minimal-agent 0.13.0 → 0.15.0

package/src/tools/webbrowser/browserRunRegistry.ts

@@ -1,12 +1,21 @@
 import { randomUUID } from 'node:crypto';
 
 export type BrowserActionGrant = 'observe' | 'interact' | 'upload' | 'submit';
+export type TabRetention =
+  | 'ephemeral'
+  | 'conversation'
+  | 'handoff'
+  | 'deliverable';
 
-const NEGATED_ACTION_PREFIX = /(?:不要|不需(?:要)?|不能|不可|别|禁止|避免|无需|切勿)(?:再|自动|直接|帮我|替我|进行)?[\s\S]{0,8}$|(?:do\s+not|don't|never|must\s+not|should\s+not|without)(?:\s+\w+){0,4}\s*$/i;
+const NEGATED_ACTION_PREFIX =
+  /(?:不要|不需(?:要)?|不能|不可|别|禁止|避免|无需|切勿)(?:再|自动|直接|帮我|替我|进行)?[\s\S]{0,8}$|(?:do\s+not|don't|never|must\s+not|should\s+not|without)(?:\s+\w+){0,4}\s*$/i;
 
 function hasUnnegatedAction(text: string, pattern: RegExp): boolean {
   for (const match of text.matchAll(pattern)) {
-    const prefix = text.slice(Math.max(0, (match.index ?? 0) - 32), match.index);
+    const prefix = text.slice(
+      Math.max(0, (match.index ?? 0) - 32),
+      match.index,
+    );
     if (!NEGATED_ACTION_PREFIX.test(prefix)) return true;
   }
   return false;
@@ -19,11 +28,20 @@ function hasUnnegatedAction(text: string, pattern: RegExp): boolean {
  */
 export function inferBrowserTaskGrants(text: string): BrowserActionGrant[] {
   const grants: BrowserActionGrant[] = ['observe', 'interact'];
-  if (hasUnnegatedAction(text, /\b(?:upload|attach)\b(?!\s+(?:page|button|history|settings))|上传(?!页|页面|按钮|记录|历史|设置)|选择文件|添加文件/giu)) grants.push('upload');
-  if (hasUnnegatedAction(
-    text,
-    /\b(?:publish|submit|purchase|buy|delete|place\s+(?:an?\s+)?order|send)\b(?!\s+(?:page|button|history|record|settings))|发布(?!页|页面|按钮|记录|历史|设置)|发表(?!页|页面|按钮|记录|历史|设置)|提交(?!页|页面|按钮|记录|历史|设置)|购买|下单|删除|发送/giu,
-  )) grants.push('submit');
+  if (
+    hasUnnegatedAction(
+      text,
+      /\b(?:upload|attach)\b(?!\s+(?:page|button|history|settings))|上传(?!页|页面|按钮|记录|历史|设置)|选择文件|添加文件/giu,
+    )
+  )
+    grants.push('upload');
+  if (
+    hasUnnegatedAction(
+      text,
+      /\b(?:publish|submit|purchase|buy|delete|place\s+(?:an?\s+)?order|send)\b(?!\s+(?:page|button|history|record|settings))|发布(?!页|页面|按钮|记录|历史|设置)|发表(?!页|页面|按钮|记录|历史|设置)|提交(?!页|页面|按钮|记录|历史|设置)|购买|下单|删除|发送/giu,
+    )
+  )
+    grants.push('submit');
   return grants;
 }
 
@@ -31,8 +49,14 @@ export function inferBrowserTaskGrants(text: string): BrowserActionGrant[] {
 export function isLikelyIrreversibleBrowserTarget(label: string): boolean {
   const normalized = label.replace(/\s+/g, ' ').trim();
   if (!normalized) return false;
-  return /(?:^|[\s"'=:_-])(?:publish|submit|purchase|buy now|place order|delete|confirm delete|send)(?:$|[\s"'=:_-])/i.test(normalized)
-    || /(?:发布|发表|提交|立即购买|确认购买|确认下单|删除|确认删除|发送)(?:$|[\s：:）)】\]])/.test(normalized);
+  return (
+    /(?:^|[\s"'=:_-])(?:publish|submit|purchase|buy now|place order|delete|confirm delete|send)(?:$|[\s"'=:_-])/i.test(
+      normalized,
+    ) ||
+    /(?:发布|发表|提交|立即购买|确认购买|确认下单|删除|确认删除|发送)(?:$|[\s：:）)】\]])/.test(
+      normalized,
+    )
+  );
 }
 
 export interface BrowserRunContext {
@@ -47,6 +71,13 @@ export interface BrowserRunContext {
   allowedUploadRoots: string[];
 }
 
