dsh-browser-plus 0.5.0 → 0.5.1

package/src/browser-electron/host-main.ts

@@ -9,7 +9,8 @@
  *   <- { id, op: 'ping' } | { id, op: 'createView', viewId, key?, label? } |
  *      { id, op: 'destroyView', viewId } | { id, op: 'showView', viewId } |
  *      { id, op: 'label', viewId, label } | { id, op: 'listWindows' } |
- *      { id, op: 'command', viewId, method, params }
+ *      { id, op: 'command', viewId, method, params } |
+ *      { id, op: 'printToPdf', viewId, options }
  *   -> { id, ok: true, result? } | { id, ok: false, err }
  *
  * The parent never parses stderr, so diagnostics may go there freely.
@@ -28,7 +29,7 @@ import { taskSummaryUrl } from './task-summary.js'
 import { taskThumbnailDataUrl, type ThumbnailImage } from './task-thumbnail.js'
 import { exportCookiesForAuth, selectCookiesForClear } from './auth-cookies.js'
 import { resolveBrowserIconPath } from './icon.js'
-import { createBootstrap, createPatch, type ChromeBookmark, type ChromePatchOperation, type ChromeTabSummary, type ChromeTaskSummary, type ChromeTrailEntry, type ChromeWorkspaceState } from './chrome-state.js'
+import { createBootstrap, createPatch, type ChromeBookmark, type ChromePatchOperation, type ChromeTabSummary, type ChromeTaskSummary, type ChromeTaskTodo, type ChromeTrailEntry, type ChromeWorkspaceState } from './chrome-state.js'
 
 // Isolate this host's profile from the DSH app's default Electron userData:
 // several Electron instances sharing Roaming\Electron fight over the GPU
@@ -274,6 +275,13 @@ function saveBookmarksToDisk(): void {
  */
 let chromeBookmarkBar = false
 
+/**
+ * Where the human parked the floating orb. Remembered in the chrome prefs file
+ * rather than in the page: a page-side position is per ORIGIN and would jump
+ * back to the corner on the next site.
+ */
+let chromeOrbPosition: { x: number; y: number } | undefined
+
 let chromePrefsFile: string | undefined
 
 function loadPrefsFromDisk(): void {
@@ -282,12 +290,24 @@ function loadPrefsFromDisk(): void {
     const parsed: unknown = JSON.parse(readFileSync(chromePrefsFile, 'utf8'))
     if (typeof parsed !== 'object' || parsed === null) return
     chromeBookmarkBar = (parsed as { bookmarkBar?: unknown }).bookmarkBar === true
+    const orb = (parsed as { orbPosition?: unknown }).orbPosition
+    if (typeof orb === 'object' && orb !== null) {
+      const { x, y } = orb as { x?: unknown; y?: unknown }
+      if (typeof x === 'number' && typeof y === 'number' && Number.isFinite(x) && Number.isFinite(y)) {
+        chromeOrbPosition = { x: Math.round(x), y: Math.round(y) }
+      }
+    }
   } catch { /* first run, or an unreadable file */ }
 }
 
 function savePrefsToDisk(): void {
   if (chromePrefsFile === undefined) return
-  try { writeFileSync(chromePrefsFile, JSON.stringify({ bookmarkBar: chromeBookmarkBar }, null, 2), 'utf8') } catch { /* read-only profile */ }
+  try {
+    writeFileSync(chromePrefsFile, JSON.stringify({
+      bookmarkBar: chromeBookmarkBar,
+      ...chromeOrbPosition !== undefined ? { orbPosition: chromeOrbPosition } : {},
+    }, null, 2), 'utf8')
+  } catch { /* read-only profile */ }
 }
 
