(4);
+  });
+});
+
+// ---------------------------------------------------------------------------
+// [watch] enabled (kimi-code 2.0.1+) — top-level section, off by default since
+// 2.0.2; mirrors the [background] handling
+// ---------------------------------------------------------------------------
+
+function watchOf(raw: unknown): Record<string, unknown> {
+  const w = (raw as { watch?: unknown })?.watch;
+  return w && typeof w === "object" && !Array.isArray(w)
+    ? (w as 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 an explicit value", () => {
+    expect(getAgentSettings({ watch: { enabled: true } }).watch?.enabled).toBe(true);
+    expect(getAgentSettings({ watch: { enabled: false } }).watch?.enabled).toBe(false);
+  });
+
+  it("writes the top-level [watch] section", () => {
+    const next = setAgentSettings({}, { watch: { enabled: true } }) as {
+      watch?: Record<string, unknown>;
+    };
+    expect(next.watch).toEqual({ enabled: true });
+  });
+
+  it("keeps the section on saves that do not touch it", () => {
+    const next = setAgentSettings(
+      { watch: { enabled: true } },
+      { thinking: { effort: "high" } }
+    ) as { watch?: Record<string, unknown> };
+    expect(next.watch).toEqual({ enabled: true });
+  });
+
+  it("does not disturb the other managed sections", () => {
+    const next = setAgentSettings(
+      { background: { max_running_tasks: 2 } },
+      { watch: { enabled: true } }
+    ) as { background?: Record<string, unknown> };
+    expect(next.background?.max_running_tasks).toBe(2);
+    expect(watchOf(next).enabled).toBe(true);
+  });
+});
+
+
 describe("permission.dangerous_command_guard", () => {
   it("reads as undefined when the key is absent (default on)", () => {
     expect(

src/lib/agent-settings.ts

@@ -13,6 +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: false,
+  },
   permission: { rules: [] },
   hooks: [],
 };
@@ -76,6 +81,10 @@ export function getAgentSettings(rawOther: unknown): AgentSettings {
       ...DEFAULT_SETTINGS.background,
       ...getSection<AgentSettings["background"]>(rawOther, "background"),
     },
+    watch: {
+      ...DEFAULT_SETTINGS.watch,
+      ...getSection<AgentSettings["watch"]>(rawOther, "watch"),
+    },
     permission: {
       rules: sectionPermission?.rules ?? [],
       // `dangerous_command_guard` stays undefined unless the config carries
@@ -111,6 +120,7 @@ export function setAgentSettings(
     thinking: { ...current.thinking, ...patch.thinking },
     loop_control: { ...current.loop_control, ...patch.loop_control },
     background: { ...current.background, ...patch.background },
+    watch: { ...current.watch, ...patch.watch },
     permission,
     hooks: patch.hooks ?? current.hooks ?? [],
   };
@@ -129,9 +139,16 @@ export function setAgentSettings(
   if (!opts?.legacyV1) {
     delete next.loop_control?.max_retries_per_step;
   }
+  // `compaction_max_attempts` (kimi-code 0.43.0+): 0 / undefined means "leave
+  // the key out" — the config then falls back to the upstream default (5)
+  // instead of pinning a literal 0.
+  if (!next.loop_control?.compaction_max_attempts) {
+    delete next.loop_control?.compaction_max_attempts;
+  }
   if (next.thinking) root.thinking = next.thinking;
   if (next.loop_control) root.loop_control = next.loop_control;
   if (next.background) root.background = next.background;
+  if (next.watch) root.watch = next.watch;
   // Keep `[permission]` when it carries rules *or* an explicit
   // dangerous_command_guard — dropping the section would silently lose a
   // guard=false write (absent key = upstream default on). An empty rules

