dsh-client-auto-continue 0.7.5 → 0.8.0

package/README.md

@@ -31,7 +31,7 @@
 
 ## What It Does
 
-For [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) (`dsh web`): whenever a request in the web GUI gets interrupted by a **non-human cause**, the plugin simulates the user typing **「继续」** and sends it, so the agent keeps working without manual intervention. The message enters the session log exactly like a manual prompt — the model sees it, and the interrupted work resumes.
+For [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) (`dsh web`): whenever a request in the web GUI gets interrupted by a **non-human cause**, the plugin simulates the user typing **「继续」** and sends it, so the agent keeps working without manual intervention. The message enters the session log exactly like a manual prompt — the model sees it, and the interrupted work resumes. Since 0.8.0 the engine runs **inside the host process** (single instance), so it keeps watching even with every browser tab closed, and multiple open tabs can never double-send.
 
 ![demo](docs/demo.svg)
 
@@ -62,15 +62,11 @@ It watches the live event streams and reacts to:
 
 ## How It Works
 
-The plugin opens two extra SSE streams in the browser — `events.mux` (session events) and `events.host` (host events). The host supports multiple consumers, so this never interferes with the built-in runtime. On an interruption it waits a **grace period** (default 3 s) — if the host starts a new turn by itself (`turn/start`), the auto-continue is cancelled — then calls `sessions.prompt` in `queue` mode with the configured text.
+The host-side engine subscribes to the session event firehose inside the dsh host process — exactly one engine, regardless of how many tabs are open (the duplicate-send class of bugs cannot exist by construction). On an interruption it waits a **grace period** (default 3 s) — if the host starts a new turn by itself (`turn/start`), the auto-continue is cancelled — then sends the configured text through the agent registry (`agent.followup`, the same queue the Send button uses).
 
-On page load / reconnect it also scans the most recently updated sessions: a session whose last turn ended with a non-human reason **within the scan window** (default 15 minutes), with no later `turn/start` or user message, gets resumed automatically too (e.g. the host crashed while the browser was closed). Before every send the plugin also asks the host one authoritative question — is the latest session event already this exact message? If the same text is still queued (the last event is that `user/message`), it skips; once the message has been processed (the last event is a turn end or anything else), a fresh send is allowed, so legitimate consecutive resumes are never blocked.
+On host boot it also scans the live sessions: a session whose last turn ended with a non-human reason **within the scan window** (default 15 minutes), with no later `turn/start` or user message, gets resumed automatically too (e.g. the host crashed while the browser was closed — the agent-loop resumes the session and the engine picks it up).
 
-With the page open in several tabs, a cross-tab atomic send lock (Web Locks API with a mutex-stamp fallback), a shared per-session cooldown stamp, and a shared send counter that hard-caps consecutive sends at `maxConsecutive` guarantee exactly one tab sends — no duplicated 「继续」, even when several tabs watch the same session.
-
-All knobs live in the plugin's settings card — see [Configuration](#configuration).
-
----
+The browser half is a thin shell: the settings card, plus a status bridge that shows notifications (with Resume now / Pause this session 1h buttons, routed back to the host engine) and feeds the card's stats / paused-sessions panels.
 
 ## Quick Start
 
@@ -214,8 +210,8 @@ auto-continue:
 | Scan on load / reconnect | `on` | Scan recently interrupted sessions on load / reconnect |
 | Scan limit | `8` | Max sessions scanned (running / subagent sessions excluded) |
 | Scan window (ms) | `900000` | Scan only considers interruptions inside this window |
-| Reconnect scan delay (ms) | `5000` | Delay before scanning after a reconnect |
-| Reconnect backoff (ms) | `3000` | SSE reconnect backoff |
+| Reconnect scan delay (ms) | `5000` | Legacy — unused since the engine moved into the host (kept for config compatibility) |
+| Reconnect backoff (ms) | `3000` | Legacy — unused since the engine moved into the host (kept for config compatibility) |
 | Verbose logs | `on` | `[auto-continue]` console logs |
 | Classify errors | `on` | Auto-resume transient failures only; auth / balance / model errors are skipped and notified |
 | Backoff factor | `2` | Cooldown multiplier per consecutive failure (2 = 20s → 40s → 80s…) |
@@ -232,7 +228,7 @@ The plugin is browser-only and touches **no files, credentials, or network beyon
 