 /** Only these raster types are admitted; anything else keeps the letter fallback. */
@@ -486,6 +506,25 @@ const dialogPolicies = new Map<string, { behavior: 'accept' | 'dismiss'; promptT
  * browser_network. A ring (not a stream) on purpose: the agent asks after the fact,
  * and an unbounded log would grow for the life of the tab.
  */
+/**
+ * The slice of Electron's `printToPDF` surface this host uses, named locally.
+ *
+ * Electron is an optional dependency, so CI (npm ci --omit=optional) has neither
+ * its types nor the global `Electron` namespace — a reference to either one fails
+ * the build there while passing locally.
+ */
+interface PrintToPdfOptions {
+  landscape?: boolean
+  printBackground?: boolean
+  pageSize?: { width: number; height: number }
+  margins?: { top?: number; bottom?: number; left?: number; right?: number }
+  scaleFactor?: number
+}
+
+interface PrintToPdfHost {
+  printToPDF: (options: PrintToPdfOptions) => Promise<Buffer>
+}
+
 interface ConsoleEntry { level: string; text: string; at: string }
 interface NetworkEntry { method: string; url: string; status?: number; mime?: string; kind?: string; failed?: string; ms?: number; at: string }
 const CONSOLE_CAP = 200
@@ -507,6 +546,13 @@ const taskViewIds = new Map<string, Set<string>>()
 const taskThumbnails = new Map<string, string>()
 const taskThumbnailVersions = new Map<string, number>()
 const taskStates = new Map<string, HostTaskState>()
+/**
+ * The Agent's todo list per task, mirrored from the DSH session projection by the
+ * parent (see `setTaskTodos`). Kept here so the visible surface can be handed the
+ * current plan the moment it becomes visible — the list itself never rides in a
+ * task summary, because summaries reach every page's main world.
+ */
+const taskTodos = new Map<string, readonly ChromeTaskTodo[]>()
 const thumbnailTimers = new Map<string, ReturnType<typeof setTimeout>>()
 const thumbnailDirty = new Set<string>()
 let thumbnailCaptureInFlight = false
@@ -599,6 +645,17 @@ function makeWindow(): BrowserWindow {
  * the page and swallow clicks in the top 84px (the chrome is drawn there today).
  */
 const CHROME_FRAME_HEIGHT = 84
+/** Height of the bookmark bar when it is shown. It becomes part of the chrome frame. */
+const BOOKMARK_BAR_HEIGHT = 34
+/**
+ * How tall the chrome frame is right now: tab strip + toolbar, plus the bookmark
+ * bar when it is on. The page view starts below ALL of it, so a site's
+ * `position: fixed` header can never end up hidden under the bar — the same
+ * reason the toolbar itself lives in a view instead of a page padding.
+ */
+function chromeFrameHeight(): number {
+  return CHROME_FRAME_HEIGHT + (chromeBookmarkBar ? BOOKMARK_BAR_HEIGHT : 0)
+}
 let chromeFrame: WebContentsView | undefined
 /** Why the frame view could not be created, if it could not. Published to the chrome. */
 let chromeFrameError = ''
@@ -685,7 +742,7 @@ function handleChromeAction(view: WebContentsView, viewId: string, chromeToken:
   const binding = (params ?? {}) as { name?: unknown; payload?: unknown }
   if (binding.name === '__dshBrowserTaskAction' && typeof binding.payload === 'string') {
     try {
-      const action = JSON.parse(binding.payload) as { type?: unknown; taskKey?: unknown; tabId?: unknown; tasks?: unknown; trail?: unknown; control?: unknown; factor?: unknown; url?: unknown; title?: unknown; visible?: unknown; tabs?: unknown; action?: unknown; id?: unknown; open?: unknown; left?: unknown; width?: unknown; toIndex?: unknown; cookies?: unknown; pinned?: unknown }
+      const action = JSON.parse(binding.payload) as { type?: unknown; taskKey?: unknown; tabId?: unknown; tasks?: unknown; trail?: unknown; control?: unknown; factor?: unknown; url?: unknown; title?: unknown; visible?: unknown; tabs?: unknown; action?: unknown; id?: unknown; open?: unknown; left?: unknown; width?: unknown; toIndex?: unknown; cookies?: unknown; pinned?: unknown; x?: unknown; y?: unknown }
