chore(version): bump version to 1.2.0

chore(deps): bump the production-dependencies group with 2 updates

README-zh.md

@@ -40,7 +40,7 @@
 
 ```html
 <script
-    src="https://registry.npmmirror.com/page-agent/1.1.0/files/dist/iife/page-agent.demo.js"
+    src="https://registry.npmmirror.com/page-agent/1.2.0/files/dist/iife/page-agent.demo.js"
     crossorigin="true"
 ></script>
 ```
@@ -50,8 +50,8 @@
 
 | Mirrors | URL                                                                                |
 | ------- | ---------------------------------------------------------------------------------- |
-| Global  | https://cdn.jsdelivr.net/npm/page-agent@1.1.0/dist/iife/page-agent.demo.js         |
+| Global  | https://cdn.jsdelivr.net/npm/page-agent@1.2.0/dist/iife/page-agent.demo.js         |
-| China   | https://registry.npmmirror.com/page-agent/1.1.0/files/dist/iife/page-agent.demo.js |
+| China   | https://registry.npmmirror.com/page-agent/1.2.0/files/dist/iife/page-agent.demo.js |
 
 ### NPM 安装
 

README.md

@@ -40,7 +40,7 @@ Fastest way to try PageAgent with our free Demo LLM:
 
 ```html
 <script
-    src="https://cdn.jsdelivr.net/npm/page-agent@1.1.0/dist/iife/page-agent.demo.js"
+    src="https://cdn.jsdelivr.net/npm/page-agent@1.2.0/dist/iife/page-agent.demo.js"
     crossorigin="true"
 ></script>
 ```
@@ -50,8 +50,8 @@ Fastest way to try PageAgent with our free Demo LLM:
 
 | Mirrors | URL                                                                                |
 | ------- | ---------------------------------------------------------------------------------- |
-| Global  | https://cdn.jsdelivr.net/npm/page-agent@1.1.0/dist/iife/page-agent.demo.js         |
+| Global  | https://cdn.jsdelivr.net/npm/page-agent@1.2.0/dist/iife/page-agent.demo.js         |
-| China   | https://registry.npmmirror.com/page-agent/1.1.0/files/dist/iife/page-agent.demo.js |
+| China   | https://registry.npmmirror.com/page-agent/1.2.0/files/dist/iife/page-agent.demo.js |
 
 ### NPM Installation
 

package.json

@@ -1,7 +1,7 @@
 {
     "name": "root",
     "private": true,
-    "version": "1.1.0",
+    "version": "1.2.0",
     "type": "module",
     "workspaces": [
         "packages/page-controller",
@@ -21,8 +21,7 @@
     },
     "homepage": "https://alibaba.github.io/page-agent/",
     "engines": {
-        "node": ">=20.0.0",
+        "node": "^20.19.0 || ^22.13.0 || >=24"
-        "npm": ">=10.0.0"
     },
     "scripts": {
         "cleanup": "rm -rf packages/*/dist",
@@ -39,27 +38,30 @@
         "@commitlint/cli": "^20.3.1",
         "@commitlint/config-conventional": "^20.3.1",
         "@eslint/js": "^9.39.2",
-        "@microsoft/api-extractor": "^7.55.2",
+        "@microsoft/api-extractor": "^7.56.3",
+        "@tailwindcss/vite": "^4.1.18",
         "@trivago/prettier-plugin-sort-imports": "^6.0.2",
-        "@types/node": "^25.0.9",
+        "@types/node": "^25.2.2",
+        "@vitejs/plugin-react-swc": "^4.1.0",
         "chalk": "^5.6.2",
         "concurrently": "^9.2.1",
-        "dotenv": "^17.2.3",
+        "dotenv": "^17.2.4",
         "eslint": "^9.39.2",
         "eslint-config-prettier": "^10.1.8",
-        "eslint-plugin-react-dom": "^2.7.2",
+        "eslint-plugin-react-dom": "^2.12.2",
         "eslint-plugin-react-hooks": "^7.0.1",
         "eslint-plugin-react-refresh": "^0.5.0",
-        "eslint-plugin-react-x": "^2.7.2",
+        "eslint-plugin-react-x": "^2.12.2",
         "globals": "^17.0.0",
         "husky": "^9.1.7",
         "lint-staged": "^16.2.4",
         "prettier": "^3.8.0",
         "typescript": "^5.9.3",
-        "typescript-eslint": "^8.54.0",
+        "typescript-eslint": "^8.55.0",
         "unplugin-dts": "^1.0.0-beta.6",
         "vite": "^7.3.1",
-        "vite-plugin-css-injected-by-js": "^3.5.2"
+        "vite-plugin-css-injected-by-js": "^3.5.2",
+        "vite-bundle-analyzer": "^1.3.6"
     },
     "overrides": {
         "typescript": "^5.9.3"

packages/core/package.json

@@ -1,7 +1,7 @@
 {
     "name": "@page-agent/core",
     "private": false,
-    "version": "1.1.0",
+    "version": "1.2.0",
     "type": "module",
     "main": "./dist/esm/page-agent-core.js",
     "module": "./dist/esm/page-agent-core.js",
@@ -45,7 +45,7 @@
     "dependencies": {
         "chalk": "^5.6.2",
         "zod": "^4.3.5",
-        "@page-agent/llms": "1.1.0",
+        "@page-agent/llms": "1.2.0",
-        "@page-agent/page-controller": "1.1.0"
+        "@page-agent/page-controller": "1.2.0"
     }
 }

packages/core/src/PageAgentCore.ts

@@ -3,15 +3,15 @@
  * All rights reserved.
  */
 import { InvokeError, LLM, type Tool } from '@page-agent/llms'
-import type { PageController } from '@page-agent/page-controller'
+import type { BrowserState, PageController } from '@page-agent/page-controller'
 import chalk from 'chalk'
-import zod from 'zod'
+import * as zod from 'zod'
 
 import { type PageAgentConfig } from './config'
 import { DEFAULT_MAX_STEPS } from './config/constants'
 import SYSTEM_PROMPT from './prompts/system_prompt.md?raw'
 import { tools } from './tools'
-import {
+import type {
 	AgentActivity,
 	AgentReflection,
 	AgentStatus,
@@ -21,17 +21,25 @@ import {
 	MacroToolInput,
 	MacroToolResult,
 } from './types'
-import { normalizeResponse, trimLines, uid, waitFor } from './utils'
+import { assert, normalizeResponse, uid, waitFor } from './utils'
-import { assert } from './utils/assert'
 
 export { type PageAgentConfig }
 export { tool, type PageAgentTool } from './tools'
 export type * from './types'
 
 /**
- * AI agent for browser DOM automation.
+ * AI agent for browser automation.
  *
  * @remarks
+ * ## Re-act Agent Loop
+ * - step
+ *    - observe (gather information about current environment and context)
+ *    - think (LLM calling)
+ *      - reflection (evaluate history, generate memory, short-term planning)
+ *      - action (give the action to approach the next goal)
+ *    - act (execute the action)
+ * - loop
+ *
  * ## Event System
  * - `statuschange` - Agent status transitions (idle → running → completed/error)
  * - `historychange` - History events updated (persistent, part of agent memory)
@@ -50,14 +58,16 @@ export type * from './types'
  *    - Types: thinking, executing, executed, retrying, error
  */
 export class PageAgentCore extends EventTarget {
-	config: PageAgentConfig & { maxSteps: number }
+	readonly id = uid()
-	id = uid()
+	readonly config: PageAgentConfig & { maxSteps: number }
-	tools: typeof tools
+	readonly tools: typeof tools
+	/** PageController for DOM operations */
+	readonly pageController: PageController
+
 	task = ''
 	taskId = ''
-
+	/** History events */
-	/** Agent execution status */
+	history: HistoricalEvent[] = []
-	#status: AgentStatus = 'idle'
 
 	/**
 	 * Callback for when agent needs user input (ask_user tool)
@@ -66,23 +76,21 @@ export class PageAgentCore extends EventTarget {
 	 */
 	onAskUser?: (question: string) => Promise<string>
 
+	#status: AgentStatus = 'idle'
 	#llm: LLM
 	#abortController = new AbortController()
+	#observations: string[] = []
 
-	/** PageController for DOM operations */
+	/** internal states during a single task execution */
-	pageController: PageController
+	#states = {
-
+		/** Accumulated wait time in seconds */
-	/** Runtime states for tracking across steps */
-	states = {
-		/** Accumulated wait time in seconds, used by wait tool */
 		totalWaitTime: 0,
-		/** Last known URL for detecting navigation */
+		/** For detecting navigation */
 		lastURL: '',
+		/** Browser state */
+		browserState: null as BrowserState | null,
 	}
 
-	/** History events */
-	history: HistoricalEvent[] = []
-
 	constructor(config: PageAgentConfig & { pageController: PageController }) {
 		super()
 
@@ -166,12 +174,13 @@ export class PageAgentCore extends EventTarget {
 	}
 
 	/**
-	 * Push a persistent observation to the history event stream.
+	 * Push a observation message to the history event stream.
-	 * This will be visible in <agent_history> and remain in memory across steps.
+	 * This will be visible in <agent_history> and remain persistent in memory across steps.
+	 * @experimental @internal
+	 * @note history change will be emitted before next step starts
 	 */
 	pushObservation(content: string): void {
-		this.history.push({ type: 'observation', content })
+		this.#observations.push(content)
-		this.#emitHistoryChange()
 	}
 
 	async execute(task: string): Promise<ExecutionResult> {
@@ -203,30 +212,25 @@ export class PageAgentCore extends EventTarget {
 		this.#setStatus('running')
 		this.#emitHistoryChange()
 
-		// Reset states
+		// Reset internal states
-		this.states = {
+		this.#states = { totalWaitTime: 0, lastURL: '', browserState: null }
-			totalWaitTime: 0,
-			lastURL: '',
-		}
 
-		try {
+		let step = 0
-			let step = 0
 
-			while (true) {
+		while (true) {
-				await this.#generateObservations(step)
+			try {
+				console.group(`step: ${step}`)
 
 				await onBeforeStep?.(this, step)
 
-				console.group(`step: ${step}`)
+				// observe
 
-				// abort
+				console.log(chalk.blue.bold('👀 Observing...'))
-				if (this.#abortController.signal.aborted) throw new Error('AbortError')
 
-				// Thinking
+				this.#states.browserState = await this.pageController.getBrowserState()
-				console.log(chalk.blue('Thinking...'))
+				await this.#handleObservations(step)
-				this.#emitActivity({ type: 'thinking' })
 