src/lib/models-dev-full.json

@@ -1,5 +1,5 @@
 {
-  "last_updated": "2026-09-13",
+  "last_updated": "2026-09-14",
   "providers": {
     "subconscious": {
       "id": "subconscious",
@@ -16470,6 +16470,20 @@
           "output_limit": 32768,
           "image": true
         },
+        {
+          "id": "alibaba/deepseek-v4.1-flash",
+          "name": "DeepSeek V4.1 Flash (Alibaba Cloud)",
+          "cost": {
+            "input": 0.3,
+            "output": 1.2,
+            "cache_read": 0.03
+          },
+          "context": 1000000,
+          "output_limit": 393216,
+          "reasoning": true,
+          "tool_call": true,
+          "image": true
+        },
         {
           "id": "alibaba/qwen-coder-plus",
           "name": "Qwen Coder Plus (Alibaba Cloud)",
@@ -16730,7 +16744,7 @@
           "id": "scx-ai-gp/glm-5.2",
           "name": "GLM-5.2 (SCX.ai)",
           "cost": {
-            "input": 0.8,
+            "input": 0.88,
             "output": 2.55,
             "cache_read": 0.16
           },
@@ -18921,6 +18935,17 @@
           "reasoning": true,
           "tool_call": true
         },
+        {
+          "id": "consensusprotocol/deepseek-v4.1-flash",
+          "name": "DeepSeek V4.1 Flash (Consensus Protocol)",
+          "cost": {
+            "input": 0.2,
+            "output": 0.6,
+            "cache_read": 0.005
+          },
+          "context": 1048576,
+          "output_limit": 384000
+        },
         {
           "id": "consensusprotocol/glm-5.3-flash",
           "name": "GLM-5.3 Flash (Consensus Protocol)",
@@ -38111,9 +38136,9 @@
           "id": "gemma-4-26b-a4b-it",
           "name": "Gemma 4 26B A4B IT",
           "cost": {
-            "input": 0.102,
+            "input": 0.1,
-            "output": 0.356,
+            "output": 0.374,
-            "cache_read": 0.051
+            "cache_read": 0.05
           },
           "context": 256000,
           "output_limit": 25600,
@@ -38129,7 +38154,7 @@
             "output": 4.311648,
             "cache_read": 0.047907
           },
-          "context": 1048576,
+          "context": 1000000,
           "output_limit": 262144,
           "reasoning": true,
           "tool_call": true,
@@ -38195,9 +38220,9 @@
           "id": "minimax-m2.7",
           "name": "MiniMax-M2.7",
           "cost": {
-            "input": 0.396,
+            "input": 0.462,
-            "output": 1.464,
+            "output": 1.728,
-            "cache_read": 0.198
+            "cache_read": 0.231
           },
           "context": 262100,
           "output_limit": 6553,
@@ -38265,7 +38290,7 @@
           "cost": {
             "input": 0.2,
             "output": 0.4,
-            "cache_read": 0.04
+            "cache_write": 0.04
           },
           "context": 1000000,
           "output_limit": 384000,
@@ -38281,7 +38306,7 @@
             "output": 4.3552,
             "cache_read": 0.206872
           },
-          "context": 262000,
+          "context": 256000,
           "output_limit": 16000,
           "reasoning": true,
           "tool_call": true,
@@ -38305,12 +38330,12 @@
           "id": "deepseek-v4.1-flash",
           "name": "DeepSeek V4.1 Flash",
           "cost": {
-            "input": 0.3,
+            "input": 0.32664,
-            "output": 1.2,
+            "output": 1.30656,
-            "cache_read": 0.03
+            "cache_read": 0.032664
           },
-          "context": 1048576,
+          "context": 1000000,
-          "output_limit": 26214,
+          "output_limit": 13107,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true,
@@ -38425,9 +38450,9 @@
           "id": "glm-5",
           "name": "GLM-5",
           "cost": {
-            "input": 0.86,
+            "input": 0.94,
-            "output": 2.752,
+            "output": 3.008,
-            "cache_read": 0.43
+            "cache_read": 0.47
           },
           "context": 202752,
           "output_limit": 20275,
@@ -38452,9 +38477,9 @@
           "id": "kimi-k2.5",
           "name": "Kimi K2.5",
           "cost": {
-            "input": 0.5584,
+            "input": 0.5344,
-            "output": 2.935,
+            "output": 2.815,
-            "cache_read": 0.2792
+            "cache_read": 0.2672
           },
           "context": 262144,
           "output_limit": 26214,
@@ -38467,7 +38492,7 @@
           "name": "GLM-5.1",
           "cost": {
             "input": 1.318,
-            "output": 4.308,
+            "output": 4.268,
             "cache_read": 0.659
           },
           "context": 202750,
@@ -38496,7 +38521,7 @@
           "cost": {
             "input": 2.4,
             "output": 4.8,
-            "cache_read": 0.2
+            "cache_write": 0.2
           },
           "context": 1000000,
           "output_limit": 384000,
@@ -38508,9 +38533,9 @@
           "id": "gpt-oss-120b",
           "name": "GPT OSS 120B",
           "cost": {
-            "input": 0.168,
+            "input": 0.178,
-            "output": 0.66,
+            "output": 0.68,
-            "cache_read": 0.084
+            "cache_read": 0.089
           },
           "context": 128072,
           "output_limit": 13107,
@@ -65693,6 +65718,21 @@
           "tool_call": true,
           "structured_output": true
         },
+        {
+          "id": "Qwen/Qwen3.8-Flash",
+          "name": "Qwen3.8 Flash",
+          "cost": {
+            "input": 0.113,
+            "output": 0.382,
+            "cache_read": 0.0141
+          },
+          "context": 1000000,
+          "output_limit": 131072,
+          "tool_call": true,
+          "structured_output": true,
+          "image": true,
+          "video": true