+export interface BrowserTabMetadata {
+  tabId: string;
+  retention: TabRetention;
+  title?: string;
+  updatedAt: number;
+}
+
 type BrowserRunInput = {
   conversationKey: string;
   runId?: string;
@@ -67,14 +98,17 @@ export class BrowserRunRegistry {
   private readonly runs = new Map<string, BrowserRunContext>();
   private readonly conversationTabs = new Map<string, Set<string>>();
   private readonly tabOwners = new Map<string, string>();
+  private readonly tabMetadata = new Map<string, BrowserTabMetadata>();
 
   ensure(input: BrowserRunInput): BrowserRunContext {
     const key = this.key(input.conversationKey, input.runId ?? 'workbench');
     const existing = this.runs.get(key);
     if (existing) {
       if (input.grants) existing.grants = [...new Set(input.grants)];
-      if (input.taskGrantId !== undefined) existing.taskGrantId = input.taskGrantId;
-      if (input.allowedUploadRoots) existing.allowedUploadRoots = [...new Set(input.allowedUploadRoots)];
+      if (input.taskGrantId !== undefined)
+        existing.taskGrantId = input.taskGrantId;
+      if (input.allowedUploadRoots)
+        existing.allowedUploadRoots = [...new Set(input.allowedUploadRoots)];
       return existing;
     }
     const context: BrowserRunContext = {
@@ -82,7 +116,9 @@ export class BrowserRunRegistry {
       runId: input.runId ?? randomUUID(),
       profileId: input.profileId ?? 'default',
       claimedTabIds: [],
-      grants: [...new Set<BrowserActionGrant>(input.grants ?? ['observe', 'interact'])],
+      grants: [
+        ...new Set<BrowserActionGrant>(input.grants ?? ['observe', 'interact']),
+      ],
       allowedUploadRoots: [...new Set(input.allowedUploadRoots ?? [])],
       ...(input.taskGrantId ? { taskGrantId: input.taskGrantId } : {}),
     };
@@ -92,7 +128,9 @@ export class BrowserRunRegistry {
 
   find(conversationKey: string, runId?: string): BrowserRunContext | undefined {
     if (runId) return this.runs.get(this.key(conversationKey, runId));
-    return [...this.runs.values()].find((item) => item.conversationKey === conversationKey);
+    return [...this.runs.values()].find(
+      (item) => item.conversationKey === conversationKey,
+    );
   }
 
   claimedTabs(conversationKey: string): string[] {
@@ -103,13 +141,33 @@ export class BrowserRunRegistry {
     return this.tabOwners.get(tabId);
   }
 