 - It opens the same two read-only event streams the web UI already uses (no extra server, no third-party endpoints)
 - The engine's **only automatic write** is `sessions.prompt` — the same call the Send button makes — with the text you configured (saving the settings card writes the `auto-continue` section of `~/.dsh/settings.yaml` through the normal settings API, exactly like any other setting)
-- Browser storage is limited to small `localStorage` keys: cross-tab send locks and counts (a hard cap that stops duplicate sends even with several tabs open), per-session pauses, and the daily stats counters
+- No browser storage at all: the single host-side engine keeps its cooldowns, send caps, pauses and stats in process memory
 - Browser notifications are opt-in (`notify` setting) and permission is requested on first use only
 
 ---
@@ -243,7 +239,7 @@ The plugin is browser-only and touches **no files, credentials, or network beyon
 npm run typecheck   # tsc --noEmit
 npm run build       # lib/client.js + lib/index.js + lib/types
 npm run watch       # rebuild on change; host HMR hot-reloads without a page refresh
-npm run test        # node tests/simulate.mjs — 43 behavioral scenarios
+npm run test        # node tests/simulate-host.mjs — 15 host-side behavioral scenarios
 ```
 
 While `npm run watch` runs, the profile's client-hmr row polls `lib/client.js` every 500 ms and hot-reloads the plugin in the browser — no server restart needed for code changes.

package/README.zh.md

@@ -31,7 +31,7 @@
 
 ## 它做什么
 
-适用于 [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness)(`dsh web`): 当 webui 里的请求因为**非人为因素**中断时, 插件模拟用户输入 **「继续」** 并自动发送, 让 Agent 继续干活, 无需手动干预。消息与手动输入完全等价——进入会话日志、对模型可见, 中断的任务随即恢复。
+适用于 [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness)(`dsh web`): 当 webui 里的请求因为**非人为因素**中断时, 插件模拟用户输入 **「继续」** 并自动发送, 让 Agent 继续干活, 无需手动干预。消息与手动输入完全等价——进入会话日志、对模型可见, 中断的任务随即恢复。自 0.8.0 起引擎跑在**宿主进程内**(单实例): 浏览器标签全关掉也在值守, 多标签页同时打开也不可能重复发送。
 
 ![demo](docs/demo-zh.svg)
 
@@ -62,15 +62,11 @@
 
 ## 工作原理
 
-插件在浏览器里额外打开两条 SSE 流——`events.mux`(会话事件)与 `events.host`(宿主事件)。宿主支持多消费者, 与内置运行时互不干扰。检测到中断后先等待一个**宽限期**(默认 3 秒)——若宿主自行开启了新回合(`turn/start`), 自动继续即取消——然后以 `queue` 模式调用 `sessions.prompt` 发送配置的文本。