+        },
         {
           "id": "Qwen/Qwen3-VL-235B-A22B-Instruct",
           "name": "Qwen3 VL 235B A22B Instruct",
@@ -66211,8 +66251,8 @@
             "input": 0.2275,
             "output": 0.91
           },
-          "context": 131072,
+          "context": 40960,
-          "output_limit": 8192,
+          "output_limit": 16384,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true
@@ -66976,12 +67016,12 @@
           "id": "~deepseek/deepseek-v4-flash-latest",
           "name": "DeepSeek: DeepSeek V4 Flash Latest",
           "cost": {
-            "input": 0.0352,
+            "input": 0.04,
-            "output": 0.1056,
+            "output": 0.1,
-            "cache_read": 0.00112
+            "cache_read": 0.01
           },
           "context": 1048576,
-          "output_limit": 131072,
+          "output_limit": 393216,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true
@@ -67591,7 +67631,7 @@
             "output": 0.45
           },
           "context": 262144,
-          "output_limit": 16384,
+          "output_limit": 235929,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true
@@ -67641,8 +67681,8 @@
             "output": 2.2,
             "cache_read": 0.1
           },
-          "context": 256000,
+          "context": 202800,
-          "output_limit": 32768,
+          "output_limit": 182520,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true
@@ -68842,7 +68882,8 @@
           "name": "MoonshotAI: Kimi Latest",
           "cost": {
             "input": 2.1,
-            "output": 10.95
+            "output": 10.95,
+            "cache_read": 0.23
           },
           "context": 1048576,
           "output_limit": 943718,
@@ -70605,12 +70646,12 @@
           "id": "~z-ai/glm-latest",
           "name": "Z.ai: GLM Latest",
           "cost": {
-            "input": 0.936,
+            "input": 0.92,
-            "output": 3.168,
+            "output": 3.1372,
-            "cache_read": 0.1872
+            "cache_read": 0.184
           },
-          "context": 262144,
+          "context": 1048576,
-          "output_limit": 235929,
+          "output_limit": 943718,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true
@@ -70718,7 +70759,8 @@
           "name": "Inference.net: Schematron V2 Small",
           "cost": {
             "input": 0.05,
-            "output": 0.23
+            "output": 0.23,
+            "cache_read": 0.05
           },
           "context": 128000,
           "output_limit": 4096,
@@ -70729,7 +70771,8 @@
           "name": "Inference.net: Schematron V2 Turbo",
           "cost": {
             "input": 0.03,
-            "output": 0.15
+            "output": 0.15,
+            "cache_read": 0.03
           },
           "context": 128000,
           "output_limit": 8192,
@@ -70997,8 +71040,8 @@
             "output": 4.4,
             "cache_read": 0.26
           },
-          "context": 202752,
+          "context": 1048576,
-          "output_limit": 182476,
+          "output_limit": 131072,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true
@@ -71095,8 +71138,8 @@
             "output": 4.4,
             "cache_read": 0.26
           },
-          "context": 1048576,
+          "context": 1048575,
-          "output_limit": 131072,
+          "output_limit": 943717,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true
@@ -81488,11 +81531,11 @@
           "id": "qwen/qwen3-14b",
           "name": "Qwen3 14B",
           "cost": {
-            "input": 0.2275,
+            "input": 0.12,
-            "output": 0.91
+            "output": 0.24
           },
           "context": 131072,
-          "output_limit": 8192,
+          "output_limit": 16384,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true
@@ -82261,12 +82304,12 @@
           "id": "~deepseek/deepseek-v4-flash-latest",
           "name": "DeepSeek V4 Flash Latest",
           "cost": {
-            "input": 0.0352,
+            "input": 0.04,
-            "output": 0.1056,
+            "output": 0.1,
-            "cache_read": 0.00112
+            "cache_read": 0.01
           },
           "context": 1310720,
-          "output_limit": 131072,
+          "output_limit": 393216,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true
@@ -82859,11 +82902,11 @@
           "id": "nvidia/nemotron-3-super-120b-a12b",
           "name": "Nemotron 3 Super 120B A12B",
           "cost": {
-            "input": 0.085,
+            "input": 0.08,
-            "output": 0.4
+            "output": 0.45
           },
           "context": 262144,
-          "output_limit": 16384,
+          "output_limit": 235929,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true
@@ -82909,12 +82952,12 @@
           "id": "nvidia/nemotron-3-ultra-550b-a55b",
           "name": "Nemotron 3 Ultra 550B A55B",
           "cost": {
-            "input": 0.625,
+            "input": 0.6,
-            "output": 3.125,
+            "output": 2.4,
-            "cache_read": 0.1875
+            "cache_read": 0.12
           },
           "context": 262144,
-          "output_limit": 32768,
+          "output_limit": 182520,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true
@@ -84157,7 +84200,8 @@
           "name": "Kimi Latest",
           "cost": {
             "input": 2.1,
-            "output": 10.95
+            "output": 10.95,
+            "cache_read": 0.23
           },
           "context": 1048576,
           "output_limit": 943718,
@@ -84224,9 +84268,9 @@
           "id": "deepseek/deepseek-v4-pro-0813",
           "name": "DeepSeek V4 Pro 0813",
           "cost": {
-            "input": 0.57948,
+            "input": 0.9834,
-            "output": 1.73844,
