2026.10.11: no-response timeout and abort, core v0.67.0

package.json

@@ -2,7 +2,7 @@
   "name": "thinkwatch-lite",
   "private": true,
   "description": "Desktop app for a local AI API gateway on macOS, Windows and Linux",
-  "version": "2026.10.10",
+  "version": "2026.10.11",
   "type": "module",
   "packageManager": "pnpm@11.13.0",
   "scripts": {

release-notes/2026.10.11.md

@@ -0,0 +1,22 @@
+**Upgrade notes:**
+- The bundled core is now 0.67.0, with control-plane protocol 43. A remote server must be upgraded to core 0.67.0 together with the app: this version does not connect to a server on 0.66.0, and 2026.10.10 and earlier do not connect to 0.67.0.
+- **The request history and stored payloads are cleared on the first start.**
+- The failover settings for a slow-starting stream are replaced by the no-response timeout. A configuration that still has `failover.next_on_slow_start` or `failover.stream_start_wait_secs` makes the gateway start in safe mode; the repair it offers deletes them.
+- Requests converted for Claude now use prompt caching: cache writes are billed at the cache-write rate and reads at the much lower cache-read rate, and both are priced and shown with the request.
+
+**No response and abort:**
+- **Settings › Failover** has a **No-response timeout**, 300 seconds by default. When an upstream sends no content for that long, the request moves to the next upstream; when none is left, the client gets a timeout error instead of waiting indefinitely. After an answer has started, it ends with an error the client understands. Keep-alive pings do not count as content. It replaces the switch for moving on when a stream is slow to start.
+- A running request can be stopped with **Abort request** in its menu on the Traffic page, and every running request of a session with **Abort session**. The request's details show **Aborted** or **No response** with the time waited.
+
+**Codex:**
+- Through a route that converts to Claude, GLM, DeepSeek or another format, Codex keeps all its tools. For some models Codex declares its tools in a way that was dropped, and it then reported that it could not read files or run commands. Its history, compaction of long sessions and instructions in the middle of a conversation also arrive now.
+- Requests converted for Claude are cached, so the turns of a long session cost much less.
+
+**Upstreams:**
+- Models can also be added by hand in an upstream's dialog, in **Models**, for any upstream and while creating one. They are marked **Manual** and saved with the upstream.
+
+**Traffic:**
+- Long upstream, model and key names no longer widen the table: they are cut with an ellipsis and shown in full on hover, also in the row menus.
+
+**ChatGPT account:**
+- Signing in on another device works again. OpenAI changed the format of a value in its sign-in answer, which stopped the sign-in with an error.

scripts/shots/core/en/overview.json

@@ -66,14 +66,13 @@
   "default_route": "default",
   "failover": {
     "failures_to_pause": 3,
+    "idle_timeout_secs": 300,
     "max_pause_secs": 600,
-    "next_on_slow_start": false,
     "no_balance_pause_secs": 1800,
     "pause_secs": 60,
     "quota_pause_secs": 3600,
     "rate_limit_max_pause_secs": 3600,
-    "slot_wait_secs": 30,
-    "stream_start_wait_secs": 15
+    "slot_wait_secs": 30
   },
   "groups": [
     {

scripts/shots/core/en/status.json

@@ -1,5 +1,5 @@
 {
-  "api_version": 38,
+  "api_version": 43,
   "clients": 4,
   "config_path": "",
   "gateway_addr": null,
@@ -14,5 +14,5 @@
     "reachable": []
   },
   "uptime_secs": 0,
-  "version": "0.62.0"
+  "version": "0.66.0"
 }

scripts/shots/core/zh/overview.json

@@ -66,14 +66,13 @@
   "default_route": "default",
   "failover": {
     "failures_to_pause": 3,
+    "idle_timeout_secs": 300,
     "max_pause_secs": 600,
-    "next_on_slow_start": false,
     "no_balance_pause_secs": 1800,
     "pause_secs": 60,
     "quota_pause_secs": 3600,
     "rate_limit_max_pause_secs": 3600,
-    "slot_wait_secs": 30,
-    "stream_start_wait_secs": 15
+    "slot_wait_secs": 30
   },
   "groups": [
     {

scripts/shots/core/zh/status.json

@@ -1,5 +1,5 @@
 {
-  "api_version": 38,
+  "api_version": 43,
   "clients": 4,
   "config_path": "",
   "gateway_addr": null,
@@ -14,5 +14,5 @@
     "reachable": []
   },
   "uptime_secs": 0,
-  "version": "0.62.0"
+  "version": "0.66.0"
 }