-				// invoke LLM
+				// assemble prompts
 
 				const messages = [
 					{ role: 'system' as const, content: this.#getSystemPrompt() },
@@ -235,12 +239,17 @@ export class PageAgentCore extends EventTarget {
 
 				const tools = { AgentOutput: this.#packMacroTool() }
 
+				// invoke LLM
+
+				console.log(chalk.blue.bold('🧠 Thinking...'))
+				this.#emitActivity({ type: 'thinking' })
+
 				const result = await this.#llm.invoke(messages, tools, this.#abortController.signal, {
 					toolChoiceName: 'AgentOutput',
 					normalizeResponse,
 				})
 
-				// assemble history event
+				// assemble history
 
 				const macroResult = result.toolResult as MacroToolResult
 				const input = macroResult.input
@@ -270,22 +279,12 @@ export class PageAgentCore extends EventTarget {
 
 				//
 
-				console.log(chalk.green('Step finished:'), actionName)
-				console.groupEnd()
-
 				await onAfterStep?.(this, this.history)
 
-				step++
+				console.groupEnd()
-				if (step > this.config.maxSteps) {
+
-					this.#onDone(false)
+				// finish task if done
-					const result: ExecutionResult = {
+
-						success: false,
-						data: 'Step count exceeded maximum limit',
-						history: this.history,
-					}
-					await onAfterTask?.(this, result)
-					return result
-				}
 				if (actionName === 'done') {
 					const success = action.input?.success ?? false
 					const text = action.input?.text || 'no text provided'
@@ -299,19 +298,32 @@ export class PageAgentCore extends EventTarget {
 					await onAfterTask?.(this, result)
 					return result
 				}
+			} catch (error: unknown) {
+				console.groupEnd() // to prevent nested groups
+				console.error('Task failed', error)
+				const errorMessage = String(error)
+				this.#emitActivity({ type: 'error', message: errorMessage })
+				this.#onDone(false)
+				const result: ExecutionResult = {
+					success: false,
+					data: errorMessage,
+					history: this.history,
+				}
+				await onAfterTask?.(this, result)
+				return result
 			}
-		} catch (error: unknown) {
+
-			console.error('Task failed', error)
+			step++
-			const errorMessage = String(error)
+			if (step > this.config.maxSteps) {
-			this.#emitActivity({ type: 'error', message: errorMessage })
+				this.#onDone(false)
-			this.#onDone(false)
+				const result: ExecutionResult = {
-			const result: ExecutionResult = {
+					success: false,
-				success: false,
+					data: 'Step count exceeded maximum limit',
-				data: errorMessage,
+					history: this.history,
-				history: this.history,
+				}
+				await onAfterTask?.(this, result)
+				return result
 			}
-			await onAfterTask?.(this, result)
-			return result
 		}
 	}
 
@@ -344,13 +356,13 @@ export class PageAgentCore extends EventTarget {
 		})
 
 		return {
-			description: 'You MUST call this tool every step. Outputs your reflections and next action.',
+			description: 'You MUST call this tool every step!',
 			inputSchema: macroToolSchema as zod.ZodType<MacroToolInput>,
 			execute: async (input: MacroToolInput): Promise<MacroToolResult> => {
 				// abort
 				if (this.#abortController.signal.aborted) throw new Error('AbortError')
 
-				console.log(chalk.blue.bold('MacroTool execute'), input)
+				console.log(chalk.blue.bold('MacroTool input'), input)
 				const action = input.action
 
 				const toolName = Object.keys(action)[0]
@@ -371,7 +383,7 @@ export class PageAgentCore extends EventTarget {
 
 				// Find the corresponding tool
 				const tool = tools.get(toolName)
-				assert(tool, `Tool ${toolName} not found. (@note should have been caught before this!!!)`)
+				assert(tool, `Tool ${toolName} not found`)
 
 				console.log(chalk.blue.bold(`Executing tool: ${toolName}`), toolInput)
 
@@ -395,9 +407,11 @@ export class PageAgentCore extends EventTarget {
 					duration,
 				})
 
-				// Reset wait time for non-wait tools
+				// counting wait time
-				if (toolName !== 'wait') {
+				if (toolName === 'wait') {
-					this.states.totalWaitTime = 0
+					this.#states.totalWaitTime += toolInput?.seconds || 0
+				} else {
+					this.#states.totalWaitTime = 0
 				}
 
 				// Return structured result
@@ -427,17 +441,17 @@ export class PageAgentCore extends EventTarget {
 	}
 
 	/**
-	 * Get instructions from config and format as XML block
+	 * Get instructions from config
 	 */
 	async #getInstructions(): Promise<string> {
 		const { instructions } = this.config
 		if (!instructions) return ''
 
 		const systemInstructions = instructions.system?.trim()
-		const url = await this.pageController.getCurrentUrl()
 		let pageInstructions: string | undefined
 
-		if (instructions.getPageInstructions) {
+		const url = this.#states.browserState?.url || ''
+		if (instructions.getPageInstructions && url) {
 			try {
 				pageInstructions = instructions.getPageInstructions(url)?.trim()
 			} catch (error) {
@@ -465,23 +479,28 @@ export class PageAgentCore extends EventTarget {
 	}
 
 	/**
-	 * Generate observations before each step
+	 * Generate system observations before each step
-	 * - URL change detection
-	 * - Too many steps warning
 	 * @todo loop detection
 	 * @todo console error
 	 */
-	async #generateObservations(stepCount: number): Promise<void> {
+	async #handleObservations(step: number): Promise<void> {
+		// Accumulated wait time warning
+		if (this.#states.totalWaitTime >= 3) {
+			this.pushObservation(
+				`You have waited ${this.#states.totalWaitTime} seconds accumulatively. DO NOT wait any longer unless you have a good reason.`
+			)
+		}
+
 		// Detect URL change
-		const currentURL = await this.pageController.getCurrentUrl()
+		const currentURL = this.#states.browserState?.url || ''
-		if (currentURL !== this.states.lastURL) {
+		if (currentURL !== this.#states.lastURL) {
 			this.pushObservation(`Page navigated to → ${currentURL}`)
-			this.states.lastURL = currentURL
+			this.#states.lastURL = currentURL
 			await waitFor(0.5) // wait for page to stabilize
 		}
 
-		// Warn about remaining steps
+		// Remaining steps warning
-		const remaining = this.config.maxSteps - stepCount
+		const remaining = this.config.maxSteps - step
 		if (remaining === 5) {
 			this.pushObservation(
 				`⚠️ Only ${remaining} steps remaining. Consider wrapping up or calling done with partial results.`
@@ -491,12 +510,25 @@ export class PageAgentCore extends EventTarget {
 				`⚠️ Critical: Only ${remaining} steps left! You must finish the task or call done immediately.`
 			)
 		}
+
+		// Push observations to history and emit
+		if (this.#observations.length > 0) {
+			for (const content of this.#observations) {
+				this.history.push({ type: 'observation', content })
+				console.log(chalk.cyan('Observation:'), content)
+			}
+			this.#observations = []
+			this.#emitHistoryChange()
+		}
 	}
 
 	async #assembleUserPrompt(): Promise<string> {
+		const browserState = this.#states.browserState!
+
 		let prompt = ''
 
 		// <instructions> (optional)
+
 		prompt += await this.#getInstructions()
 
 		// <agent_state>
@@ -506,38 +538,36 @@ export class PageAgentCore extends EventTarget {
 
 		const stepCount = this.history.filter((e) => e.type === 'step').length
 
-		prompt += `<agent_state>
+		prompt += '<agent_state>\n'
-			<user_request>
+		prompt += '<user_request>\n'
-			${this.task}
+		prompt += `${this.task}\n`
-			</user_request>
+		prompt += '</user_request>\n'
-			<step_info>
+		prompt += '<step_info>\n'
-			Step ${stepCount + 1} of ${this.config.maxSteps} max possible steps
+		prompt += `Step ${stepCount + 1} of ${this.config.maxSteps} max possible steps\n`
-			Current date and time: ${new Date().toLocaleString()}
+		prompt += `Current time: ${new Date().toLocaleString()}\n`
-			</step_info>
+		prompt += '</step_info>\n'
-			</agent_state>
+		prompt += '</agent_state>\n\n'
-		`
 
 		// <agent_history>
 		//  - <step_N> for steps
 		//  - <sys> for observations and system messages
 
-		prompt += '\n<agent_history>\n'
+		prompt += '<agent_history>\n'
 
 		let stepIndex = 0
 		for (const event of this.history) {
 			if (event.type === 'step') {
 				stepIndex++
-				prompt += `<step_${stepIndex}>
+				prompt += `<step_${stepIndex}>\n`
-				Evaluation of Previous Step: ${event.reflection.evaluation_previous_goal}
+				prompt += `Evaluation of Previous Step: ${event.reflection.evaluation_previous_goal}\n`
-				Memory: ${event.reflection.memory}
+				prompt += `Memory: ${event.reflection.memory}\n`
-				Next Goal: ${event.reflection.next_goal}
+				prompt += `Next Goal: ${event.reflection.next_goal}\n`
-				Action Results: ${event.action.output}
+				prompt += `Action Results: ${event.action.output}\n`
-				</step_${stepIndex}>
+				prompt += `</step_${stepIndex}>\n`
-			`
 			} else if (event.type === 'observation') {
 				prompt += `<sys>${event.content}</sys>\n`
 			} else if (event.type === 'user_takeover') {
-				prompt += `<sys>User took over control and made changes to the page.</sys>\n`
+				prompt += `<sys>User took over control and made changes to the page</sys>\n`
 			} else if (event.type === 'error') {
 				// Error events are mainly for panel rendering, not included in LLM context
 				// to avoid polluting the agent's reasoning with transient errors
@@ -548,9 +578,18 @@ export class PageAgentCore extends EventTarget {
 
 		// <browser_state>
 
-		prompt += await this.#getBrowserState()
+		let pageContent = browserState.content
+		if (this.config.transformPageContent) {
+			pageContent = await this.config.transformPageContent(pageContent)
+		}
 
-		return trimLines(prompt)
+		prompt += '<browser_state>\n'
+		prompt += browserState.header + '\n'
+		prompt += pageContent + '\n'
+		prompt += browserState.footer + '\n\n'
+		prompt += '</browser_state>\n\n'
+
+		return prompt
 	}
 
 	#onDone(success = true) {
@@ -560,23 +599,6 @@ export class PageAgentCore extends EventTarget {
 		this.#abortController.abort()
 	}
 
-	async #getBrowserState(): Promise<string> {
-		const state = await this.pageController.getBrowserState()
-
-		let content = state.content
-		if (this.config.transformPageContent) {
-			content = await this.config.transformPageContent(content)
-		}
-
-		return trimLines(`<browser_state>
-			${state.header}
-			${content}
-			${state.footer}
-			
-			</browser_state>
-		`)
-	}
-
 	dispose() {
 		console.log('Disposing PageAgent...')
 		this.pageController.dispose()

packages/core/src/prompts/system_prompt.md

@@ -128,7 +128,6 @@ Exhibit the following reasoning patterns to successfully achieve the <user_reque
 Here are examples of good output patterns. Use them as reference but never copy them directly.
 