+            "output": 2.9502,
-            "cache_read": 0.019316
+            "cache_read": 0.03278
           },
           "context": 1048576,
           "output_limit": 384000,
@@ -84238,9 +84282,9 @@
           "id": "deepseek/deepseek-v4-flash-0731",
           "name": "DeepSeek V4 Flash 0731",
           "cost": {
-            "input": 0.04,
+            "input": 0.06,
-            "output": 0.08,
+            "output": 0.12,
-            "cache_read": 0.008
+            "cache_read": 0.012
           },
           "context": 1310720,
           "output_limit": 943718,
@@ -84252,9 +84296,9 @@
           "id": "deepseek/deepseek-v4-flash",
           "name": "DeepSeek V4 Flash",
           "cost": {
-            "input": 0.04788,
+            "input": 0.08554,
-            "output": 0.09576,
+            "output": 0.17108,
-            "cache_read": 0.009576
+            "cache_read": 0.017108
           },
           "context": 1048576,
           "output_limit": 384000,
@@ -85829,12 +85873,12 @@
           "id": "~z-ai/glm-latest",
           "name": "GLM Latest",
           "cost": {
-            "input": 0.936,
+            "input": 0.92,
-            "output": 3.168,
+            "output": 3.1372,
-            "cache_read": 0.1872
+            "cache_read": 0.184
           },
           "context": 1310720,
-          "output_limit": 235929,
+          "output_limit": 943718,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true
@@ -85942,7 +85986,8 @@
           "name": "Schematron V2 Small",
           "cost": {
             "input": 0.05,
-            "output": 0.23
+            "output": 0.23,
+            "cache_read": 0.05
           },
           "context": 128000,
           "output_limit": 4096,
@@ -85953,7 +85998,8 @@
           "name": "Schematron V2 Turbo",
           "cost": {
             "input": 0.03,
-            "output": 0.15
+            "output": 0.15,
+            "cache_read": 0.03
           },
           "context": 128000,
           "output_limit": 8192,
@@ -86062,9 +86108,9 @@
           "id": "tencent/hy3",
           "name": "Hy3",
           "cost": {
-            "input": 0.0825,
+            "input": 0.132,
-            "output": 0.33,
+            "output": 0.528,
-            "cache_read": 0.020625
+            "cache_read": 0.033
           },
           "context": 262144,
           "input_limit": 192000,
@@ -86217,12 +86263,12 @@
           "id": "z-ai/glm-5.2",
           "name": "GLM-5.2",
           "cost": {
-            "input": 0.6,
+            "input": 0.6832,
-            "output": 2,
+            "output": 2.1472,
-            "cache_read": 0.15
+            "cache_read": 0.12688
           },
           "context": 1048576,
-          "output_limit": 182476,
+          "output_limit": 131072,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true
@@ -86231,9 +86277,9 @@
           "id": "z-ai/glm-5.3-flash",
           "name": "GLM-5.3-Flash",
           "cost": {
-            "input": 0.075,
+            "input": 0.15,
-            "output": 0.25,
+            "output": 0.5,
-            "cache_read": 0.015
+            "cache_read": 0.03
           },
           "context": 1310720,
           "output_limit": 131072,
@@ -86315,12 +86361,12 @@
           "id": "z-ai/glm-5.3",
           "name": "GLM-5.3",
           "cost": {
-            "input": 1.092,
+            "input": 1.4,
-            "output": 3.432,
+            "output": 4.4,
-            "cache_read": 0.2028
+            "cache_read": 0.26
           },
           "context": 1310720,
-          "output_limit": 131072,
+          "output_limit": 943717,
           "reasoning": true,
           "tool_call": true,
           "structured_output": true
@@ -102924,8 +102970,9 @@
           "id": "qwen/deepseek-v4-pro-0813",
           "name": "DeepSeek V4 Pro 0813 (Alibaba)",
           "cost": {
-            "input": 0.5808,
+            "input": 0.66,
-            "output": 1.7424
+            "output": 1.98,
+            "cache_read": 0.066
           },
           "context": 1000000,
           "output_limit": 384000,
@@ -102937,8 +102984,9 @@
           "id": "qwen/deepseek-v4-flash-0731",
           "name": "DeepSeek V4 Flash 0731 (Alibaba)",
           "cost": {
-            "input": 0.176,
+            "input": 0.22,
-            "output": 0.528
+            "output": 0.66,
+            "cache_read": 0.022
           },
           "context": 1000000,
           "output_limit": 384000,
@@ -103080,6 +103128,21 @@
           "output_limit": 8192,
           "reasoning": true
         },
+        {
+          "id": "qwen/deepseek-v4.1-flash",
+          "name": "DeepSeek V4.1 Flash (Alibaba)",
+          "cost": {
+            "input": 0.15,
+            "output": 0.6,
+            "cache_read": 0.015
+          },
+          "context": 1000000,
+          "output_limit": 384000,
+          "reasoning": true,
+          "tool_call": true,
+          "structured_output": true,
+          "image": true
+        },
         {
           "id": "qwen/qwen3-coder-next",
           "name": "Qwen3 Coder Next",
@@ -104143,6 +104206,20 @@
           "reasoning": true,
           "tool_call": true
         },
+        {
+          "id": "tensorx/deepseek/deepseek-v4.1-flash",
+          "name": "DeepSeek V4.1 Flash (TensorX)",
+          "cost": {
+            "input": 0.5,
+            "output": 1.5,
+            "cache_read": 0.125
+          },
+          "context": 1048576,
+          "output_limit": 384000,
+          "reasoning": true,
+          "tool_call": true,
+          "image": true
+        },
         {
           "id": "tensorx/moonshotai/kimi-k2.5",
           "name": "Kimi K2.5 (TensorX)",