scripts/shots/mock/core.ts

@@ -104,6 +104,8 @@ export const CORE: { [N in WebviewEndpoint]: Handler<N> } = {
   Sessions: (req) => sessions(req.limit ?? 200),
   SessionDetail: (_req, [id]) => ({ session: sessionView(id!) ?? notFound(`Session ${id}`), turns: turns(id!) }),
   SessionTranscript: (req, [id]) => (sessionView(id!) ? transcript(id!, req.from_turn ?? 0) : notFound(`Session ${id}`)),
+  AbortRequest: refuse,
+  AbortSession: refuse,
   SpeedQuote: refuse,
   SpeedRun: refuse,
   ReplayQuote: refuse,

src-tauri/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "thinkwatch-lite"
-version = "2026.10.10"
+version = "2026.10.11"
 edition = "2024"
 # **这个数决定依赖能升到哪一版。**edition 2024 的解析器只挑声明的 Rust
 # 版本编得动的依赖：写 1.85 的时候，`cargo update` 一直停在旧的 time 和
@@ -26,12 +26,12 @@ license = "MIT"
 # twcore 二进制必须和这里编译进去的协议镜像来自同一个 core 版本 —— 打包脚本正是
 # 从 tw-api 锁到的 tag 去取二进制的（见 scripts/fetch-core.sh）。
 [workspace.dependencies]
-tw-api = { git = "https://github.com/ThinkWatchProject/ThinkWatch-Core.git", tag = "v0.66.0" }
-tw-types = { git = "https://github.com/ThinkWatchProject/ThinkWatch-Core.git", tag = "v0.66.0" }
-tw-yaml = { git = "https://github.com/ThinkWatchProject/ThinkWatch-Core.git", tag = "v0.66.0" }
-tw-guard = { git = "https://github.com/ThinkWatchProject/ThinkWatch-Core.git", tag = "v0.66.0" }
-tw-watch = { git = "https://github.com/ThinkWatchProject/ThinkWatch-Core.git", tag = "v0.66.0" }
-tw-link = { git = "https://github.com/ThinkWatchProject/ThinkWatch-Core.git", tag = "v0.66.0" }
+tw-api = { git = "https://github.com/ThinkWatchProject/ThinkWatch-Core.git", tag = "v0.67.0" }
+tw-types = { git = "https://github.com/ThinkWatchProject/ThinkWatch-Core.git", tag = "v0.67.0" }
+tw-yaml = { git = "https://github.com/ThinkWatchProject/ThinkWatch-Core.git", tag = "v0.67.0" }
+tw-guard = { git = "https://github.com/ThinkWatchProject/ThinkWatch-Core.git", tag = "v0.67.0" }
+tw-watch = { git = "https://github.com/ThinkWatchProject/ThinkWatch-Core.git", tag = "v0.67.0" }
+tw-link = { git = "https://github.com/ThinkWatchProject/ThinkWatch-Core.git", tag = "v0.67.0" }
 
 [lib]
 name = "thinkwatch_lite_lib"

src-tauri/src/call.rs

@@ -76,9 +76,11 @@ webview_endpoints![
     History,
     HistorySearch,
     RequestDetail,
+    AbortRequest,
     Sessions,
     SessionDetail,
     SessionTranscript,
+    AbortSession,
     // 测速、回放、试路由
     SpeedQuote,
     SpeedRun,

src-tauri/src/clients/desktop_rule.rs

@@ -477,8 +477,8 @@ mod tests {
             "security": {"redact": "observe", "inspect_tools": "observe", "content": "observe"},
             "retention": {"body_days": 7, "row_days": 90, "body_max_bytes": 0, "body_bytes_now": 0},
             "failover": {"failures_to_pause": 3, "pause_secs": 60, "max_pause_secs": 600, "no_balance_pause_secs": 1800,
-                         "quota_pause_secs": 3600, "rate_limit_max_pause_secs": 3600, "stream_start_wait_secs": 15,
-                         "next_on_slow_start": false, "slot_wait_secs": 30}
+                         "quota_pause_secs": 3600, "rate_limit_max_pause_secs": 3600, "idle_timeout_secs": 300,
+                         "slot_wait_secs": 30}
         })
     }
 