+  markTab(
+    context: BrowserRunContext,
+    tabId: string,
+    metadata: Pick<BrowserTabMetadata, 'retention' | 'title'>,
+  ): BrowserTabMetadata {
+    this.assertTab(context, tabId);
+    const next: BrowserTabMetadata = {
+      tabId,
+      ...metadata,
+      updatedAt: Date.now(),
+    };
+    this.tabMetadata.set(tabId, next);
+    return next;
+  }
+
+  tabInfo(tabId: string): BrowserTabMetadata | undefined {
+    return this.tabMetadata.get(tabId);
+  }
+
   claimTab(context: BrowserRunContext, tabId: string): void {
     const owner = this.tabOwners.get(tabId);
     if (owner && owner !== context.conversationKey) {
       throw new Error('该浏览器标签属于另一个会话，已阻止跨会话操作');
     }
     this.tabOwners.set(tabId, context.conversationKey);
-    const claimed = this.conversationTabs.get(context.conversationKey) ?? new Set<string>();
+    const claimed =
+      this.conversationTabs.get(context.conversationKey) ?? new Set<string>();
     claimed.add(tabId);
     this.conversationTabs.set(context.conversationKey, claimed);
     context.claimedTabIds = [...claimed];
@@ -124,16 +182,34 @@ export class BrowserRunRegistry {
    * 测试里先撞上（基准集跑完不清理，把 browserTabs 的用例挤挂了），但这是生产同款问题。
    */
   forgetConversation(conversationKey: string): void {
-    for (const tabId of this.claimedTabs(conversationKey)) this.releaseTab(conversationKey, tabId);
+    for (const tabId of this.claimedTabs(conversationKey))
+      this.releaseTab(conversationKey, tabId);
     this.conversationTabs.delete(conversationKey);
     for (const [key, run] of [...this.runs.entries()]) {
       if (run.conversationKey === conversationKey) this.runs.delete(key);
     }
   }
 
+  /**
+   * Invalidate every tab lease after the browser process is rebuilt. Tab ids
+   * are process-local and can be reused by the replacement daemon; retaining
+   * old owners would either block the new owner or route it across sessions.
+   * Run contexts remain alive so the active task can claim a replacement tab.
+   */
+  invalidateAllTabs(): void {
+    this.tabOwners.clear();
+    this.conversationTabs.clear();
+    this.tabMetadata.clear();
+    for (const context of this.runs.values()) {
+      context.claimedTabIds = [];
+      context.activeTabId = undefined;
+    }
+  }
+
   releaseTab(conversationKey: string, tabId: string): void {
     if (this.tabOwners.get(tabId) !== conversationKey) return;
     this.tabOwners.delete(tabId);
+    this.tabMetadata.delete(tabId);
     const claimed = this.conversationTabs.get(conversationKey);
     claimed?.delete(tabId);
     if (claimed?.size === 0) this.conversationTabs.delete(conversationKey);

package/src/tools/webbrowser/browserService.js

@@ -5,10 +5,36 @@
  */
 export class BrowserService {
     tails = new Map();
+    async waitForPrevious(previous, signal) {
+        if (!signal)
+            return previous.catch(() => undefined);
+        if (signal.aborted)
+            throw signal.reason instanceof Error
+                ? signal.reason
+                : new Error('Browser action cancelled');
+        await new Promise((resolve, reject) => {
+            let settled = false;
+            const cleanup = () => signal.removeEventListener('abort', onAbort);
+            const finish = (fn) => {
+                if (settled)
+                    return;
+                settled = true;
+                cleanup();
+                fn();
+            };
+            const onAbort = () => finish(() => reject(signal.reason instanceof Error
+                ? signal.reason
+                : new Error('Browser action cancelled')));
+            signal.addEventListener('abort', onAbort, { once: true });
+            void previous.then(() => finish(resolve), () => finish(resolve));
+        });
+    }
     async exclusive(input, task) {
         const previous = this.tails.get(input.sessionName) ?? Promise.resolve();
         let release;
-        const current = new Promise((resolve) => { release = resolve; });
+        const current = new Promise((resolve) => {
+            release = resolve;
+        });
         // 链上永远接一个已消化的 previous：任何一环 reject 都不能顺着队列传下去，
         // 否则后来者在 `await previous` 处抛出。
         const tail = previous.catch(() => undefined).then(() => current);
@@ -19,7 +45,7 @@ export class BrowserService {
         // 当前 current 只 resolve 不 reject，所以实际安全 —— 但那是一条没写下来的不变量，
         // 谁往链上加一个 .then 都可能打破它。把 await 挪进来，安全性就不再依赖它。
         try {
-            await previous.catch(() => undefined);
+            await this.waitForPrevious(previous, input.signal);
             const run = async (args) => {
                 if (input.signal?.aborted) {
                     return { code: -1, stdout: '', stderr: 'Browser action cancelled' };

package/src/tools/webbrowser/browserService.ts

@@ -8,15 +8,54 @@ import type { AgentBrowserRun, AgentBrowserRunner } from './browser.js';
 export class BrowserService {
   private tails = new Map<string, Promise<void>>();
 