src/lib/models-dev.json

@@ -1,5 +1,5 @@
 {
-  "last_updated": "2026-09-13",
+  "last_updated": "2026-09-14",
   "subconscious/subconscious/glm-5.2": {
     "name": "GLM-5.2",
     "context": 1000000,
@@ -13257,6 +13257,18 @@
       "cache_write": 0.25
     }
   },
+  "llmgateway-providers/alibaba/deepseek-v4.1-flash": {
+    "name": "DeepSeek V4.1 Flash (Alibaba Cloud)",
+    "context": 1000000,
+    "reasoning": true,
+    "tool_call": true,
+    "image": true,
+    "cost": {
+      "input": 0.3,
+      "output": 1.2,
+      "cache_read": 0.03
+    }
+  },
   "llmgateway-providers/alibaba/qwen-coder-plus": {
     "name": "Qwen Coder Plus (Alibaba Cloud)",
     "context": 131072,
@@ -13482,7 +13494,7 @@
     "reasoning": true,
     "tool_call": true,
     "cost": {
-      "input": 0.8,
+      "input": 0.88,
       "output": 2.55,
       "cache_read": 0.16
     }
@@ -15353,6 +15365,15 @@
       "cache_read": 0.01
     }
   },
+  "llmgateway-providers/consensusprotocol/deepseek-v4.1-flash": {
+    "name": "DeepSeek V4.1 Flash (Consensus Protocol)",
+    "context": 1048576,
+    "cost": {
+      "input": 0.2,
+      "output": 0.6,
+      "cache_read": 0.005
+    }
+  },
   "llmgateway-providers/consensusprotocol/glm-5.3-flash": {
     "name": "GLM-5.3 Flash (Consensus Protocol)",
     "context": 1048576,
@@ -31345,14 +31366,14 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.102,
+      "input": 0.1,
-      "output": 0.356,
+      "output": 0.374,
-      "cache_read": 0.051
+      "cache_read": 0.05
     }
   },
   "hyper/deepseek-v4-pro-0813": {
     "name": "DeepSeek V4 Pro 0813",
-    "context": 1048576,
+    "context": 1000000,
     "reasoning": true,
     "tool_call": true,
     "structured_output": true,
@@ -31416,9 +31437,9 @@
     "reasoning": true,
     "tool_call": true,
     "cost": {
-      "input": 0.396,
+      "input": 0.462,
-      "output": 1.464,
+      "output": 1.728,
-      "cache_read": 0.198
+      "cache_read": 0.231
     }
   },
   "hyper/kimi-k2.6": {
@@ -31477,12 +31498,12 @@
     "cost": {
       "input": 0.2,
       "output": 0.4,
-      "cache_read": 0.04
+      "cache_write": 0.04
     }
   },
   "hyper/kimi-k2.7-code": {
     "name": "Kimi K2.7 Code",
-    "context": 262000,
+    "context": 256000,
     "reasoning": true,
     "tool_call": true,
     "structured_output": true,
@@ -31506,15 +31527,15 @@
   },
   "hyper/deepseek-v4.1-flash": {
     "name": "DeepSeek V4.1 Flash",
-    "context": 1048576,
+    "context": 1000000,
     "reasoning": true,
     "tool_call": true,
     "structured_output": true,
     "image": true,
     "cost": {
-      "input": 0.3,
+      "input": 0.32664,
-      "output": 1.2,
+      "output": 1.30656,
-      "cache_read": 0.03
+      "cache_read": 0.032664
     }
   },
   "hyper/qwen3.7-flash": {
@@ -31613,9 +31634,9 @@
     "reasoning": true,
     "tool_call": true,
     "cost": {
-      "input": 0.86,
+      "input": 0.94,
-      "output": 2.752,
+      "output": 3.008,
-      "cache_read": 0.43
+      "cache_read": 0.47
     }
   },
   "hyper/qwen3.8-max": {
@@ -31637,9 +31658,9 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.5584,
+      "input": 0.5344,
-      "output": 2.935,
+      "output": 2.815,
-      "cache_read": 0.2792
+      "cache_read": 0.2672
     }
   },
   "hyper/glm-5.1": {
@@ -31650,7 +31671,7 @@
     "structured_output": true,
     "cost": {
       "input": 1.318,
-      "output": 4.308,
+      "output": 4.268,
       "cache_read": 0.659
     }
   },
@@ -31675,7 +31696,7 @@
     "cost": {
       "input": 2.4,
       "output": 4.8,
-      "cache_read": 0.2
+      "cache_write": 0.2
     }
   },
   "hyper/gpt-oss-120b": {
@@ -31685,9 +31706,9 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.168,
+      "input": 0.178,
-      "output": 0.66,
+      "output": 0.68,
-      "cache_read": 0.084
+      "cache_read": 0.089
     }
   },
   "hyper/glm-5.3": {
@@ -54422,6 +54443,19 @@
       "output": 0.55
     }
   },
+  "deepinfra/Qwen/Qwen3.8-Flash": {
+    "name": "Qwen3.8 Flash",
+    "context": 1000000,
+    "tool_call": true,
+    "structured_output": true,
+    "image": true,
+    "video": true,
+    "cost": {
+      "input": 0.113,
+      "output": 0.382,
+      "cache_read": 0.0141
+    }
+  },
   "deepinfra/Qwen/Qwen3-VL-235B-A22B-Instruct": {
     "name": "Qwen3 VL 235B A22B Instruct",
     "context": 262144,
@@ -54852,7 +54886,7 @@
   },
   "kilo/qwen/qwen3-14b": {
     "name": "Qwen: Qwen3 14B",
-    "context": 131072,
+    "context": 40960,
     "reasoning": true,
     "tool_call": true,
     "structured_output": true,
@@ -55512,9 +55546,9 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.0352,
+      "input": 0.04,
-      "output": 0.1056,
+      "output": 0.1,
-      "cache_read": 0.00112
+      "cache_read": 0.01
     }
   },
   "kilo/dots-studio/dots-3-note-preview:free": {
@@ -56067,7 +56101,7 @@
   },
   "kilo/nvidia/nemotron-3-ultra-550b-a55b": {
     "name": "Nemotron 3 Ultra 550B A55B",
-    "context": 256000,
+    "context": 202800,
     "reasoning": true,
     "tool_call": true,
     "structured_output": true,
@@ -57115,7 +57149,8 @@
     "video": true,
     "cost": {
       "input": 2.1,
-      "output": 10.95
+      "output": 10.95,
+      "cache_read": 0.23
     }
   },
   "kilo/ibm-granite/granite-4.2-8b": {
@@ -58592,14 +58627,14 @@
   },
   "kilo/~z-ai/glm-latest": {
     "name": "Z.ai: GLM Latest",
-    "context": 262144,
+    "context": 1048576,
     "reasoning": true,
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.936,
+      "input": 0.92,
-      "output": 3.168,
+      "output": 3.1372,
-      "cache_read": 0.1872
+      "cache_read": 0.184
     }
   },
   "kilo/moonshotai/kimi-k2-0905": {
@@ -58692,7 +58727,8 @@
     "structured_output": true,
     "cost": {
       "input": 0.05,
-      "output": 0.23
+      "output": 0.23,