       // Authenticate before acting: only our injected chrome knows this
       // view's token, so a forged payload never reaches the dispatcher.
       if (!authorizeChromeAction(action, chromeToken)) return
@@ -699,7 +756,17 @@ function handleChromeAction(view: WebContentsView, viewId: string, chromeToken:
         && typeof action.trail === 'boolean') {
         workspacePanels = { tasks: action.tasks, trail: action.trail }
         if (workspacePanels.tasks && visibleTaskKey !== undefined) scheduleVisibleTaskThumbnail(visibleTaskKey)
+        // Opening the panel is the moment the other rows become visible, so hand
+        // the surface every image we already have for them.
+        if (workspacePanels.tasks) { pushCachedTaskThumbnails(); pushCachedTaskTodos() }
         queueChromePatch({ op: 'panels.set', panels: workspacePanels })
+      } else if (action.type === 'orb-move'
+        && typeof action.x === 'number'
+        && typeof action.y === 'number'
+        && Number.isFinite(action.x)
+        && Number.isFinite(action.y)) {
+        chromeOrbPosition = { x: Math.round(action.x), y: Math.round(action.y) }
+        savePrefsToDisk()
       } else if (action.type === 'set-control-owner'
         && typeof action.taskKey === 'string'
         && (action.control === 'agent' || action.control === 'human')
@@ -727,6 +794,9 @@ function handleChromeAction(view: WebContentsView, viewId: string, chromeToken:
         // The bar is a profile-wide preference, so the host owns it and the
         // chrome reads it back from the bootstrap / patch stream.
         chromeBookmarkBar = action.visible
+    // 书签栏现在占的是帧视图的高度，所以开关一变就得重排：
+    // 只发 patch 的话，帧还是 84px 高、页面视图还从 84 开始 —— 书签栏就会压住页面。
+    layoutViews()
         savePrefsToDisk()
         queueChromePatch({ op: 'bookmarkbar.set', visible: chromeBookmarkBar })
       } else if (action.type === 'bookmark-remove' && typeof action.url === 'string') {
@@ -903,7 +973,7 @@ function ensureChromeFrame(): void {
     // did nothing (its binding was never installed).
     try { frame.webContents.debugger.attach('1.3') } catch { /* already attached */ }
     win.contentView.addChildView(frame)
-    frame.setBounds({ x: 0, y: 0, width: win.getContentSize()[0] ?? 0, height: CHROME_FRAME_HEIGHT })
+    frame.setBounds({ x: 0, y: 0, width: win.getContentSize()[0] ?? 0, height: chromeFrameHeight() })
     // HIDDEN until the frame's chrome can drive the page. A visible frame sits on
     // top of the page view, so a person's clicks land on it — and the frame's
     // chrome can only do the things it can do *in its own document* today: its
@@ -955,11 +1025,12 @@ function layoutViews(): void {
     try {
       // Below the chrome frame: the page viewport is genuinely smaller now, so a
       // sticky/fixed header lands at the top of the page instead of under the chrome.
+      const frameHeight = chromeFrameHeight()
       entry.webContentsView.setBounds({
         x: 0,
-        y: CHROME_FRAME_HEIGHT,
+        y: frameHeight,
         width: width ?? 0,
-        height: Math.max(0, (height ?? 0) - CHROME_FRAME_HEIGHT),
+        height: Math.max(0, (height ?? 0) - frameHeight),
       })
     } catch { /* destroyed */ }
   }
@@ -975,7 +1046,7 @@ function raiseChromeFrame(): void {
   try { win.contentView.removeChildView(chromeFrame) } catch { /* not attached */ }
   try {
     win.contentView.addChildView(chromeFrame)
-    chromeFrame.setBounds({ x: 0, y: 0, width: win.getContentSize()[0] ?? 0, height: CHROME_FRAME_HEIGHT })
+    chromeFrame.setBounds({ x: 0, y: 0, width: win.getContentSize()[0] ?? 0, height: chromeFrameHeight() })
   } catch { /* destroyed */ }
 }
 