-  async exclusive<T>(input: {
-    sessionName: string;
-    bin: string;
-    runner: AgentBrowserRunner;
-    signal?: AbortSignal;
-  }, task: (run: (args: string[]) => Promise<AgentBrowserRun>) => Promise<T>): Promise<T> {
+  private async waitForPrevious(
+    previous: Promise<void>,
+    signal?: AbortSignal,
+  ): Promise<void> {
+    if (!signal) return previous.catch(() => undefined);
+    if (signal.aborted)
+      throw signal.reason instanceof Error
+        ? signal.reason
+        : new Error('Browser action cancelled');
+    await new Promise<void>((resolve, reject) => {
+      let settled = false;
+      const cleanup = () => signal.removeEventListener('abort', onAbort);
+      const finish = (fn: () => void) => {
+        if (settled) return;
+        settled = true;
+        cleanup();
+        fn();
+      };
+      const onAbort = () =>
+        finish(() =>
+          reject(
+            signal.reason instanceof Error
+              ? signal.reason
+              : new Error('Browser action cancelled'),
+          ),
+        );
+      signal.addEventListener('abort', onAbort, { once: true });
+      void previous.then(
+        () => finish(resolve),
+        () => finish(resolve),
+      );
+    });
+  }
+
+  async exclusive<T>(
+    input: {
+      sessionName: string;
+      bin: string;
+      runner: AgentBrowserRunner;
+      signal?: AbortSignal;
+    },
+    task: (run: (args: string[]) => Promise<AgentBrowserRun>) => Promise<T>,
+  ): Promise<T> {
     const previous = this.tails.get(input.sessionName) ?? Promise.resolve();
     let release!: () => void;
-    const current = new Promise<void>((resolve) => { release = resolve; });
+    const current = new Promise<void>((resolve) => {
+      release = resolve;
+    });
     // 链上永远接一个已消化的 previous：任何一环 reject 都不能顺着队列传下去，
     // 否则后来者在 `await previous` 处抛出。
     const tail = previous.catch(() => undefined).then(() => current);
@@ -28,7 +67,7 @@ export class BrowserService {
     // 当前 current 只 resolve 不 reject，所以实际安全 —— 但那是一条没写下来的不变量，
     // 谁往链上加一个 .then 都可能打破它。把 await 挪进来，安全性就不再依赖它。
     try {
-      await previous.catch(() => undefined);
+      await this.waitForPrevious(previous, input.signal);
       const run = async (args: string[]): Promise<AgentBrowserRun> => {
         if (input.signal?.aborted) {
           return { code: -1, stdout: '', stderr: 'Browser action cancelled' };
@@ -38,7 +77,8 @@ export class BrowserService {
       return await task(run);
     } finally {
       release();
-      if (this.tails.get(input.sessionName) === tail) this.tails.delete(input.sessionName);
+      if (this.tails.get(input.sessionName) === tail)
+        this.tails.delete(input.sessionName);
     }
   }
 