 <evaluation_examples>
-- Positive Examples:
 "evaluation_previous_goal": "Successfully navigated to the product page and found the target information. Verdict: Success"
 "evaluation_previous_goal": "Clicked the login button and user authentication form appeared. Verdict: Success"
 </evaluation_examples>
@@ -139,17 +138,16 @@ Here are examples of good output patterns. Use them as reference but never copy
 
 <next_goal_examples>
 "next_goal": "Click on the 'Add to Cart' button to proceed with the purchase flow."
-"next_goal": "Extract details from the first item on the page."
 </next_goal_examples>
 </examples>
 
 <output>
-You must ALWAYS respond with a valid JSON in this exact format:
-
 {
   "evaluation_previous_goal": "Concise one-sentence analysis of your last action. Clearly state success, failure, or uncertain.",
   "memory": "1-3 concise sentences of specific memory of this step and overall progress. You should put here everything that will help you track progress in future steps. Like counting pages visited, items found, etc.",
-  "next_goal": "State the next immediate goal and action to achieve it, in one clear sentence."
+  "next_goal": "State the next immediate goal and action to achieve it, in one clear sentence.",
-  "action":{"action": {// action-specific parameter}}
+  "action":{
+    "Action name": {// Action parameters}
+  }
 }
 </output>

packages/core/src/tools/index.ts

@@ -2,7 +2,7 @@
  * Internal tools for PageAgent.
  * @note Adapted from browser-use
  */
-import zod, { type z } from 'zod'
+import * as zod from 'zod'
 
 import type { PageAgentCore } from '../PageAgentCore'
 import { waitFor } from '../utils'
@@ -13,7 +13,7 @@ import { waitFor } from '../utils'
 export interface PageAgentTool<TParams = any> {
 	// name: string
 	description: string
-	inputSchema: z.ZodType<TParams>
+	inputSchema: zod.ZodType<TParams>
 	execute: (this: PageAgentCore, args: TParams) => Promise<string>
 }
 
@@ -31,14 +31,13 @@ tools.set(
 	'done',
 	tool({
 		description:
-			'Complete task - provide a summary of results for the user. Set success=True if task completed successfully, false otherwise. Text should be your response to the user summarizing results.',
+			'Complete task. Text is your final response to the user — keep it concise unless the user explicitly asks for detail.',
 		inputSchema: zod.object({
 			text: zod.string(),
 			success: zod.boolean().default(true),
 		}),
 		execute: async function (this: PageAgentCore, input) {
 			// @note main loop will handle this one
-			// this.onDone(input.text, input.success)
 			return Promise.resolve('Task completed')
 		},
 	})
@@ -47,25 +46,17 @@ tools.set(
 tools.set(
 	'wait',
 	tool({
-		description:
+		description: 'Wait for x seconds. Can be used to wait until the page or data is fully loaded.',
-			'Wait for x seconds. default 1s (max 10 seconds, min 1 second). This can be used to wait until the page or data is fully loaded.',
 		inputSchema: zod.object({
 			seconds: zod.number().min(1).max(10).default(1),
 		}),
 		execute: async function (this: PageAgentCore, input) {
+			// try to subtract LLM calling time from the actual wait time
 			const lastTimeUpdate = await this.pageController.getLastUpdateTime()
 			const actualWaitTime = Math.max(0, input.seconds - (Date.now() - lastTimeUpdate) / 1000)
 			console.log(`actualWaitTime: ${actualWaitTime} seconds`)
 			await waitFor(actualWaitTime)
 
-			this.states.totalWaitTime += input.seconds
-
-			if (this.states.totalWaitTime >= 3) {
-				this.pushObservation(
-					`You have waited ${this.states.totalWaitTime} seconds accumulatively. Do NOT wait any longer unless you have a good reason.`
-				)
-			}
-
 			return `✅ Waited for ${input.seconds} seconds.`
 		},
 	})
@@ -106,7 +97,7 @@ tools.set(
 tools.set(
 	'input_text',
 	tool({
-		description: 'Click and input text into a input interactive element',
+		description: 'Click and type text into an interactive input element',
 		inputSchema: zod.object({
 			index: zod.int().min(0),
 			text: zod.string(),
@@ -140,8 +131,7 @@ tools.set(
 tools.set(
 	'scroll',
 	tool({
-		description:
+		description: 'Scroll the page vertically. Use index for scroll elements (dropdowns/custom UI).',
-			'Scroll the page by specified number of pages (set down=True to scroll down, down=False to scroll up, num_pages=number of pages to scroll like 0.5 for half page, 1.0 for one page, etc.). Optional index parameter to scroll within a specific element or its scroll container (works well for dropdowns and custom UI components). Optional pixels parameter to scroll by a specific number of pixels instead of pages.',
 		inputSchema: zod.object({
 			down: zod.boolean().default(true),
 			num_pages: zod.number().min(0).max(10).optional().default(0.1),
@@ -158,11 +148,14 @@ tools.set(
 	})
 )
 
+/**
+ * @todo Tables need a dedicated parser to extract structured data. This tool is useless.
+ */
 tools.set(
 	'scroll_horizontally',
 	tool({
 		description:
-			'Scroll the page or element horizontally (set right=True to scroll right, right=False to scroll left, pixels=number of pixels to scroll). Optional index parameter to scroll within a specific element or its scroll container (works well for wide tables).',
+			'Scroll the page horizontally, or within a specific element by index. Useful for wide tables.',
 		inputSchema: zod.object({
 			right: zod.boolean().default(true),
 			pixels: zod.number().int().min(0),
@@ -190,8 +183,6 @@ tools.set(
 	})
 )
 
-// @todo get_dropdown_options
-// @todo select_dropdown_option
 // @todo send_keys
 // @todo upload_file
 // @todo go_back

packages/core/src/utils/assert.ts

@@ -1,17 +0,0 @@
-import chalk from 'chalk'
-
-/**
- * Simple assertion function that throws an error if the condition is falsy
- * @param condition - The condition to assert
- * @param message - Optional error message
- * @throws Error if condition is falsy
- */
-export function assert(condition: unknown, message?: string, silent?: boolean): asserts condition {
-	if (!condition) {
-		const errorMessage = message ?? 'Assertion failed'
-
-		if (!silent) console.error(chalk.red(`❌ assert: ${errorMessage}`))
-
-		throw new Error(errorMessage)
-	}
-}

packages/core/src/utils/index.ts

@@ -1,27 +1,7 @@
+import chalk from 'chalk'
+
 export { normalizeResponse } from './autoFixer'
 
-/**
- * Wait until condition becomes true
- * @returns Returns when condition becomes true, throws otherwise
- * @param timeout Timeout in milliseconds, default 0 means no timeout, throws error on timeout
- */
-export async function waitUntil(check: () => boolean, timeout = 60 * 60_1000): Promise<boolean> {
-	if (check()) return true
-
-	return new Promise((resolve, reject) => {
-		const start = Date.now()
-		const interval = setInterval(() => {
-			if (check()) {
-				clearInterval(interval)
-				resolve(true)
-			} else if (Date.now() - start > timeout) {
-				clearInterval(interval)
-				reject(new Error('Timeout waiting for condition to become true'))
-			}
-		}, 100)
-	})
-}
-
 export async function waitFor(seconds: number): Promise<void> {
 	await new Promise((resolve) => setTimeout(resolve, seconds * 1000))
 }
@@ -37,15 +17,6 @@ export function truncate(text: string, maxLength: number): string {
 
 //
 
-export function trimLines(text: string): string {
-	return text
-		.split('\n')
-		.map((line) => line.trim())
-		.join('\n')
-}
-
-//
-
 export function randomID(existingIDs?: string[]): string {
 	let id = Math.random().toString(36).substring(2, 11)
 
@@ -85,3 +56,19 @@ export function uid() {
 	ids.push(id)
 	return id
 }
+
+/**
+ * Simple assertion function that throws an error if the condition is falsy
+ * @param condition - The condition to assert
+ * @param message - Optional error message
+ * @throws Error if condition is falsy
+ */
+export function assert(condition: unknown, message?: string, silent?: boolean): asserts condition {
+	if (!condition) {
+		const errorMessage = message ?? 'Assertion failed'
+
+		if (!silent) console.error(chalk.red(`❌ assert: ${errorMessage}`))
+
+		throw new Error(errorMessage)
+	}
+}

packages/extension/docs/extension_api.md

@@ -33,7 +33,6 @@ import type {
   AgentStatus,
   ExecutionResult,
   HistoricalEvent,
-  LLMConfig,
 } from '@page-agent/core'
 
 // Wait for extension injection (up to 1 second)
@@ -48,18 +47,13 @@ async function waitForExtension(timeout = 1000): Promise<boolean> {
 
 // Usage
 if (await waitForExtension()) {
-  const result = await window.PAGE_AGENT_EXT!.execute(
+  const result = await window.PAGE_AGENT_EXT!.execute('Click the login button', {
-    'Click the login button',
+    baseURL: 'https://api.openai.com/v1',
-    {
+    apiKey: 'your-api-key',
-      baseURL: 'https://api.openai.com/v1',
+    model: 'gpt-5.2',
-      apiKey: 'your-api-key',
+    onStatusChange: (status) => console.log('Status:', status),
-      model: 'gpt-5.2',
+    onActivity: (activity) => console.log('Activity:', activity),
-    },
+  })
-    {
-      onStatusChange: (status) => console.log('Status:', status),
-      onActivity: (activity) => console.log('Activity:', activity),
-    }
-  )
   console.log('Result:', result)
 }
 ```
@@ -76,7 +70,7 @@ Extension version string (e.g., `"1.0.0"`). This is exposed separately to allow
 
 Main API namespace object containing:
 
-#### `PAGE_AGENT_EXT.execute(task, llmConfig, hooks?)`
+#### `PAGE_AGENT_EXT.execute(task, config)`
 
 Execute an agent task.
 
@@ -85,8 +79,7 @@ Execute an agent task.
 | Name | Type | Required | Description |
 |------|------|----------|-------------|
 | `task` | `string` | Yes | Task description |
-| `llmConfig` | `LLMConfig` | Yes | LLM configuration |
+| `config` | `ExecuteConfig` | Yes | Execution configuration (LLM settings, options, and event callbacks) |
-| `hooks` | `ExecuteHooks` | No | Event callbacks |
 
 **Returns:** `Promise<ExecutionResult>`
 
@@ -104,21 +97,26 @@ import type {
   AgentStatus,
   ExecutionResult,
   HistoricalEvent,
-  LLMConfig,
 } from '@page-agent/core'
 
-export interface ExecuteHooks {
+export interface ExecuteConfig {
+  baseURL: string
+  apiKey: string
+  model: string
+
+  /**
+   * Whether to include the initial tab (that holds this main world script) in the task.
+   * @default true
+   */
+  includeInitialTab?: boolean
+
   onStatusChange?: (status: AgentStatus) => void
   onActivity?: (activity: AgentActivity) => void
   onHistoryUpdate?: (history: HistoricalEvent[]) => void
   onDispose?: () => void
 }
 
-export type Execute = (
+export type Execute = (task: string, config: ExecuteConfig) => Promise<ExecutionResult>
-  task: string,
-  llmConfig: LLMConfig,
-  hooks?: ExecuteHooks
-) => Promise<ExecutionResult>
 ```
 