@@ -1282,6 +1353,7 @@ function chromeWorkspaceState(selectedTaskKey = visibleTaskKey): ChromeWorkspace
     trail: activeTraceForTask(selectedTaskKey),
     bookmarks: chromeBookmarks,
     bookmarkBar: chromeBookmarkBar,
+    ...chromeOrbPosition !== undefined ? { orbPosition: chromeOrbPosition } : {},
     frameError: chromeFrameError,
     windowProbe: (() => {
       const win = window
@@ -1313,6 +1385,9 @@ function chromePatchScript(operations: readonly ChromePatchOperation[]): string
   const patch = createPatch(chromeEpoch, ++chromeRevision, operations)
   return ';window.__dshChromePatch = ' + JSON.stringify(patch)
     + ';try { window.__dshChromeApply?.(window.__dshChromePatch) } catch {}'
+    // 补丁里含 Agent 的计划（task.todos），而默认主世界下页面脚本读得到 window ——
+    // 应用完立刻把这份明文抹掉，别让它一直挂在页面上。
+    + ';try { delete window.__dshChromePatch } catch { window.__dshChromePatch = undefined }'
 }
 
 
@@ -1405,6 +1480,50 @@ function queueFrameAddress(): void {
   runChromeScript(frame, `window.__dshChromeAddress && window.__dshChromeAddress(${JSON.stringify(url)})`)
 }
 
+/**
+ * Replay every task's last known image to the visible surface.
+ *
+ * A task's picture never travels in the task summaries: those are injected into
+ * every visited page's main world, and the picture is screen content. So a
+ * chrome surface only ever learns an image from a targeted 'task.thumbnail'
+ * patch, and the capture path queues those for the VISIBLE task alone. Without
+ * this replay, the surface that shows the panel paints its own task's image and
+ * a placeholder for every other one — even for tasks the human already looked
+ * at, which is exactly what the panel looked like before this existed.
+ *
+ * Safe by construction: the patches go to {@link chromeSurfaces}, i.e. the
+ * visible task's page and the chrome frame, which are the only surfaces that
+ * can show the panel in the first place.
+ */
+/** Store one task's plan and push it to the surface that shows the panel. */
+function updateTaskTodos(key: string, todos: unknown): void {
+  if (!Array.isArray(todos)) return
+  const clean: ChromeTaskTodo[] = []
+  for (const item of todos) {
+    const entry = item as { content?: unknown; status?: unknown }
+    if (typeof entry?.content !== 'string' || entry.content === '') continue
+    const status = entry.status === 'completed' || entry.status === 'in_progress' ? entry.status : 'pending'
+    clean.push({ content: entry.content.slice(0, 160), status })
+    if (clean.length >= 40) break
+  }
+  taskTodos.set(key, clean)
+  queueChromePatch({ op: 'task.todos', key, todos: clean })
+}
+
+/**
+ * Hand every known plan to the visible surface. The orb is always on screen, so
+ * a surface that has just become visible has never seen any of them.
+ */
+function pushCachedTaskTodos(): void {
+  for (const [key, todos] of taskTodos) queueChromePatch({ op: 'task.todos', key, todos })
+}
+
+function pushCachedTaskThumbnails(): void {
+  for (const [key, dataUrl] of taskThumbnails) {
+    queueChromePatch({ op: 'task.thumbnail', key, version: taskThumbnailVersions.get(key) ?? 0, dataUrl })
+  }
+}
+
 function scheduleVisibleTaskThumbnail(taskKey: string, delayMs = 360): void {
   if (taskKey !== visibleTaskKey) return
   thumbnailDirty.add(taskKey)
@@ -1494,6 +1613,10 @@ function switchVisibleTask(taskKey: string): void {
   // changed with it. Queued after the bootstrap: resetChromeDelivery() inside
   // pushVisibleChromeState() drops anything queued before it.
   queueTabsSet()
+  // 新露出来的那个表面此前没收到过别的任务的图像（补丁只发给可见表面），所以把
+  // 宿主手里已有的图都补给它 —— 否则切过去之后面板里除了当前任务全是占位符。
+  pushCachedTaskThumbnails()
+  pushCachedTaskTodos()
   // 必须在 pushVisibleChromeState() **之后**排队：它里面的 resetChromeDelivery() 会把
   // 先排的东西丢掉（切标签就是走这条路）。导航走 applyPageChrome，不会播。
   queueChromePatch({ op: 'reveal' })
@@ -1575,6 +1698,10 @@ function applyPageChrome(view: WebContentsView, viewId: string): void {
     // (whose own chrome is not on screen but whose title belongs in the strip).
     queueTabsSet()
     scheduleVisibleTaskThumbnail(taskKey, 550)
+    // 每一次真导航都会重跑整个 chrome 脚本，页面侧那份 todo 缓存跟着归零。
+    // 缩略图有 scheduleVisibleTaskThumbnail 兜住，计划没有对应物 —— 漏了它，
+    // Agent 一导航球上就只剩状态圆点（这是本轮功能的卖点场景，必现）。
+    pushCachedTaskTodos()
   }
 }
 