package/src/tools/webbrowser/browserStream.js

@@ -1,5 +1,5 @@
+import { BrowserControlCoordinator, } from './browserControl.js';
 import { getStreamPort, waitForStreamPort } from './browserDaemon.js';
-import { BrowserControlCoordinator } from './browserControl.js';
 class SessionBrowserStream {
     sessionName;
     socket = null;
@@ -8,26 +8,39 @@ class SessionBrowserStream {
     port = null;
     mode = 'browse';
     currentUrl = 'about:blank';
-    status = { connected: false, screencasting: false };
+    status = {
+        connected: false,
+        screencasting: false,
+    };
+    reconnectTimer = null;
+    reconnectAttempt = 0;
+    disposed = false;
     picker;
     marked = [];
     control;
     constructor(sessionName) {
         this.sessionName = sessionName;
-        this.control = new BrowserControlCoordinator({ conversationKey: `browser:${sessionName}` });
+        this.control = new BrowserControlCoordinator({
+            conversationKey: `browser:${sessionName}`,
+        });
         this.control.subscribe((state) => this.emit({ type: 'control_state', state }));
         this.control.subscribeInterventions((event) => this.emit({
-            type: event.type === 'started' ? 'human_intervention_started' : 'human_intervention_resolved',
+            type: event.type === 'started'
+                ? 'human_intervention_started'
+                : 'human_intervention_resolved',
             event,
         }));
     }
     async connect() {
+        this.disposed = false;
         if (this.socket?.readyState === WebSocket.OPEN)
             return;
         if (this.connecting)
             return this.connecting;
         this.connecting = (async () => {
-            this.port = (await getStreamPort(this.sessionName)) ?? (await waitForStreamPort(this.sessionName));
+            this.port =
+                (await getStreamPort(this.sessionName)) ??
+                    (await waitForStreamPort(this.sessionName));
             await new Promise((resolve, reject) => {
                 const socket = new WebSocket(`ws://127.0.0.1:${this.port}`);
                 const timer = setTimeout(() => {
@@ -37,6 +50,7 @@ class SessionBrowserStream {
                 socket.onopen = () => {
                     clearTimeout(timer);
                     this.socket = socket;
+                    this.reconnectAttempt = 0;
                     this.status = { ...this.status, connected: true };
                     this.emit({ type: 'status', ...this.status });
                     resolve();
@@ -48,16 +62,33 @@ class SessionBrowserStream {
                 socket.onclose = () => {
                     if (this.socket === socket)
                         this.socket = null;
-                    this.status = { ...this.status, connected: false, screencasting: false };
+                    this.status = {
+                        ...this.status,
+                        connected: false,
+                        screencasting: false,
+                    };
                     this.emit({ type: 'status', ...this.status });
+                    this.scheduleReconnect();
                 };
                 socket.onmessage = (event) => this.onMessage(event.data);
             });
-        })().finally(() => { this.connecting = null; });
+        })().finally(() => {
+            this.connecting = null;
+        });
         return this.connecting;
     }
-    disconnect() { this.socket?.close(); this.socket = null; }
-    dispose() { this.disconnect(); this.listeners.clear(); }
+    disconnect() {
+        this.disposed = true;
+        if (this.reconnectTimer)
+            clearTimeout(this.reconnectTimer);
+        this.reconnectTimer = null;
+        this.socket?.close();
+        this.socket = null;
+    }
+    dispose() {
+        this.disconnect();
+        this.listeners.clear();
+    }
     send(frame) {
         if (this.socket?.readyState === WebSocket.OPEN)
             this.socket.send(JSON.stringify(frame));
@@ -66,22 +97,49 @@ class SessionBrowserStream {
         this.listeners.add(listener);
         return () => this.listeners.delete(listener);
     }
-    getStatus() { return this.status; }
-    getPort() { return this.port; }
-    isConnected() { return this.socket?.readyState === WebSocket.OPEN; }
-    setMode(mode) { this.mode = mode; this.emit({ type: 'mode', mode }); }
-    getMode() { return this.mode; }
-    getCurrentUrl() { return this.currentUrl; }
-    setCurrentUrl(url) { this.currentUrl = url; this.emit({ type: 'url', url }); }
-    setPicker(picker) { this.picker = picker; }
+    publish(event) {
+        this.emit(event);
+    }
+    getStatus() {
+        return this.status;
+    }
+    getPort() {
+        return this.port;
+    }
+    isConnected() {
+        return this.socket?.readyState === WebSocket.OPEN;