+      "cache_read": 0.05
     }
   },
   "kilo/inference-net/schematron-v2-turbo": {
@@ -58701,7 +58737,8 @@
     "structured_output": true,
     "cost": {
       "input": 0.03,
-      "output": 0.15
+      "output": 0.15,
+      "cache_read": 0.03
     }
   },
   "kilo/cohere/north-mini-code:free": {
@@ -58919,7 +58956,7 @@
   },
   "kilo/z-ai/glm-5.2": {
     "name": "GLM-5.2",
-    "context": 202752,
+    "context": 1048576,
     "reasoning": true,
     "tool_call": true,
     "structured_output": true,
@@ -59003,7 +59040,7 @@
   },
   "kilo/z-ai/glm-5.3": {
     "name": "GLM-5.3",
-    "context": 1048576,
+    "context": 1048575,
     "reasoning": true,
     "tool_call": true,
     "structured_output": true,
@@ -67739,8 +67776,8 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.2275,
+      "input": 0.12,
-      "output": 0.91
+      "output": 0.24
     }
   },
   "openrouter/qwen/qwen3.6-plus": {
@@ -68399,9 +68436,9 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.0352,
+      "input": 0.04,
-      "output": 0.1056,
+      "output": 0.1,
-      "cache_read": 0.00112
+      "cache_read": 0.01
     }
   },
   "openrouter/dots-studio/dots-3-note-preview:free": {
@@ -68906,8 +68943,8 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.085,
+      "input": 0.08,
-      "output": 0.4
+      "output": 0.45
     }
   },
   "openrouter/nvidia/nemotron-3-ultra-550b-a55b:free": {
@@ -68948,9 +68985,9 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.625,
+      "input": 0.6,
-      "output": 3.125,
+      "output": 2.4,
-      "cache_read": 0.1875
+      "cache_read": 0.12
     }
   },
   "openrouter/nvidia/nemotron-3-nano-30b-a3b": {
@@ -70028,7 +70065,8 @@
     "video": true,
     "cost": {
       "input": 2.1,
-      "output": 10.95
+      "output": 10.95,
+      "cache_read": 0.23
     }
   },
   "openrouter/ibm-granite/granite-4.2-8b": {
@@ -70083,9 +70121,9 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.57948,
+      "input": 0.9834,
-      "output": 1.73844,
+      "output": 2.9502,
-      "cache_read": 0.019316
+      "cache_read": 0.03278
     }
   },
   "openrouter/deepseek/deepseek-v4-flash-0731": {
@@ -70095,9 +70133,9 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.04,
+      "input": 0.06,
-      "output": 0.08,
+      "output": 0.12,
-      "cache_read": 0.008
+      "cache_read": 0.012
     }
   },
   "openrouter/deepseek/deepseek-v4-flash": {
@@ -70107,9 +70145,9 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.04788,
+      "input": 0.08554,
-      "output": 0.09576,
+      "output": 0.17108,
-      "cache_read": 0.009576
+      "cache_read": 0.017108
     }
   },
   "openrouter/deepseek/deepseek-v4.1-flash": {
@@ -71428,9 +71466,9 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.936,
+      "input": 0.92,
-      "output": 3.168,
+      "output": 3.1372,
-      "cache_read": 0.1872
+      "cache_read": 0.184
     }
   },
   "openrouter/moonshotai/kimi-k2-0905": {
@@ -71523,7 +71561,8 @@
     "structured_output": true,
     "cost": {
       "input": 0.05,
-      "output": 0.23
+      "output": 0.23,
+      "cache_read": 0.05
     }
   },
   "openrouter/inference-net/schematron-v2-turbo": {
@@ -71532,7 +71571,8 @@
     "structured_output": true,
     "cost": {
       "input": 0.03,
-      "output": 0.15
+      "output": 0.15,
+      "cache_read": 0.03
     }
   },
   "openrouter/cohere/north-mini-code:free": {
@@ -71625,9 +71665,9 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.0825,
+      "input": 0.132,
-      "output": 0.33,
+      "output": 0.528,
-      "cache_read": 0.020625
+      "cache_read": 0.033
     }
   },
   "openrouter/tencent/hy4-preview": {
@@ -71755,9 +71795,9 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.6,
+      "input": 0.6832,
-      "output": 2,
+      "output": 2.1472,
-      "cache_read": 0.15
+      "cache_read": 0.12688
     }
   },
   "openrouter/z-ai/glm-5.3-flash": {
@@ -71769,9 +71809,9 @@
     "image": true,
     "video": true,
     "cost": {
-      "input": 0.075,
+      "input": 0.15,
-      "output": 0.25,
+      "output": 0.5,
-      "cache_read": 0.015
+      "cache_read": 0.03
     }
   },
   "openrouter/z-ai/glm-4.5": {
@@ -71839,9 +71879,9 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 1.092,
+      "input": 1.4,
-      "output": 3.432,
+      "output": 4.4,
-      "cache_read": 0.2028
+      "cache_read": 0.26
     }
   },
   "openrouter/z-ai/glm-5v-turbo": {
@@ -85559,8 +85599,9 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.5808,
+      "input": 0.66,
-      "output": 1.7424
+      "output": 1.98,
+      "cache_read": 0.066
     }
   },
   "edenai/qwen/deepseek-v4-flash-0731": {
@@ -85570,8 +85611,9 @@
     "tool_call": true,
     "structured_output": true,
     "cost": {
-      "input": 0.176,
+      "input": 0.22,
-      "output": 0.528
+      "output": 0.66,
+      "cache_read": 0.022
     }
   },
   "edenai/qwen/qwen3-coder-plus": {
@@ -85688,6 +85730,19 @@
       "output": 2.4
     }
   },
+  "edenai/qwen/deepseek-v4.1-flash": {
+    "name": "DeepSeek V4.1 Flash (Alibaba)",
+    "context": 1000000,
+    "reasoning": true,
+    "tool_call": true,
+    "structured_output": true,
+    "image": true,
+    "cost": {
+      "input": 0.15,
+      "output": 0.6,
+      "cache_read": 0.015
+    }
+  },
   "edenai/qwen/qwen3-coder-next": {
     "name": "Qwen3 Coder Next",
     "context": 262144,
@@ -86603,6 +86658,18 @@
       "cache_read": 0.0625
     }
   },
+  "edenai/tensorx/deepseek/deepseek-v4.1-flash": {
+    "name": "DeepSeek V4.1 Flash (TensorX)",
+    "context": 1048576,
+    "reasoning": true,
+    "tool_call": true,
+    "image": true,
+    "cost": {
+      "input": 0.5,
+      "output": 1.5,
+      "cache_read": 0.125
+    }
+  },
   "edenai/tensorx/moonshotai/kimi-k2.5": {
     "name": "Kimi K2.5 (TensorX)",
     "context": 262144,