src-tauri/src/notices/fixtures/snapshot.json

@@ -175,8 +175,7 @@
    "no_balance_pause_secs": 1800,
    "quota_pause_secs": 3600,
    "rate_limit_max_pause_secs": 3600,
-   "stream_start_wait_secs": 15,
-   "next_on_slow_start": false,
+   "idle_timeout_secs": 300,
    "slot_wait_secs": 30
   },
   "price_sheets": []

src-tauri/tauri.conf.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://schema.tauri.app/config/2",
   "productName": "ThinkWatch Lite",
-  "version": "2026.10.10",
+  "version": "2026.10.11",
   "identifier": "app.thinkwatch.lite",
   "build": {
     "beforeDevCommand": "pnpm dev",

src/control.ts

@@ -41,9 +41,11 @@ export const WEBVIEW_ENDPOINTS = [
   "History",
   "HistorySearch",
   "RequestDetail",
+  "AbortRequest",
   "Sessions",
   "SessionDetail",
   "SessionTranscript",
+  "AbortSession",
   "SpeedQuote",
   "SpeedRun",
   "ReplayQuote",

src/events.test.ts

@@ -50,6 +50,15 @@ describe("从事件缝出一行", () => {
     expect(rows.get(1)?.model).toBe("claude-sonnet-4-5");
   });
 
+  /** 跑在 WebSocket 连接上的那一行说得出来：它不能单独中止（`abortable`） */
+  it("开始事件说得出是不是 WebSocket", () => {
+    const rows = new Map<number, RequestRow>();
+    applyEvent(rows, started({ method: "WS", path: "/v1/responses" }));
+    applyEvent(rows, started({ id: 2 }));
+    expect(rows.get(1)?.ws).toBe(true);
+    expect(rows.get(2)?.ws).toBeUndefined();
+  });
+
   /** 会话是网关在开始的那一刻定的：正在跑的那一条已经在它的会话里，不用等落库 */
   it("开始就带着会话", () => {
     const rows = new Map<number, RequestRow>();

src/generated/tw-api.ts

@@ -1,6 +1,18 @@
 // Generated by tw-api (`tw_api::ts::export_all`). Do not edit by hand.
 
-export const CONTROL_API_VERSION = 42;
+export const CONTROL_API_VERSION = 43;
+
+/**
+ * 手动中止（`POST /request/{id}/abort`、`POST /sessions/{id}/abort`）叫停了哪几个请求。
+ *
+ * **叫停是立刻的，结局随后到**：每个请求照常报一条 [`Event::RequestFailed`]（`source` 是
+ * [`FailureSource::Aborted`]），通常在这条响应之前或紧跟着。叫停之后才跑完的不改。
+ */
+export type Aborted = { 
+/**
+ * 叫停的请求号，开始得早的在前
+ */
+requests: Array<number>, };
 
 /**
  * 一个账号上游登的是哪个账号。
@@ -219,10 +231,10 @@ suggestions: Array<AliasSuggestion>, };
 /**
  * 尝试链里一跳的结果。
  */
-export type AttemptOutcome = "served" | "status" | "error" | "estimated" | "slow_start";
+export type AttemptOutcome = "served" | "status" | "error" | "estimated" | "idle_timeout" | "aborted";
 
 /**
- * 放弃了的一跳（[`AttemptOutcome::SlowStart`]）上游可能已经收了钱的输入。
+ * 放弃了的一跳（[`AttemptOutcome::IdleTimeout`]）上游可能已经收了钱的输入。
  *
  * 上游在流开头报了的（Anthropic 的 `message_start`）是它报的数；没报的只有 `input`，是网关
  * 估的（`estimated`，和 [`Event::RequestStarted`] 的 `input_estimate` 同一个数）。**输出不知道**：
@@ -266,11 +278,12 @@ outcome: AttemptOutcome,
  */
 status?: number | null, 
 /**
- * `error` 时的说明。和这一跳报给客户端的那条错误是同一句。`slow_start` 时说等了多久
+ * `error` 时的说明。和这一跳报给客户端的那条错误是同一句。`idle_timeout` 时说等了多久，
+ * `aborted` 时是 `gw.request.aborted`
  */
 error?: Msg | null, ms: number, 
 /**
- * 放弃了的这一跳（`slow_start`）上游可能已经收了钱的输入（见 [`AttemptUsage`]）。估不
+ * 放弃了的这一跳（`idle_timeout`）上游可能已经收了钱的输入（见 [`AttemptUsage`]）。估不
  * 出来的（请求解不开）没有。别的结果都没有：接下请求的那一跳的用量在结局里
  */
 usage?: AttemptUsage | null, 
@@ -1501,7 +1514,7 @@ plugin_name: string,
 request_id?: number | null, message: Msg, at_ms: number, } | { "kind": "events_dropped", id: number, count: number, at_ms: number, };
 