 ### AgentStatus
@@ -164,16 +162,6 @@ type HistoricalEvent =
   | { type: 'error'; message: string; rawResponse?: unknown }
 ```
 
-### LLMConfig
-
-```typescript
-interface LLMConfig {
-  baseURL: string   // e.g. 'https://api.openai.com/v1'
-  apiKey: string
-  model: string     // e.g. 'gpt-5.2'
-}
-```
-
 ### ExecutionResult
 
 ```typescript
@@ -205,44 +193,63 @@ if (result.success) {
 }
 ```
 
-### With Event Hooks
+### Exclude Initial Tab
+
+By default, the agent includes the initial tab (where the script runs) in the task. Set `includeInitialTab: false` to exclude it:
 
 ```typescript
-await window.PAGE_AGENT_EXT!.execute(
+const result = await window.PAGE_AGENT_EXT!.execute(
-  'Navigate to the settings page',
+  'Open a new tab and search for page-agent on GitHub',
-  llmConfig,
   {
-    onStatusChange: (status) => {
+    baseURL: 'https://api.openai.com/v1',
-      updateUI({ agentStatus: status })
+    apiKey: process.env.OPENAI_API_KEY!,
-    },
+    model: 'gpt-5.2',
-    onActivity: (activity) => {
+    includeInitialTab: false,  // Agent will open new tabs only
-      switch (activity.type) {
-        case 'thinking':
-          showSpinner('Agent is thinking...')
-          break
-        case 'executing':
-          showSpinner(`Executing: ${activity.tool}`)
-          break
-        case 'executed':
-          log(`${activity.tool} completed in ${activity.duration}ms`)
-          break
-        case 'error':
-          showError(activity.message)
-          break
-      }
-    },
-    onHistoryUpdate: (history) => {
-      renderHistory(history)
-    },
   }
 )
 ```
 
+### With Event Callbacks
+
+```typescript
+await window.PAGE_AGENT_EXT!.execute('Navigate to the settings page', {
+  baseURL: 'https://api.openai.com/v1',
+  apiKey: process.env.OPENAI_API_KEY!,
+  model: 'gpt-5.2',
+  onStatusChange: (status) => {
+    updateUI({ agentStatus: status })
+  },
+  onActivity: (activity) => {
+    switch (activity.type) {
+      case 'thinking':
+        showSpinner('Agent is thinking...')
+        break
+      case 'executing':
+        showSpinner(`Executing: ${activity.tool}`)
+        break
+      case 'executed':
+        log(`${activity.tool} completed in ${activity.duration}ms`)
+        break
+      case 'error':
+        showError(activity.message)
+        break
+    }
+  },
+  onHistoryUpdate: (history) => {
+    renderHistory(history)
+  },
+})
+```
+
 ### Stop Execution
 
 ```typescript
 // Start a task
-window.PAGE_AGENT_EXT!.execute('Scroll through all pages', llmConfig)
+window.PAGE_AGENT_EXT!.execute('Scroll through all pages', {
+  baseURL: 'https://api.openai.com/v1',
+  apiKey: process.env.OPENAI_API_KEY!,
+  model: 'gpt-5.2',
+})
 
 // Later, stop it
 window.PAGE_AGENT_EXT!.dispose()
@@ -258,24 +265,25 @@ import type {
   AgentStatus,
   ExecutionResult,
   HistoricalEvent,
-  LLMConfig,
 } from '@page-agent/core'
 
+interface ExecuteConfig {
+  baseURL: string
+  apiKey: string
+  model: string
+  includeInitialTab?: boolean
+  onStatusChange?: (status: AgentStatus) => void
+  onActivity?: (activity: AgentActivity) => void
+  onHistoryUpdate?: (history: HistoricalEvent[]) => void
+  onDispose?: () => void
+}
+
 declare global {
   interface Window {
     PAGE_AGENT_EXT_VERSION?: string
     PAGE_AGENT_EXT?: {
       version: string
-      execute: (
+      execute: (task: string, config: ExecuteConfig) => Promise<ExecutionResult>
-        task: string,
-        llmConfig: LLMConfig,
-        hooks?: {
-          onStatusChange?: (status: AgentStatus) => void
-          onActivity?: (activity: AgentActivity) => void
-          onHistoryUpdate?: (history: HistoricalEvent[]) => void
-          onDispose?: () => void
-        }
-      ) => Promise<ExecutionResult>
       dispose: () => void
     }
   }

packages/extension/docs/extension_api_zh.md

@@ -33,7 +33,6 @@ import type {
   AgentStatus,
   ExecutionResult,
   HistoricalEvent,
-  LLMConfig,
 } from '@page-agent/core'
 
 // 等待插件注入（最多 1 秒）
@@ -48,18 +47,13 @@ async function waitForExtension(timeout = 1000): Promise<boolean> {
 
 // 使用
 if (await waitForExtension()) {
-  const result = await window.PAGE_AGENT_EXT!.execute(
+  const result = await window.PAGE_AGENT_EXT!.execute('点击登录按钮', {
-    '点击登录按钮',
+    baseURL: 'https://api.openai.com/v1',
-    {
+    apiKey: 'your-api-key',
-      baseURL: 'https://api.openai.com/v1',
+    model: 'gpt-5.2',
-      apiKey: 'your-api-key',
+    onStatusChange: (status) => console.log('状态:', status),
-      model: 'gpt-5.2',
+    onActivity: (activity) => console.log('活动:', activity),
-    },
+  })
-    {
-      onStatusChange: (status) => console.log('状态:', status),
-      onActivity: (activity) => console.log('活动:', activity),
-    }
-  )
   console.log('结果:', result)
 }
 ```
@@ -76,7 +70,7 @@ if (await waitForExtension()) {
 
 主 API 命名空间对象，包含：
 
-#### `PAGE_AGENT_EXT.execute(task, llmConfig, hooks?)`
+#### `PAGE_AGENT_EXT.execute(task, config)`
 
 执行 Agent 任务。
 
@@ -85,8 +79,7 @@ if (await waitForExtension()) {
 | 名称 | 类型 | 必填 | 说明 |
 |------|------|------|------|
 | `task` | `string` | 是 | 任务描述 |
-| `llmConfig` | `LLMConfig` | 是 | LLM 配置 |
+| `config` | `ExecuteConfig` | 是 | 执行配置（LLM 设置、选项和事件回调） |
-| `hooks` | `ExecuteHooks` | 否 | 事件回调 |
 
 **返回：** `Promise<ExecutionResult>`
 
@@ -104,21 +97,26 @@ import type {
   AgentStatus,
   ExecutionResult,
   HistoricalEvent,
-  LLMConfig,
 } from '@page-agent/core'
 
-export interface ExecuteHooks {
+export interface ExecuteConfig {
+  baseURL: string
+  apiKey: string
+  model: string
+
+  /**
+   * 是否将初始标签页（运行此脚本的页面）包含在任务中。
+   * @default true
+   */
+  includeInitialTab?: boolean
+
   onStatusChange?: (status: AgentStatus) => void
   onActivity?: (activity: AgentActivity) => void
   onHistoryUpdate?: (history: HistoricalEvent[]) => void
   onDispose?: () => void
 }
 
-export type Execute = (
+export type Execute = (task: string, config: ExecuteConfig) => Promise<ExecutionResult>
-  task: string,
-  llmConfig: LLMConfig,
-  hooks?: ExecuteHooks
-) => Promise<ExecutionResult>
 ```
 
 ### AgentStatus
@@ -164,16 +162,6 @@ type HistoricalEvent =
   | { type: 'error'; message: string; rawResponse?: unknown }
 ```
 
-### LLMConfig
-
-```typescript
-interface LLMConfig {
-  baseURL: string   // 例如 'https://api.openai.com/v1'
-  apiKey: string
-  model: string     // 例如 'gpt-5.2'
-}
-```
-
 ### ExecutionResult
 
 ```typescript
@@ -205,44 +193,63 @@ if (result.success) {
 }
 ```
 
+### 排除初始标签页
+
+默认情况下，Agent 会将初始标签页（运行脚本的页面）包含在任务中。设置 `includeInitialTab: false` 可以排除它：
+
+```typescript
+const result = await window.PAGE_AGENT_EXT!.execute(
+  '打开新标签页并在 GitHub 上搜索 page-agent',
+  {
+    baseURL: 'https://api.openai.com/v1',
+    apiKey: process.env.OPENAI_API_KEY!,
+    model: 'gpt-5.2',
+    includeInitialTab: false,  // Agent 只会打开新标签页
+  }
+)
+```
+
 ### 使用事件回调
 
 ```typescript
-await window.PAGE_AGENT_EXT!.execute(
+await window.PAGE_AGENT_EXT!.execute('导航到设置页面', {
-  '导航到设置页面',
+  baseURL: 'https://api.openai.com/v1',
-  llmConfig,
+  apiKey: process.env.OPENAI_API_KEY!,
-  {
+  model: 'gpt-5.2',
-    onStatusChange: (status) => {
+  onStatusChange: (status) => {
-      updateUI({ agentStatus: status })
+    updateUI({ agentStatus: status })
-    },
+  },
-    onActivity: (activity) => {
+  onActivity: (activity) => {
-      switch (activity.type) {
+    switch (activity.type) {
-        case 'thinking':
+      case 'thinking':
-          showSpinner('Agent 正在思考...')
+        showSpinner('Agent 正在思考...')
-          break
+        break
-        case 'executing':
+      case 'executing':
-          showSpinner(`正在执行: ${activity.tool}`)
+        showSpinner(`正在执行: ${activity.tool}`)
-          break
+        break
-        case 'executed':
+      case 'executed':
-          log(`${activity.tool} 完成，耗时 ${activity.duration}ms`)
+        log(`${activity.tool} 完成，耗时 ${activity.duration}ms`)
-          break
+        break
-        case 'error':
+      case 'error':
-          showError(activity.message)
+        showError(activity.message)
-          break
+        break
-      }
+    }
-    },
+  },
-    onHistoryUpdate: (history) => {
+  onHistoryUpdate: (history) => {
-      renderHistory(history)
+    renderHistory(history)
-    },
+  },
-  }
+})
-)
 ```
 
 ### 停止执行
 