src/lib/subagent-settings.test.ts

@@ -461,22 +461,23 @@ describe("validateSubagentPool", () => {
 // ---------------------------------------------------------------------------
 
 describe("experimental flag registry", () => {
-  it("mirrors the 0.42.0 v2 registry (6 flags)", () => {
+  it("mirrors the 0.43.0 v2 registry (5 flags)", () => {
     expect(EXPERIMENTAL_FLAGS.map((f) => f.id)).toEqual([
       "wait_for",
       "tool-select",
       "notify_user",
       "tower",
       "subagent_fork",
-      "auto_session_title",
     ]);
     // Removed upstream: file_history (0.41.0, turn-level file history is
-    // always on) plus secondary-model / persistence_minidb_readmodel /
+    // always on), auto_session_title (0.43.0, graduated — AI session titles
-    // remote-control / search_worker (0.42.0 — promoted or moved to
+    // are always on, #3749), plus secondary-model /
-    // dedicated config, the minidb read-model now lives under [database]),
+    // persistence_minidb_readmodel / remote-control / search_worker
-    // so the adapter must not mirror them either.
+    // (0.42.0 — promoted or moved to dedicated config, the minidb read-model
+    // now lives under [database]), so the adapter must not mirror them either.
     for (const id of [
       "file_history",
+      "auto_session_title",
       "secondary-model",
       "persistence_minidb_readmodel",
       "remote-control",

src/lib/subagent-settings.ts

@@ -33,10 +33,11 @@ export interface ExperimentalFlagDef {
   defaultEnabled?: boolean;
 }
 
-/** Known experimental flags — mirrors the kimi-code 0.42.0 v2 flag registry
+/** Known experimental flags — mirrors the kimi-code 0.43.0 v2 flag registry
  * (the per-feature flag.ts files under packages/agent-core-v2/src):
- * wait_for, tool-select, notify_user, tower, subagent_fork,
+ * wait_for, tool-select, notify_user, tower, subagent_fork.
- * auto_session_title. Removed in 0.42.0: secondary-model, remote-control
+ * Removed in 0.43.0: auto_session_title (graduated — AI session titles are
+ * always on, #3749). Removed in 0.42.0: secondary-model, remote-control
  * and search_worker (promoted or dropped upstream — no flag left), and
  * persistence_minidb_readmodel (now controlled by the `[database]` config
  * section). `acp-v2` was removed upstream earlier; `file_history` was
@@ -53,10 +54,6 @@ export const EXPERIMENTAL_FLAGS: ExperimentalFlagDef[] = [
   { id: "notify_user", envVar: "KIMI_CODE_EXPERIMENTAL_NOTIFY_USER" },
   { id: "tower", envVar: "KIMI_CODE_EXPERIMENTAL_TOWER" },
   { id: "subagent_fork", envVar: "KIMI_CODE_EXPERIMENTAL_SUBAGENT_FORK" },
-  {
-    id: "auto_session_title",
-    envVar: "KIMI_CODE_EXPERIMENTAL_AUTO_SESSION_TITLE",
-  },
 ];
 
 export const EXPERIMENTAL_MASTER_ENV = "KIMI_CODE_EXPERIMENTAL_FLAG";

src/lib/usage-display.test.ts

@@ -59,6 +59,7 @@ describe("planLabel", () => {
     expect(planLabel("five_hour", t)).toBe("usageTier5h");
     expect(planLabel("weekly_limit", t)).toBe("usageTierWeekly");
     expect(planLabel("monthly_limit", t)).toBe("usageTierMonthly");
+    expect(planLabel("booster_wallet", t)).toBe("usageTierBoosterWallet");
     expect(planLabel("NewAPI", t)).toBe("NewAPI");
     expect(planLabel("¥", t)).toBe("¥");
   });

src/lib/usage-display.ts

@@ -5,15 +5,23 @@ type TranslateFn = (
   vars?: Record<string, string | number>
 ) => string;
 