 /**
- * 上游失败之后停用多久、流开头最多等多久。和配置的 `failover` 一一对应，
+ * 上游失败之后停用多久、多久没有内容就不再等。和配置的 `failover` 一一对应，
  * 没写的是默认值 —— **界面显示的就是真在用的数**，不是「空 = 默认」。
  */
 export type FailoverView = { 
@@ -1530,13 +1543,10 @@ quota_pause_secs: number,
  */
 rate_limit_max_pause_secs: number, 
 /**
- * 流式回答的开头最多等多少秒
- */
-stream_start_wait_secs: number, 
-/**
- * 等过 `stream_start_wait_secs` 还没有内容就换下一家（最后一家照常等）
+ * 上游多少秒没有内容就不再等它（无响应超时）：从请求发出去起算，每来一段内容重新
+ * 计时。客户端还什么都没收到时换下一家，已经收到一部分的报错收尾
  */
-next_on_slow_start: boolean, 
+idle_timeout_secs: number, 
 /**
  * 一个请求合计最多等多少秒：等密钥的分钟、小时上限空出名额，和等满着（`max_concurrent`）
  * 的上游空出位置，共用这一段。0 是不等
@@ -1547,7 +1557,7 @@ slot_wait_secs: number, };
  * 一个请求失败在哪一方。和 HTTP 响应里的 `x-thinkwatch-error` 同一个词表，
  * 另外多一个 `internal`。
  */
-export type FailureSource = "auth" | "config" | "upstream" | "request" | "rate_limited" | "denied" | "internal";
+export type FailureSource = "auth" | "config" | "upstream" | "request" | "rate_limited" | "denied" | "aborted" | "internal";
 
 /**
  * 新建或修改一个策略组时交过来的定义。
@@ -4935,10 +4945,12 @@ export const ENDPOINTS = {
   TokenRateByProvider: { method: "GET", path: "/token-rate/provider", params: [], format: "json" },
   UpstreamHealth: { method: "GET", path: "/upstreams/health", params: [], format: "json" },
   RequestDetail: { method: "GET", path: "/request/{id}", params: ["id"], format: "json" },
+  AbortRequest: { method: "POST", path: "/request/{id}/abort", params: ["id"], format: "json" },
   Fixture: { method: "GET", path: "/request/{id}/fixture", params: ["id"], format: "text" },
   Sessions: { method: "GET", path: "/sessions", params: [], format: "json" },
   SessionDetail: { method: "GET", path: "/sessions/{id}", params: ["id"], format: "json" },
   SessionTranscript: { method: "GET", path: "/sessions/{id}/transcript", params: ["id"], format: "json" },
+  AbortSession: { method: "POST", path: "/sessions/{id}/abort", params: ["id"], format: "json" },
   SpeedQuote: { method: "POST", path: "/speed/quote", params: [], format: "json" },
   SpeedRun: { method: "POST", path: "/speed/run", params: [], format: "json" },
   ReplayQuote: { method: "POST", path: "/replay/quote", params: [], format: "json" },
@@ -5058,10 +5070,12 @@ export type Endpoints = {
   TokenRateByProvider: { req: Window; res: Array<TokenRateView> };
   UpstreamHealth: { req: Window; res: UpstreamHealth };
   RequestDetail: { req: null; res: RequestDetail };
+  AbortRequest: { req: null; res: Aborted };
   Fixture: { req: null; res: string };
   Sessions: { req: ListQuery; res: Array<SessionView> };
   SessionDetail: { req: null; res: SessionDetail };
   SessionTranscript: { req: TranscriptQuery; res: Transcript };
+  AbortSession: { req: null; res: Aborted };
   SpeedQuote: { req: SpeedRunRequest; res: SpeedQuote };
   SpeedRun: { req: SpeedRunRequest; res: Array<SpeedResult> };
   ReplayQuote: { req: ReplayRequest; res: ReplayQuote };