 ```typescript
 // 启动任务
-window.PAGE_AGENT_EXT!.execute('滚动浏览所有页面', llmConfig)
+window.PAGE_AGENT_EXT!.execute('滚动浏览所有页面', {
+  baseURL: 'https://api.openai.com/v1',
+  apiKey: process.env.OPENAI_API_KEY!,
+  model: 'gpt-5.2',
+})
 
 // 稍后停止
 window.PAGE_AGENT_EXT!.dispose()
@@ -258,24 +265,25 @@ import type {
   AgentStatus,
   ExecutionResult,
   HistoricalEvent,
-  LLMConfig,
 } from '@page-agent/core'
 
+interface ExecuteConfig {
+  baseURL: string
+  apiKey: string
+  model: string
+  includeInitialTab?: boolean
+  onStatusChange?: (status: AgentStatus) => void
+  onActivity?: (activity: AgentActivity) => void
+  onHistoryUpdate?: (history: HistoricalEvent[]) => void
+  onDispose?: () => void
+}
+
 declare global {
   interface Window {
     PAGE_AGENT_EXT_VERSION?: string
     PAGE_AGENT_EXT?: {
       version: string
-      execute: (
+      execute: (task: string, config: ExecuteConfig) => Promise<ExecutionResult>
-        task: string,
-        llmConfig: LLMConfig,
-        hooks?: {
-          onStatusChange?: (status: AgentStatus) => void
-          onActivity?: (activity: AgentActivity) => void
-          onHistoryUpdate?: (history: HistoricalEvent[]) => void
-          onDispose?: () => void
-        }
-      ) => Promise<ExecutionResult>
       dispose: () => void
     }
   }

packages/extension/package.json

@@ -1,7 +1,7 @@
 {
     "name": "@page-agent/ext",
     "private": true,
-    "version": "0.1.2",
+    "version": "0.1.4",
     "type": "module",
     "scripts": {
         "dev": "wxt",
@@ -10,23 +10,6 @@
         "postinstall": "wxt prepare"
     },
     "devDependencies": {
-        "@tailwindcss/vite": "^4.1.18",
-        "@types/chrome": "^0.1.34",
-        "@types/react": "^19.2.8",
-        "@types/react-dom": "^19.2.1",
-        "@vitejs/plugin-react-swc": "^4.1.0",
-        "@wxt-dev/module-react": "^1.1.5",
-        "react": "^19.2.4",
-        "react-dom": "^19.2.4",
-        "tailwindcss": "^4.1.14",
-        "tw-animate-css": "^1.4.0",
-        "wxt": "^0.20.13"
-    },
-    "dependencies": {
-        "@page-agent/core": "1.1.0",
-        "@page-agent/llms": "1.1.0",
-        "@page-agent/page-controller": "1.1.0",
-        "@page-agent/ui": "1.1.0",
         "@radix-ui/react-hover-card": "^1.1.15",
         "@radix-ui/react-icons": "^1.3.2",
         "@radix-ui/react-label": "^2.1.8",
@@ -34,14 +17,32 @@
         "@radix-ui/react-slot": "^1.2.4",
         "@radix-ui/react-switch": "^1.2.6",
         "@radix-ui/react-tooltip": "^1.2.8",
+        "@types/chrome": "^0.1.34",
+        "@types/react": "^19.2.13",
+        "@types/react-dom": "^19.2.1",
+        "@wxt-dev/module-react": "^1.1.5",
         "class-variance-authority": "^0.7.1",
         "clsx": "^2.1.1",
+        "idb": "^8.0.3",
         "lucide-react": "^0.563.0",
-        "motion": "^12.30.0",
+        "motion": "^12.34.0",
         "next-themes": "^0.4.6",
+        "react": "^19.2.4",
+        "react-dom": "^19.2.4",
         "rough-notation": "^0.5.1",
-        "simple-icons": "^16.7.0",
+        "simple-icons": "^16.8.0",
         "sonner": "^2.0.7",
-        "tailwind-merge": "^3.4.0"
+        "tailwind-merge": "^3.4.0",
+        "tailwindcss": "^4.1.14",
+        "tw-animate-css": "^1.4.0",
+        "wxt": "^0.20.14"
+    },
+    "dependencies": {
+        "@page-agent/core": "1.2.0",
+        "@page-agent/llms": "1.2.0",
+        "@page-agent/page-controller": "1.2.0",
+        "@page-agent/ui": "1.2.0",
+        "chalk": "^5.6.2",
+        "zod": "^4.3.5"
     }
 }

packages/extension/src/agent/MultiPageAgent.ts

@@ -17,7 +17,7 @@ function detectLanguage(): 'en-US' | 'zh-CN' {
  * - can be used from a side panel or a content script
  */
 export class MultiPageAgent extends PageAgentCore {
-	constructor(config: Omit<PageAgentConfig, 'pageController'>) {
+	constructor(config: Omit<PageAgentConfig, 'pageController'> & { includeInitialTab?: boolean }) {
 		// multi page controller
 		const tabsController = new TabsController()
 		const pageController = new RemotePageController(tabsController)
@@ -31,6 +31,9 @@ export class MultiPageAgent extends PageAgentCore {
 			`Default working language: **${targetLanguage}**`
 		)
 
+		// include initial tab for controlling
+		const includeInitialTab = config.includeInitialTab ?? true
+
 		/**
 		 * When the agent is in side-panel and user closed the side-panel.
 		 * There is no chance for isAgentRunning to be set false.
@@ -47,7 +50,7 @@ export class MultiPageAgent extends PageAgentCore {
 			customSystemPrompt: systemPrompt,
 
 			onBeforeTask: async (agent) => {
-				await tabsController.init(agent.task)
+				await tabsController.init(agent.task, includeInitialTab)
 
 				heartBeatInterval = window.setInterval(() => {
 					chrome.storage.local.set({
@@ -71,6 +74,11 @@ export class MultiPageAgent extends PageAgentCore {
 				})
 			},
 
+			onBeforeStep: async (agent) => {
+				// make sure the current tab is loaded before the step starts
+				await tabsController.waitUntilTabLoaded(tabsController.currentTabId!)
+			},
+
 			onDispose: () => {
 				if (heartBeatInterval) {
 					window.clearInterval(heartBeatInterval)

packages/extension/src/agent/RemotePageController.background.ts

@@ -8,13 +8,21 @@ export function handlePageControlMessage(
 	sender: chrome.runtime.MessageSender,
 	sendResponse: (response: unknown) => void
 ): true | undefined {
+	const PREFIX = '[RemotePageController.background]'
+
+	function debug(...messages: any[]) {
+		console.debug(`\x1b[90m${PREFIX}\x1b[0m`, ...messages)
+	}
+
 	const { action, payload, targetTabId } = message
 
 	if (action === 'get_my_tab_id') {
+		debug('get_my_tab_id', sender.tab?.id)
 		sendResponse({ tabId: sender.tab?.id || null })
 		return
 	}
 
+	// proxy to content script
 	chrome.tabs
 		.sendMessage(targetTabId, {
 			type: 'PAGE_CONTROL',
@@ -25,6 +33,7 @@ export function handlePageControlMessage(
 			sendResponse(result)
 		})
 		.catch((error) => {
+			console.error(PREFIX, error)
 			sendResponse({
 				success: false,
 				error: error instanceof Error ? error.message : String(error),

packages/extension/src/agent/RemotePageController.content.ts

@@ -12,10 +12,14 @@ export function initPageController() {
 		.then((response) => {
 			return (response as { tabId: number | null }).tabId
 		})
+		.catch((error) => {
+			console.error('[RemotePageController.ContentScript]: Failed to get my tab id', error)
+			return null
+		})
 
 	function getPC(): PageController {
 		if (!pageController) {
-			pageController = new PageController({ enableMask: false })
+			pageController = new PageController({ enableMask: false, viewportExpansion: 400 })
 		}
 		return pageController
 	}

packages/extension/src/agent/RemotePageController.ts

@@ -2,13 +2,31 @@ import type { BrowserState } from '@page-agent/page-controller'
 
 import type { TabsController } from './TabsController'
 
+const PREFIX = '[RemotePageController]'
+
+function debug(...messages: any[]) {
+	console.debug(`\x1b[90m${PREFIX}\x1b[0m`, ...messages)
+}
+
+function sendMessage(message: {
+	type: 'PAGE_CONTROL'
+	action: string
+	targetTabId: number
+	payload?: any
+}): Promise<any> {
+	return chrome.runtime.sendMessage(message).catch((error) => {
+		console.error(PREFIX, message.action, error)
+		return null
+	})
+}
+
 /**
  * Agent side page controller.
  * - live in the agent env (extension page or content script)
  * - communicates with remote PageController via sw
  */
 export class RemotePageController {
-	private tabsController: TabsController
+	tabsController: TabsController
 
 	constructor(tabsController: TabsController) {
 		this.tabsController = tabsController
@@ -18,46 +36,46 @@ export class RemotePageController {
 		return this.tabsController.currentTabId
 	}
 
-	async getCurrentUrl(): Promise<string> {
+	private async getCurrentUrl(): Promise<string> {
 		if (!this.currentTabId) return ''
 		const { url } = await this.tabsController.getTabInfo(this.currentTabId)
 		return url || ''
 	}
 
-	async getCurrentTitle(): Promise<string> {
+	private async getCurrentTitle(): Promise<string> {
 		if (!this.currentTabId) return ''
 		const { title } = await this.tabsController.getTabInfo(this.currentTabId)
 		return title || ''
 	}
 
-	get currentTabTitle(): Promise<string> {
-		return this.getCurrentTitle()
-	}
-
 	async getLastUpdateTime(): Promise<number> {
 		if (!this.currentTabId) throw new Error('tabsController not initialized.')
-
+		return sendMessage({
-		return await chrome.runtime.sendMessage({
 			type: 'PAGE_CONTROL',
 			action: 'get_last_update_time',
 			targetTabId: this.currentTabId,
 		})
 	}
 
-	// getBrowserState
 	async getBrowserState(): Promise<BrowserState> {
-		let browserState = {} as BrowserState
+		if (!this.currentTabId) throw new Error('tabsController not initialized.')
 