+宿主侧引擎在 dsh 宿主进程内订阅会话事件 firehose——**全局只有一个引擎**, 无论开多少个标签页(重复发送这类 bug 从构造上不可能存在)。检测到中断后先等待一个**宽限期**(默认 3 秒)——若宿主自行开启了新回合(`turn/start`), 自动继续即取消——然后经 agent 注册表(`agent.followup`, 与「发送」按钮同一个排队通道)发送配置的文本。
 
-页面启动 / 重连时, 插件还会扫描最近更新的会话: 若某个会话的最后一个回合在**扫描时间窗**(默认 15 分钟)内以非人为原因结束, 且之后没有新的 `turn/start` 或用户消息, 也会被自动续跑(例如浏览器关闭期间宿主崩溃的情况)。每次发送前插件还会问宿主一个权威问题——会话的**最后一条事件**是否正是这条消息: 若相同文本还在排队(最后一条就是那条 `user/message`)则跳过; 一旦该消息已被处理(最后一条是回合结束或其他事件)就允许再发, 因此合法的连续续跑不会被误挡。
+宿主启动时还会扫描存活会话: 最后一个回合在**扫描时间窗**(默认 15 分钟)内以非人为原因结束、且其后没有新回合或用户消息的会话, 会被自动续跑(例如浏览器关闭期间宿主崩溃——agent-loop 恢复会话后引擎接着接手)。
 
-多个标签页同时打开时, 通过跨标签页原子发送锁(Web Locks API, 旧环境回退互斥戳)+ 共享的每会话冷却记录 + 以 `maxConsecutive` 为硬上限的共享发送计数, 保证**只有一个标签页发送**——即使多个标签页盯着同一会话, 也不会重复出现两条「继续」。
-
-所有参数都在插件的设置卡片中调整——见 [配置](#配置)。
-
----
+浏览器半侧是瘦壳: 设置卡片 + 一条状态桥(展示通知, 带「立即续跑 / 暂停该会话 1 小时」按钮并把动作回传给宿主引擎; 驱动卡片里的统计与暂停面板)。
 
 ## 快速开始
 
@@ -214,8 +210,8 @@ auto-continue:
 | 启动/重连扫描 | 开 | 页面启动 / 重连时扫描最近中断的会话 |
 | 扫描会话数 | `8` | 扫描最多检查的会话数(不含运行中 / 子代理会话) |
 | 扫描时间窗 (ms) | `900000` | 扫描只处理该时间窗内的中断 |
-| 重连扫描延迟 (ms) | `5000` | 重连后等待宿主恢复再扫描 |
-| 重连退避 (ms) | `3000` | SSE 流断开后的重连间隔 |
+| 重连扫描延迟 (ms) | `5000` | 遗留字段 — 引擎迁入宿主后不再使用(仅为配置兼容保留) |
+| 重连退避 (ms) | `3000` | 遗留字段 — 引擎迁入宿主后不再使用(仅为配置兼容保留) |
 | 详细日志 | 开 | 控制台输出 `[auto-continue]` 日志 |
 | 错误分类 | 开 | 仅自动恢复临时性错误; 认证 / 余额 / 模型等永久性错误跳过并通知 |
 | 退避系数 | `2` | 连续失败时冷却间隔的倍率(2 = 20s → 40s → 80s…) |
@@ -232,7 +228,7 @@ auto-continue:
 
 - 只复用 webui 本身就在用的两条只读事件流(无额外服务、无第三方端点)
 - 引擎**唯一会自动执行的写入**是 `sessions.prompt`——与点「发送」按钮完全相同的调用, 内容为你配置的文本(设置卡片里保存配置会通过常规设置 API 写入 `~/.dsh/settings.yaml` 的 `auto-continue` 段落, 与任何其他设置一样)
-- 浏览器存储仅限于少量 `localStorage` 键: 跨标签页发送锁与发送计数(多标签页同时打开时也绝不重复刷屏的硬上限)、会话级暂停、每日统计计数
+- 不使用任何浏览器存储: 宿主侧单实例引擎的冷却、发送上限、暂停与统计全部保存在进程内存里
 - 浏览器通知是可选开启的(`notify` 设置), 仅在首次使用时请求一次权限
 
 ---
@@ -243,7 +239,7 @@ auto-continue:
 npm run typecheck   # tsc --noEmit
 npm run build       # lib/client.js + lib/index.js + lib/types
 npm run watch       # 监听变更自动重建; 宿主 HMR 免刷新热重载
-npm run test        # node tests/simulate.mjs — 43 个行为场景
+npm run test        # node tests/simulate-host.mjs — 15 个 host 侧行为场景
 ```
 
 `npm run watch` 运行时, profile 的 client-hmr 行每 500ms 轮询 `lib/client.js` 并在浏览器中热重载插件——改代码无需重启服务。