@@ -1642,7 +1769,7 @@ function authorizeChromeAction(action: unknown, token: string): boolean {
 }
 
 /** Handle one command. */
-async function handle(op: string, msg: { id: number; viewId?: string; method?: string; params?: Record<string, unknown>; expression?: string; url?: string; savePath?: string; cookies?: unknown[]; entry?: unknown; key?: string; label?: string; task?: Record<string, unknown>; domain?: string; name?: string; all?: boolean; behavior?: string; promptText?: string; clear?: boolean }): Promise<void> {
+async function handle(op: string, msg: { id: number; viewId?: string; method?: string; params?: Record<string, unknown>; expression?: string; url?: string; savePath?: string; cookies?: unknown[]; entry?: unknown; key?: string; label?: string; task?: Record<string, unknown>; todos?: unknown; domain?: string; name?: string; all?: boolean; behavior?: string; promptText?: string; clear?: boolean; options?: Record<string, unknown> }): Promise<void> {
   try {
     switch (op) {
       case 'ping':
@@ -1667,6 +1794,7 @@ async function handle(op: string, msg: { id: number; viewId?: string; method?: s
             const trail = activeTraceForTask(entryView.taskKey).at(-1)
             if (trail !== undefined) operations.push({ op: 'trail.append', taskKey: entryView.taskKey, entry: trail })
             scheduleVisibleTaskThumbnail(entryView.taskKey)
+            pushCachedTaskTodos()
           }
           if (operations.length > 0) queueChromePatch(...operations)
         }
@@ -1918,6 +2046,7 @@ async function handle(op: string, msg: { id: number; viewId?: string; method?: s
             if (thumbnailTimer !== undefined) clearTimeout(thumbnailTimer)
             thumbnailTimers.delete(entry.taskKey)
             taskThumbnails.delete(entry.taskKey)
+            taskTodos.delete(entry.taskKey)
             taskThumbnailVersions.delete(entry.taskKey)
             thumbnailDirty.delete(entry.taskKey)
             thumbnailLastCapturedAt.delete(entry.taskKey)
@@ -1984,8 +2113,16 @@ async function handle(op: string, msg: { id: number; viewId?: string; method?: s
         const expression = msg.expression
         if (typeof expression !== 'string') throw new Error('chromeEval: expression must be a string')
         try { frame.webContents.debugger.attach('1.3') } catch { /* already attached */ }
+        // 隔离模式（默认）下，帧里的 chrome 也住在它自己的世界 —— 读它必须带上下文，
+        // 否则读的是帧的主世界（那里什么都没有）。
+        const frameContextId = CHROME_WORLD === 'isolated' ? await ensureChromeContext(frame) : undefined
         frame.webContents.debugger
-          .sendCommand('Runtime.evaluate', { expression, returnByValue: true, awaitPromise: true })
+          .sendCommand('Runtime.evaluate', {
+            expression,
+            returnByValue: true,
+            awaitPromise: true,
+            ...frameContextId === undefined ? {} : { contextId: frameContextId },
+          })
           .then(
             (result: unknown) => {
               // The client resolves with `msg.result`, so the value goes in there.
@@ -2014,6 +2151,19 @@ async function handle(op: string, msg: { id: number; viewId?: string; method?: s
         reply(msg.id, { ok: true })
         return
       }
+      case 'focusWindow': {
+        const viewId = msg.viewId
+        if (viewId === undefined) throw new Error('focusWindow missing viewId')
+        if (!views.has(viewId)) throw new Error(`focusWindow: unknown view ${viewId}`)
+        const win = ensureWindow()
+        // The same three steps capture uses: a minimized window has to be
+        // restored, and a window behind another one only comes forward on focus.
+        try { if (!win.isVisible()) win.show() } catch { /* closing */ }
+        try { win.restore() } catch { /* not minimized */ }
+        try { win.focus() } catch { /* closing */ }
+        reply(msg.id, { ok: true })
+        return
+      }
       case 'label': {
         const viewId = msg.viewId
         const label = msg.label
@@ -2053,6 +2203,13 @@ async function handle(op: string, msg: { id: number; viewId?: string; method?: s
         reply(msg.id, { ok: true, result: { task: task ?? null } })
         return
       }
+      case 'setTaskTodos': {
+        const key = msg.key