-		if (!this.currentTabId || !isContentScriptAllowed(await this.currentTabUrl)) {
+		let browserState = {} as BrowserState
+		debug('getBrowserState', this.currentTabId)
+
+		const currentUrl = await this.getCurrentUrl()
+		const currentTitle = await this.getCurrentTitle()
+
+		if (!this.currentTabId || !isContentScriptAllowed(currentUrl)) {
 			browserState = {
-				url: await this.currentTabUrl,
+				url: currentUrl,
-				title: await this.currentTabTitle,
+				title: currentTitle,
 				header: '',
-				content: '(empty page)',
+				content: '(empty page. either current page is not readable or not loaded yet.)',
 				footer: '',
 			}
 		} else {
-			browserState = await chrome.runtime.sendMessage({
+			browserState = await sendMessage({
 				type: 'PAGE_CONTROL',
 				action: 'get_browser_state',
 				targetTabId: this.currentTabId,
@@ -67,61 +85,58 @@ export class RemotePageController {
 		const sum = await this.tabsController.summarizeTabs()
 		browserState.header = sum + '\n\n' + (browserState.header || '')
 
+		debug('getBrowserState: success', this.currentTabId, browserState)
+
 		return browserState
 	}
 
-	// updateTree
 	async updateTree(): Promise<void> {
-		if (!this.currentTabId || !isContentScriptAllowed(await this.currentTabUrl)) {
+		if (!this.currentTabId || !isContentScriptAllowed(await this.getCurrentUrl())) {
 			return
 		}
 
-		await chrome.runtime.sendMessage({
+		await sendMessage({
 			type: 'PAGE_CONTROL',
 			action: 'update_tree',
 			targetTabId: this.currentTabId,
 		})
 	}
 
-	// cleanUpHighlights
 	async cleanUpHighlights(): Promise<void> {
-		if (!this.currentTabId || !isContentScriptAllowed(await this.currentTabUrl)) {
+		if (!this.currentTabId || !isContentScriptAllowed(await this.getCurrentUrl())) {
 			return
 		}
 
-		await chrome.runtime.sendMessage({
+		await sendMessage({
 			type: 'PAGE_CONTROL',
 			action: 'clean_up_highlights',
 			targetTabId: this.currentTabId,
 		})
 	}
 
-	// clickElement
 	async clickElement(...args: any[]): Promise<DomActionReturn> {
-		return this.remoteCallDomAction('click_element', args)
+		const res = await this.remoteCallDomAction('click_element', args)
+		// @note may cause page navigation, wait for 1 second to ensure the page loading started
+		await new Promise((resolve) => setTimeout(resolve, 1000))
+		return res
 	}
 
-	// inputText
 	async inputText(...args: any[]): Promise<DomActionReturn> {
 		return this.remoteCallDomAction('input_text', args)
 	}
 
-	// selectOption
 	async selectOption(...args: any[]): Promise<DomActionReturn> {
 		return this.remoteCallDomAction('select_option', args)
 	}
 
-	// scroll
 	async scroll(...args: any[]): Promise<DomActionReturn> {
 		return this.remoteCallDomAction('scroll', args)
 	}
 
-	// scrollHorizontally
 	async scrollHorizontally(...args: any[]): Promise<DomActionReturn> {
 		return this.remoteCallDomAction('scroll_horizontally', args)
 	}
 
-	// executeJavascript
 	async executeJavascript(...args: any[]): Promise<DomActionReturn> {
 		return this.remoteCallDomAction('execute_javascript', args)
 	}
@@ -133,35 +148,26 @@ export class RemotePageController {
 	/** @note Managed by content script via storage polling. */
 	dispose(): void {}
 
-	private async preCheck() {
-		if (!this.currentTabId) {
-			return 'RemotePageController not initialized.'
-		}
-
-		if (!isContentScriptAllowed(await this.currentTabUrl)) {
-			return 'Operation not allowed on this page. Use open_new_tab to navigate to a web page first.'
-		}
-
-		return null
-	}
-
 	private async remoteCallDomAction(action: string, payload: any[]): Promise<DomActionReturn> {
-		const preCheckError = await this.preCheck()
+		if (!this.currentTabId) {
-		if (preCheckError) {
+			return { success: false, message: 'RemotePageController not initialized.' }
-			return { success: false, message: preCheckError }
 		}
 
-		return await chrome.runtime.sendMessage({
+		if (!isContentScriptAllowed(await this.getCurrentUrl())) {
+			return {
+				success: false,
+				message:
+					'Operation not allowed on this page. Use open_new_tab to navigate to a web page first.',
+			}
+		}
+
+		return sendMessage({
 			type: 'PAGE_CONTROL',
 			action: action,
 			targetTabId: this.currentTabId!,
 			payload,
 		})
 	}
-
-	private get currentTabUrl(): Promise<string> {
-		return this.getCurrentUrl()
-	}
 }
 
 interface DomActionReturn {

packages/extension/src/agent/TabsController.background.ts

@@ -3,6 +3,12 @@
  */
 import type { TabAction } from './TabsController'
 
+const PREFIX = '[TabsController.background]'
+
+function debug(...messages: any[]) {
+	console.debug(`\x1b[90m${PREFIX}\x1b[0m`, ...messages)
+}
+
 export function handleTabControlMessage(
 	message: { type: 'TAB_CONTROL'; action: TabAction; payload: any },
 	sender: chrome.runtime.MessageSender,
@@ -12,10 +18,12 @@ export function handleTabControlMessage(
 
 	switch (action as TabAction) {
 		case 'get_active_tab': {
+			debug('get_active_tab')
 			chrome.tabs
 				.query({ active: true, currentWindow: true })
 				.then((tabs) => {
 					const tabId = tabs.length > 0 ? tabs[0].id || null : null
+					debug('get_active_tab: success', tabId)
 					sendResponse({ success: true, tabId })
 				})
 				.catch((error) => {
@@ -25,11 +33,12 @@ export function handleTabControlMessage(
 		}
 
 		case 'get_tab_info': {
+			debug('get_tab_info', payload)
 			chrome.tabs
 				.get(payload.tabId)
 				.then((tab) => {
-					const result = { title: tab.title || '', url: tab.url || '' }
+					debug('get_tab_info: success', tab)
-					sendResponse(result)
+					sendResponse(tab)
 				})
 				.catch((error) => {
 					sendResponse({ error: error instanceof Error ? error.message : String(error) })
@@ -38,10 +47,11 @@ export function handleTabControlMessage(
 		}
 
 		case 'open_new_tab': {
+			debug('open_new_tab', payload)
 			chrome.tabs
 				.create({ url: payload.url, active: false })
 				.then((newTab) => {
-					// @todo: wait for tab to be fully loaded
+					debug('open_new_tab: success', newTab)
 					sendResponse({ success: true, tabId: newTab.id, windowId: newTab.windowId })
 				})
 				.catch((error) => {
@@ -51,20 +61,22 @@ export function handleTabControlMessage(
 		}
 
 		case 'create_tab_group': {
+			debug('create_tab_group', payload)
 			chrome.tabs
 				.group({ tabIds: payload.tabIds, createProperties: { windowId: payload.windowId } })
 				.then((groupId) => {
-					console.log('Created tab group', groupId)
+					debug('create_tab_group: success', groupId)
 					sendResponse({ success: true, groupId })
 				})
 				.catch((error) => {
-					console.error('Failed to create tab group', error)
+					console.error(PREFIX, 'Failed to create tab group', error)
 					sendResponse({ error: error instanceof Error ? error.message : String(error) })
 				})
 			return true // async response
 		}
 
 		case 'update_tab_group': {
+			debug('update_tab_group', payload)
 			chrome.tabGroups
 				.update(payload.groupId, payload.properties)
 				.then(() => {
@@ -77,6 +89,7 @@ export function handleTabControlMessage(
 		}
 
 		case 'add_tab_to_group': {
+			debug('add_tab_to_group', payload)
 			chrome.tabs
 				.group({ tabIds: payload.tabId, groupId: payload.groupId })
 				.then(() => {
@@ -89,6 +102,7 @@ export function handleTabControlMessage(
 		}
 
 		case 'close_tab': {
+			debug('close_tab', payload)
 			chrome.tabs
 				.remove(payload.tabId)
 				.then(() => {
@@ -105,3 +119,42 @@ export function handleTabControlMessage(
 			return
 	}
 }
+
+export function setupTabChangeEvents() {
+	console.log('[TabsController.background] setupTabChangeEvents')
+
+	chrome.tabs.onCreated.addListener((tab) => {
+		debug('onCreated', tab)
+		chrome.runtime
+			.sendMessage({ type: 'TAB_CHANGE', action: 'created', payload: { tab } })
+			.catch((error) => {
+				debug('onCreated error:', error)
+			})
+	})
+
+	chrome.tabs.onRemoved.addListener((tabId, removeInfo) => {
+		debug('onRemoved', tabId, removeInfo)
+		chrome.runtime
+			.sendMessage({
+				type: 'TAB_CHANGE',
+				action: 'removed',
+				payload: { tabId, removeInfo },
+			})
+			.catch((error) => {
+				debug('onRemoved error:', error)
+			})
+	})
+
+	chrome.tabs.onUpdated.addListener((tabId, changeInfo, tab) => {
+		debug('onUpdated', tabId, changeInfo)
+		chrome.runtime
+			.sendMessage({
+				type: 'TAB_CHANGE',
+				action: 'updated',
+				payload: { tabId, changeInfo, tab },
+			})
+			.catch((error) => {
+				debug('onUpdated error:', error)
+			})
+	})
+}

packages/extension/src/agent/TabsController.ts

@@ -1,3 +1,20 @@
+const PREFIX = '[TabsController]'
+
+function debug(...messages: any[]) {
+	console.debug(`\x1b[90m${PREFIX}\x1b[0m`, ...messages)
+}
+
+function sendMessage(message: {
+	type: 'TAB_CONTROL'
+	action: TabAction
+	payload?: any
+}): Promise<any> {
+	return chrome.runtime.sendMessage(message).catch((error) => {
+		console.error(PREFIX, message.action, error)
+		return null
+	})
+}
+
 /**
  * Controller for managing browser tabs.
  * - live in the agent env (extension page or content script)
@@ -12,7 +29,9 @@ export class TabsController extends EventTarget {
 	private task: string = ''
 	private windowId: number | null = null
 
-	async init(task: string) {
+	async init(task: string, includeInitialTab: boolean = true) {
+		debug('init', task, includeInitialTab)
+
 		this.task = task
 		this.tabs = []
 		this.currentTabId = null
@@ -20,23 +39,36 @@ export class TabsController extends EventTarget {
 		this.initialTabId = null
 		this.windowId = null
 