+/**
+ * Kimi quota 的加力钱包行 id（src-tauri/src/services/coding_plan.rs 的
+ * PLAN_BOOSTER_WALLET）。它是金额行而非配额百分比行，planLabel() 与
+ * UsageFooter 的金额行判定共用此常量，避免字面量散落。
+ */
+export const BOOSTER_WALLET_PLAN = "booster_wallet";
+
 /**
  * Map a raw Rust plan/tier name to a localized label. Unknown names (proper
  * nouns like "OpenRouter", "NewAPI", currency codes) pass through unchanged.
  *
  * Coverage note: ALL plan-type (套餐) queries flow through
  * src-tauri/src/services/coding_plan.rs::percent_tier, whose names come only
- * from TIER_FIVE_HOUR / TIER_WEEKLY_LIMIT / TIER_MONTHLY_LIMIT — so this
+ * from TIER_FIVE_HOUR / TIER_WEEKLY_LIMIT / TIER_MONTHLY_LIMIT /
- * mapping covers every current and future plan provider automatically. If a
+ * TIER_MONTH_CODE — so this mapping covers every current and future plan
- * new tier id is ever added there, add a case here too.
+ * provider automatically. If a new tier id is ever added there, add a case
+ * here too.
  */
 export function planLabel(name: string, t: TranslateFn): string {
   switch (name) {
@@ -25,6 +33,10 @@ export function planLabel(name: string, t: TranslateFn): string {
       return t("usageTierWeekly");
     case "monthly_limit":
       return t("usageTierMonthly");
+    case "month_code":
+      return t("usageTierMonthCode");
+    case BOOSTER_WALLET_PLAN:
+      return t("usageTierBoosterWallet");
     default:
       return name;
   }

src/lib/validation.ts

@@ -50,7 +50,10 @@ export function validateProviders(
     const reasons: string[] = [];
 
     const apiKey = provider.api_key?.trim();
-    if (!apiKey && !provider.managed) {
+    // kimi-code 2.0.0+: the credential may come from an environment variable
+    // instead of a stored key — a provider configured that way is complete.
+    const apiKeyEnv = provider.api_key_env?.trim();
+    if (!apiKey && !apiKeyEnv && !provider.managed) {
       reasons.push(t("validationNoApiKey"));
     }
 

src/types/index.ts

@@ -15,6 +15,12 @@ export interface Provider {
   provider_type: ProviderType;
   base_url: string | null;
   api_key: string | null;
+  /**
+   * Name of the process environment variable holding the credential
+   * (kimi-code 2.0.0+). Mutually exclusive with `api_key` upstream — the
+   * export writes only one of the two. Only the variable *name* is stored.
+   */
+  api_key_env?: string | null;
   env: Record<string, string>;
   /** Optional note / remark for the provider. */
   note?: string | null;
@@ -90,6 +96,9 @@ export interface UsageConfig {
   autoQueryIntervalMinutes?: number;
   /** Per-request timeout in seconds; 0/undefined = default (8s). */
   timeoutSeconds?: number;
+  /** 用量预警阈值（百分比 0-100）；0/undefined = 关闭预警。
+   *  判定见 src/hooks/useUsageQuery.ts 的 isAlertTier()。 */
+  threshold?: number;
 }
 
 /** A model discovered from a provider's API endpoint. */
@@ -125,6 +134,9 @@ export interface LoopControlConfig {
    *  KIMI_CODE_LEGACY_FLAG=1 (v1 engine compat). v2 saves strip it. */
   max_retries_per_step?: number;
   reserved_context_size?: number;
+  /** Retry budget for context compaction (kimi-code 0.43.0+). Absent/0 = the
+   *  key is not written and upstream's default (5) applies. */
+  compaction_max_attempts?: number;
 }
 
 export interface BackgroundConfig {
@@ -132,6 +144,13 @@ export interface BackgroundConfig {
   keep_alive_on_exit?: boolean;
 }
 
+/** Top-level `[watch]` section (kimi-code 2.0.1+). The key was added in
+ *  2.0.1 and its default flipped to off in 2.0.2, so an absent key reads as
+ *  disabled. `KIMI_CODE_WATCH` overrides the config value at runtime. */
+export interface WatchConfig {
+  enabled?: boolean;
+}
+
 export interface PermissionRule {
   decision?: "allow" | "deny" | "ask";
   scope?: string;
@@ -150,6 +169,8 @@ export interface AgentSettings {
   thinking?: ThinkingConfig;
   loop_control?: LoopControlConfig;
   background?: BackgroundConfig;
+  /** Top-level `[watch]` section (kimi-code 2.0.1+). */
+  watch?: WatchConfig;
   permission?: {
     rules?: PermissionRule[];
     /** kimi-code `[permission]` key. Default true when the key is