+        if (typeof key !== 'string' || msg.todos === undefined) throw new Error('setTaskTodos missing key or todos')
+        updateTaskTodos(key, msg.todos)
+        reply(msg.id, { ok: true, result: {} })
+        return
+      }
       case 'updateTask': {
         const key = msg.key
         if (typeof key !== 'string' || msg.task === undefined) throw new Error('updateTask missing key or task')
@@ -2073,6 +2230,28 @@ async function handle(op: string, msg: { id: number; viewId?: string; method?: s
         reply(msg.id, { ok: true, result })
         return
       }
+      case 'chromeWorldEval': {
+        // Run a snippet in the world the *page's* injected chrome lives in.
+        //
+        // In isolated mode the chrome's globals are not reachable from the page
+        // world, and creating a world with the same name from outside yields a
+        // different, empty context — so the host, which owns the real one, has to
+        // resolve it here. In main mode this is a plain page-world evaluate.
+        const viewId = msg.viewId
+        if (viewId === undefined) throw new Error('chromeWorldEval missing viewId')
+        const entry = views.get(viewId)
+        if (entry === undefined) throw new Error(`chromeWorldEval: unknown view ${viewId}`)
+        if (typeof msg.expression !== 'string') throw new Error('chromeWorldEval missing expression')
+        const contextId = CHROME_WORLD === 'isolated' ? await ensureChromeContext(entry.webContentsView) : undefined
+        const result = await entry.webContentsView.webContents.debugger.sendCommand('Runtime.evaluate', {
+          expression: msg.expression,
+          returnByValue: true,
+          awaitPromise: true,
+          ...contextId === undefined ? {} : { contextId },
+        })
+        reply(msg.id, { ok: true, result })
+        return
+      }
       case 'capture': {
         const viewId = msg.viewId
         if (viewId === undefined) throw new Error('capture missing viewId')
@@ -2140,6 +2319,23 @@ async function handle(op: string, msg: { id: number; viewId?: string; method?: s
         reply(msg.id, { ok: true, result: { base64, width: 0, height: 0 } })
         return
       }
+      case 'printToPdf': {
+        const viewId = msg.viewId
+        if (viewId === undefined) throw new Error('printToPdf missing viewId')
+        const entry = views.get(viewId)
+        if (entry === undefined) throw new Error(`printToPdf: unknown view ${viewId}`)
+        // Electron's debugger does not implement CDP `Page.printToPDF`; this
+        // native path is the one that actually produces a document.
+        const options = (msg.options ?? {}) as PrintToPdfOptions
+        // `printToPDF` and its option type are Electron's; Electron is an optional
+        // dependency (CI installs with --omit=optional, so neither the type nor the
+        // `Electron` namespace exists there). Name the shape locally instead of
+        // reaching for the namespace, and let this child — which only ever runs
+        // inside Electron — be the one that calls it.
+        const data = await (entry.webContentsView.webContents as unknown as PrintToPdfHost).printToPDF(options)
+        reply(msg.id, { ok: true, result: { base64: data.toString('base64') } })
+        return
+      }
       case 'download': {
         const viewId = msg.viewId
         const url = msg.url
@@ -2303,7 +2499,7 @@ void app.whenReady().then(() => {
   rl.on('line', line => {
     const text = line.trim()
     if (text === '') return
-    let msg: { id: number; op?: string; viewId?: string; method?: string; params?: Record<string, unknown>; expression?: string; url?: string; savePath?: string; cookies?: unknown[]; key?: string; label?: string; task?: Record<string, unknown>; domain?: string; name?: string; all?: boolean }
+    let msg: { id: number; op?: string; viewId?: string; method?: string; params?: Record<string, unknown>; expression?: string; url?: string; savePath?: string; cookies?: unknown[]; key?: string; label?: string; task?: Record<string, unknown>; todos?: unknown; domain?: string; name?: string; all?: boolean; options?: Record<string, unknown> }
     try {
       msg = JSON.parse(text) as typeof msg
     } catch {