-		const result = await chrome.runtime.sendMessage({
+		const result = await sendMessage({
 			type: 'TAB_CONTROL',
 			action: 'get_active_tab',
 		})
 
 		this.initialTabId = result.tabId
-		this.currentTabId = result.tabId
-
-		this.tabs.push({
-			id: result.tabId,
-			isInitial: true,
-		})
 
 		if (!this.initialTabId) {
 			throw new Error('Failed to get initial tab ID')
 		}
 
+		if (includeInitialTab) {
+			this.currentTabId = this.initialTabId
+
+			// update tab status immediately
+			const info = await sendMessage({
+				type: 'TAB_CONTROL',
+				action: 'get_tab_info',
+				payload: { tabId: this.initialTabId },
+			})
+
+			this.tabs.push({
+				id: result.tabId,
+				isInitial: true,
+				url: info.url,
+				title: info.title,
+				status: info.status,
+			})
+		}
+
 		await this.updateCurrentTabId(this.currentTabId)
 
 		const tabChangeHandler = (message: any): void => {
@@ -68,6 +100,14 @@ export class TabsController extends EventTarget {
 						}
 					}
 				}
+			} else if (message.action === 'updated') {
+				const { tabId, tab } = message.payload as { tabId: number; tab: chrome.tabs.Tab }
+				const targetTab = this.tabs.find((t) => t.id === tabId)
+				if (targetTab) {
+					targetTab.url = tab.url
+					targetTab.title = tab.title
+					targetTab.status = tab.status
+				}
 			}
 		}
 
@@ -78,8 +118,10 @@ export class TabsController extends EventTarget {
 		})
 	}
 
-	async openNewTab(url: string): Promise<{ success: boolean; tabId: number; message: string }> {
+	async openNewTab(url: string): Promise<string> {
-		const result = await chrome.runtime.sendMessage({
+		debug('openNewTab', url)
+
+		const result = await sendMessage({
 			type: 'TAB_CONTROL',
 			action: 'open_new_tab',
 			payload: { url },
@@ -102,7 +144,7 @@ export class TabsController extends EventTarget {
 		await this.switchToTab(tabId)
 
 		if (!this.tabGroupId) {
-			const result = await chrome.runtime.sendMessage({
+			const result = await sendMessage({
 				type: 'TAB_CONTROL',
 				action: 'create_tab_group',
 				payload: { tabIds: [tabId], windowId: this.windowId },
@@ -116,7 +158,7 @@ export class TabsController extends EventTarget {
 
 			this.tabGroupId = groupId
 
-			await chrome.runtime.sendMessage({
+			await sendMessage({
 				type: 'TAB_CONTROL',
 				action: 'update_tab_group',
 				payload: {
@@ -129,57 +171,43 @@ export class TabsController extends EventTarget {
 				},
 			})
 		} else {
-			await chrome.runtime.sendMessage({
+			await sendMessage({
 				type: 'TAB_CONTROL',
 				action: 'add_tab_to_group',
 				payload: { tabId: result.tabId, groupId: this.tabGroupId },
 			})
 		}
 
-		// wait for the new tab to be fully loaded
+		await this.waitUntilTabLoaded(tabId)
-		// @todo
-		await new Promise((resolve) => setTimeout(resolve, 500))
 
-		return {
+		return `✅ Opened new tab ID ${tabId} with URL ${url}`
-			success: true,
-			tabId,
-			message: `Opened new tab ID ${tabId} with URL ${url}`,
-		}
 	}
 
-	async switchToTab(tabId: number): Promise<{ success: boolean; message: string }> {
+	async switchToTab(tabId: number): Promise<string> {
+		debug('switchToTab', tabId)
+
 		const targetTab = this.tabs.find((t) => t.id === tabId)
 		if (!targetTab) {
-			return {
+			throw new Error(`Tab ID ${tabId} not found in tab list.`)
-				success: false,
-				message: `Tab ID ${tabId} not found in tab list.`,
-			}
 		}
 
 		await this.updateCurrentTabId(tabId)
 
-		return {
+		return `✅ Switched to tab ID ${tabId}.`
-			success: true,
-			message: `Switched to tab ID ${tabId}.`,
-		}
 	}
 
-	async closeTab(tabId: number): Promise<{ success: boolean; message: string }> {
+	async closeTab(tabId: number): Promise<string> {
+		debug('closeTab', tabId)
+
 		const targetTab = this.tabs.find((t) => t.id === tabId)
 		if (!targetTab) {
-			return {
+			throw new Error(`Tab ID ${tabId} not found in tab list.`)
-				success: false,
-				message: `Tab ID ${tabId} not found in tab list.`,
-			}
 		}
 		if (targetTab.isInitial) {
-			return {
+			throw new Error(`Cannot close the initial tab ID ${tabId}.`)
-				success: false,
-				message: `Cannot close the initial tab ID ${tabId}.`,
-			}
 		}
 
-		const result = await chrome.runtime.sendMessage({
+		const result = await sendMessage({
 			type: 'TAB_CONTROL',
 			action: 'close_tab',
 			payload: { tabId },
@@ -196,29 +224,39 @@ export class TabsController extends EventTarget {
 				}
 			}
 
-			return {
+			return `✅ Closed tab ID ${tabId}.`
-				success: true,
-				message: `Closed tab ID ${tabId}.`,
-			}
 		} else {
-			return {
+			throw new Error(`Failed to close tab ID ${tabId}: ${result.error}`)
-				success: false,
-				message: `Failed to close tab ID ${tabId}: ${result.error}`,
-			}
 		}
 	}
 
 	async updateCurrentTabId(tabId: number | null) {
+		debug('updateCurrentTabId', tabId)
+
 		this.currentTabId = tabId
 		await chrome.storage.local.set({ currentTabId: tabId })
 	}
 
 	async getTabInfo(tabId: number): Promise<{ title: string; url: string }> {
-		const result = await chrome.runtime.sendMessage({
+		// use cached tab info if available
+		const tabMeta = this.tabs.find((t) => t.id === tabId)
+		if (tabMeta && tabMeta.url && tabMeta.title) {
+			return { title: tabMeta.title, url: tabMeta.url }
+		}
+
+		// otherwise, pull the latest tab info from the background script
+		debug('getTabInfo: pulling from background script', tabId)
+		const result = await sendMessage({
 			type: 'TAB_CONTROL',
 			action: 'get_tab_info',
 			payload: { tabId },
 		})
+
+		if (tabMeta) {
+			tabMeta.url = result.url
+			tabMeta.title = result.title
+		}
+
 		return result
 	}
 
@@ -230,9 +268,24 @@ export class TabsController extends EventTarget {
 				`| ${tab.id} | ${url} | ${title} | ${this.currentTabId === tab.id ? '✅' : ''} |`
 			)
 		}
+		if (!this.tabs.length) {
+			summaries.push('\nNo tabs available. Open a tab if needed.')
+		}
+
 		return summaries.join('\n')
 	}
 
+	async waitUntilTabLoaded(tabId: number): Promise<void> {
+		const tab = this.tabs.find((t) => t.id === tabId)
+		if (!tab) throw new Error(`Tab ID ${tabId} not found in tab list.`)
+
+		if (tab.status === 'unloaded') throw new Error(`Tab ID ${tabId} is unloaded.`)
+		if (tab.status === 'complete') return
+
+		debug('waitUntilTabLoaded', tabId)
+		await waitUntil(() => tab.status === 'complete', 4_000)
+	}
+
 	dispose() {
 		this.dispatchEvent(new Event('dispose'))
 	}
@@ -251,6 +304,9 @@ export type TabAction =
 interface TabMeta {
 	id: number
 	isInitial: boolean
+	url?: string
+	title?: string
+	status?: 'loading' | 'unloaded' | 'complete'
 }
 
 const TAB_GROUP_COLORS = [
@@ -269,3 +325,33 @@ type TabGroupColor = (typeof TAB_GROUP_COLORS)[number]
 function randomColor(): TabGroupColor {
 	return TAB_GROUP_COLORS[Math.floor(Math.random() * TAB_GROUP_COLORS.length)]
 }
+
+/**
+ * Wait until condition becomes true
+ * @returns Returns when condition becomes true, throws otherwise
+ * @param timeoutMS Timeout in milliseconds, default 1 minutes, throws error on timeout
+ * @param error Error object to reject on timeout. If not provided, will resolve with false
+ */
+export async function waitUntil(
+	check: () => boolean | Promise<boolean>,
+	timeoutMS = 60_000,
+	error?: string
+): Promise<boolean> {
+	if (await check()) return true
+
+	return new Promise((resolve, reject) => {
+		const start = Date.now()
+		const poll = async () => {
+			if (await check()) return resolve(true)
+			if (Date.now() - start > timeoutMS) {
+				if (error) {
+					return reject(new Error(error))
+				} else {
+					return resolve(false)
+				}
+			}
+			setTimeout(poll, 100)
+		}
+		setTimeout(poll, 100)
+	})
+}

packages/extension/src/agent/system_prompt.md

@@ -120,7 +120,6 @@ Exhibit the following reasoning patterns to successfully achieve the <user_reque
 Here are examples of good output patterns. Use them as reference but never copy them directly.
 
 <evaluation_examples>
-- Positive Examples:
 "evaluation_previous_goal": "Successfully navigated to the product page and found the target information. Verdict: Success"
 "evaluation_previous_goal": "Clicked the login button and user authentication form appeared. Verdict: Success"
 </evaluation_examples>
@@ -131,17 +130,16 @@ Here are examples of good output patterns. Use them as reference but never copy
 
 <next_goal_examples>
 "next_goal": "Click on the 'Add to Cart' button to proceed with the purchase flow."
-"next_goal": "Extract details from the first item on the page."
 </next_goal_examples>
 </examples>
 
 <output>
-You must ALWAYS respond with a valid JSON in this exact format:
-
 {
   "evaluation_previous_goal": "Concise one-sentence analysis of your last action. Clearly state success, failure, or uncertain.",
   "memory": "1-3 concise sentences of specific memory of this step and overall progress. You should put here everything that will help you track progress in future steps. Like counting pages visited, items found, etc.",
-  "next_goal": "State the next immediate goal and action to achieve it, in one clear sentence."
+  "next_goal": "State the next immediate goal and action to achieve it, in one clear sentence.",
-  "action":{"action_name": {// action-specific parameter}}
+  "action":{
+    "Action name": {// Action parameters}
+  }
 }
 </output>

packages/extension/src/agent/tabTools.ts

@@ -6,7 +6,7 @@
  * - switch_to_tab: Switch to an existing tab
  * - close_tab: Close a tab (optionally switch to another)
  */
-import zod from 'zod'
+import * as zod from 'zod'
 
 import type { TabsController } from './TabsController'
 