+    }
+    setMode(mode) {
+        this.mode = mode;
+        this.emit({ type: 'mode', mode });
+    }
+    getMode() {
+        return this.mode;
+    }
+    getCurrentUrl() {
+        return this.currentUrl;
+    }
+    setCurrentUrl(url) {
+        this.currentUrl = url;
+        this.emit({ type: 'url', url });
+    }
+    setPicker(picker) {
+        this.picker = picker;
+    }
     pickElement(x, y) {
         return this.picker ? this.picker(x, y) : Promise.resolve(null);
     }
     markElement(element) {
-        this.marked = [...this.marked.filter((item) => item.ref !== element.ref), element];
+        this.marked = [
+            ...this.marked.filter((item) => item.ref !== element.ref),
+            element,
+        ];
         this.emit({ type: 'marked_elements', elements: this.marked });
     }
-    clearMarkedElements() { this.marked = []; this.emit({ type: 'marked_elements', elements: [] }); }
+    clearMarkedElements() {
+        this.marked = [];
+        this.emit({ type: 'marked_elements', elements: [] });
+    }
     unmarkElement(ref) {
         const next = this.marked.filter((item) => item.ref !== ref);
         if (next.length !== this.marked.length) {
@@ -102,29 +160,67 @@ class SessionBrowserStream {
         // 只发语义终点（目标元素中心 + 动作类型）；平滑移动交给面板 CSS 过渡，内核不做逐帧插值（避免 WS 洪泛 + 双层缓动打架）。
         this.emit({ type: 'agent_pointer', ...pointer });
     }
-    getControlState() { return this.control.getState(); }
-    subscribeControl(listener) { return this.control.subscribe(listener); }
-    subscribeInterventions(listener) { return this.control.subscribeInterventions(listener); }
-    beginAgentAction(input) { this.control.beginAgent(input); }
-    finishAgentAction(runId) { this.control.finishAgent(runId); }
-    requestHuman(reason, input) { return this.control.requestHuman(reason, input); }
-    takeControl(input) { this.control.takeInteractive(input); }
-    resumeControl(token) { return this.control.resume(token); }
-    stopControl(token) { return this.control.stop(token); }
+    getControlState() {
+        return this.control.getState();
+    }
+    subscribeControl(listener) {
+        return this.control.subscribe(listener);
+    }
+    subscribeInterventions(listener) {
+        return this.control.subscribeInterventions(listener);
+    }
+    beginAgentAction(input) {
+        this.control.beginAgent(input);
+    }
+    finishAgentAction(runId) {
+        this.control.finishAgent(runId);
+    }
+    requestHuman(reason, input) {
+        return this.control.requestHuman(reason, input);
+    }
+    takeControl(input) {
+        this.control.takeInteractive(input);
+    }
+    resumeControl(token) {
+        return this.control.resume(token);
+    }
+    stopControl(token) {
+        return this.control.stop(token);
+    }
     resumeIntervention(credential) {
         return this.control.resumeIntervention(credential);
     }
     cancelIntervention(credential) {
         return this.control.cancelIntervention(credential);
     }
-    waitForHuman(signal) { return this.control.waitForHuman(signal); }
+    waitForHuman(signal) {
+        return this.control.waitForHuman(signal);
+    }
     emit(event) {
-        for (const listener of this.listeners)
-            listener(event);
+        for (const listener of this.listeners) {
+            try {
+                listener(event);
+            }
+            catch {
+                // A disconnected/failed UI observer must not break browser control or
+                // prevent the remaining observers from receiving the event.
+            }
+        }
+    }
+    scheduleReconnect() {
+        if (this.disposed || !this.listeners.size || this.reconnectTimer)
+            return;
+        const delay = Math.min(10_000, 250 * 2 ** this.reconnectAttempt++);
+        this.reconnectTimer = setTimeout(() => {
+            this.reconnectTimer = null;
+            void this.connect().catch(() => this.scheduleReconnect());
+        }, delay);
     }
     onMessage(data) {
         try {
-            const text = typeof data === 'string' ? data : Buffer.from(data).toString('utf8');
+            const text = typeof data === 'string'
+                ? data
+                : Buffer.from(data).toString('utf8');
             const event = JSON.parse(text);
             if (event.type === 'url' && typeof event.url === 'string')
                 this.currentUrl = event.url;
@@ -133,7 +229,10 @@ class SessionBrowserStream {
             this.emit(event);
         }
         catch (error) {
-            this.emit({ type: 'error', message: error instanceof Error ? error.message : String(error) });
+            this.emit({
+                type: 'error',
+                message: error instanceof Error ? error.message : String(error),
+            });
         }
     }
 }