@@ -31,8 +31,11 @@ export function createTabTools(tabsController: TabsController): Record<string, T
 			}),
 			execute: async (input: unknown) => {
 				const { url } = input as { url: string }
-				const result = await tabsController.openNewTab(url)
+				try {
-				return result.message
+					return await tabsController.openNewTab(url)
+				} catch (error) {
+					return `❌ Failed: ${error instanceof Error ? error.message : String(error)}`
+				}
 			},
 		},
 
@@ -44,7 +47,11 @@ export function createTabTools(tabsController: TabsController): Record<string, T
 			}),
 			execute: async (input: unknown) => {
 				const { tab_id } = input as { tab_id: number }
-				return (await tabsController.switchToTab(tab_id)).message
+				try {
+					return await tabsController.switchToTab(tab_id)
+				} catch (error) {
+					return `❌ Failed: ${error instanceof Error ? error.message : String(error)}`
+				}
 			},
 		},
 
@@ -56,7 +63,11 @@ export function createTabTools(tabsController: TabsController): Record<string, T
 			}),
 			execute: async (input: unknown) => {
 				const { tab_id } = input as { tab_id: number }
-				return (await tabsController.closeTab(tab_id)).message
+				try {
+					return await tabsController.closeTab(tab_id)
+				} catch (error) {
+					return `❌ Failed: ${error instanceof Error ? error.message : String(error)}`
+				}
 			},
 		},
 	}

packages/extension/src/agent/useAgent.ts

@@ -75,7 +75,7 @@ export function useAgent(): UseAgentResult {
 
 	const execute = useCallback(async (task: string) => {
 		const agent = agentRef.current
-		console.log('Fire task:', task, agent)
+		console.log('🚀 [useAgent] start executing task:', task)
 		if (!agent) throw new Error('Agent not initialized')
 
 		setCurrentTask(task)

packages/extension/src/entrypoints/background.ts

@@ -1,24 +1,13 @@
 import { handlePageControlMessage } from '@/agent/RemotePageController.background'
-import { handleTabControlMessage } from '@/agent/TabsController.background'
+import { handleTabControlMessage, setupTabChangeEvents } from '@/agent/TabsController.background'
-
-chrome.runtime.onMessage.addListener((message, sender, sendResponse): true | undefined => {
-	if (message.type === 'TAB_CONTROL') {
-		return handleTabControlMessage(message, sender, sendResponse)
-	} else if (message.type === 'PAGE_CONTROL') {
-		return handlePageControlMessage(message, sender, sendResponse)
-	} else {
-		sendResponse({ error: 'Unknown message type' })
-		return
-	}
-})
-
-// ============================================================================
-// Extension Setup
-// ============================================================================
 
 export default defineBackground(() => {
 	console.log('[Background] Service Worker started')
 
+	// tab change events
+
+	setupTabChangeEvents()
+
 	// generate user auth token
 
 	chrome.storage.local.get('PageAgentExtUserAuthToken').then((result) => {
@@ -28,21 +17,20 @@ export default defineBackground(() => {
 		chrome.storage.local.set({ PageAgentExtUserAuthToken: userAuthToken })
 	})
 
+	// message proxy
+
+	chrome.runtime.onMessage.addListener((message, sender, sendResponse): true | undefined => {
+		if (message.type === 'TAB_CONTROL') {
+			return handleTabControlMessage(message, sender, sendResponse)
+		} else if (message.type === 'PAGE_CONTROL') {
+			return handlePageControlMessage(message, sender, sendResponse)
+		} else {
+			sendResponse({ error: 'Unknown message type' })
+			return
+		}
+	})
+
 	// setup
 
 	chrome.sidePanel.setPanelBehavior({ openPanelOnActionClick: true }).catch(() => {})
-
-	// Tab change events
-
-	chrome.tabs.onCreated.addListener((tab) => {
-		chrome.runtime.sendMessage({ type: 'TAB_CHANGE', action: 'created', payload: { tab } })
-	})
-
-	chrome.tabs.onRemoved.addListener((tabId, removeInfo) => {
-		chrome.runtime.sendMessage({
-			type: 'TAB_CHANGE',
-			action: 'removed',
-			payload: { tabId, removeInfo },
-		})
-	})
 })

packages/extension/src/entrypoints/content.ts

@@ -6,7 +6,7 @@ const DEBUG_PREFIX = '[Content]'
 
 export default defineContentScript({
 	matches: ['<all_urls>'],
-	runAt: 'document_idle',
+	runAt: 'document_end',
 
 	main() {
 		console.debug(`${DEBUG_PREFIX} Loaded on ${window.location.href}`)
@@ -69,11 +69,11 @@ async function exposeAgentToPage() {
 				}
 
 				try {
-					const { task, llmConfig } = payload
+					const { task, config } = payload
 
 					// create when used
 
-					multiPageAgent = new MultiPageAgent(llmConfig)
+					multiPageAgent = new MultiPageAgent(config)
 
 					// events
 

packages/extension/src/entrypoints/main-world.ts

@@ -1,19 +1,24 @@
 import type { AgentActivity, AgentStatus, ExecutionResult, HistoricalEvent } from '@page-agent/core'
-import type { LLMConfig } from '@page-agent/llms'
 
-export interface ExecuteHooks {
+export type Execute = (task: string, config: ExecuteConfig) => Promise<ExecutionResult>
+
+export interface ExecuteConfig {
+	baseURL: string
+	apiKey: string
+	model: string
+
+	/**
+	 * Whether to include the initial tab (that holds this main world script) in the task.
+	 * @default true
+	 */
+	includeInitialTab?: boolean
+
 	onStatusChange?: (status: AgentStatus) => void
 	onActivity?: (activity: AgentActivity) => void
 	onHistoryUpdate?: (history: HistoricalEvent[]) => void
 	onDispose?: () => void
 }
 
-export type Execute = (
-	task: string,
-	llmConfig: LLMConfig,
-	hooks?: ExecuteHooks
-) => Promise<ExecutionResult>
-
 export default defineUnlistedScript(() => {
 	let _lastId = 0
 	function getId() {
@@ -21,13 +26,13 @@ export default defineUnlistedScript(() => {
 		return _lastId
 	}
 
-	const execute: Execute = async (task, llmConfig, hooks) => {
+	const execute: Execute = async (task, config) => {
 		if (typeof task !== 'string') throw new Error('Task must be a string')
 		if (task.trim().length === 0) throw new Error('Task cannot be empty')
-		if (!llmConfig) throw new Error('LLM config is required')
+		if (!config) throw new Error('Config is required')
-		if (!llmConfig.baseURL) throw new Error('LLM config must have a baseURL')
+		if (!config.baseURL) throw new Error('Config must have a baseURL')
-		if (!llmConfig.apiKey) throw new Error('LLM config must have an apiKey')
+		if (!config.apiKey) throw new Error('Config must have an apiKey')
-		if (!llmConfig.model) throw new Error('LLM config must have a model')
+		if (!config.model) throw new Error('Config must have a model')
 
 		const id = getId()
 
@@ -40,30 +45,31 @@ export default defineUnlistedScript(() => {
 
 				// events
 
-				if (data.action === 'status_change_event' && hooks?.onStatusChange) {
+				if (data.action === 'status_change_event' && config.onStatusChange) {
-					hooks.onStatusChange(data.payload)
+					config.onStatusChange(data.payload)
 					return
 				}
 
-				if (data.action === 'activity_event' && hooks?.onActivity) {
+				if (data.action === 'activity_event' && config.onActivity) {
-					hooks.onActivity(data.payload)
+					config.onActivity(data.payload)
 					return
 				}
 
-				if (data.action === 'history_change_event' && hooks?.onHistoryUpdate) {
+				if (data.action === 'history_change_event' && config.onHistoryUpdate) {
-					hooks.onHistoryUpdate(data.payload)
+					config.onHistoryUpdate(data.payload)
 					return
 				}
 
-				if (data.action === 'dispose_event' && hooks?.onDispose) {
+				if (data.action === 'dispose_event' && config.onDispose) {
-					hooks.onDispose()
+					config.onDispose()
+					window.removeEventListener('message', handleMessage)
 					return
 				}
 
-				// result
-
 				if (data.action !== 'execute_result') return
 
+				// execute_result
+
 				window.removeEventListener('message', handleMessage)
 
 				if (data.error) {
@@ -73,6 +79,7 @@ export default defineUnlistedScript(() => {
 				}
 			}
 
+			// @note will be removed on dispose or result
 			window.addEventListener('message', handleMessage)
 		})
 
@@ -81,7 +88,15 @@ export default defineUnlistedScript(() => {
 				channel: 'PAGE_AGENT_EXT_REQUEST',
 				id,
 				action: 'execute',
-				payload: { task, llmConfig },
+				payload: {
+					task,
+					config: {
+						baseURL: config.baseURL,
+						apiKey: config.apiKey,
+						model: config.model,
+						includeInitialTab: config.includeInitialTab,
+					},
+				},
 			},
 			'*'
 		)

packages/extension/src/entrypoints/sidepanel/App.tsx

@@ -1,4 +1,4 @@
-import { Send, Settings, Square } from 'lucide-react'
+import { History, Send, Settings, Square } from 'lucide-react'
 import { useCallback, useEffect, useRef, useState } from 'react'
 
 import { Button } from '@/components/ui/button'
@@ -8,20 +8,47 @@ import {
 	InputGroupButton,
 	InputGroupTextarea,
 } from '@/components/ui/input-group'
+import { saveSession } from '@/lib/db'
 
 import { useAgent } from '../../agent/useAgent'
 import { ConfigPanel } from './components/ConfigPanel'
+import { HistoryDetail } from './components/HistoryDetail'
+import { HistoryList } from './components/HistoryList'
 import { ActivityCard, EventCard } from './components/cards'
 import { EmptyState, Logo, StatusDot } from './components/misc'
 
+type View =
+	| { name: 'chat' }
+	| { name: 'config' }
+	| { name: 'history' }
+	| { name: 'history-detail'; sessionId: string }
+
 export default function App() {
-	const [showConfig, setShowConfig] = useState(false)
+	const [view, setView] = useState<View>({ name: 'chat' })
 	const [inputValue, setInputValue] = useState('')
 	const historyRef = useRef<HTMLDivElement>(null)
 	const textareaRef = useRef<HTMLTextAreaElement>(null)
 
 	const { status, history, activity, currentTask, config, execute, stop, configure } = useAgent()
 
+	// Persist session when task finishes
+	const prevStatusRef = useRef(status)
+	useEffect(() => {
+		const prev = prevStatusRef.current
+		prevStatusRef.current = status
+
+		if (
+			prev === 'running' &&
+			(status === 'completed' || status === 'error') &&
+			history.length > 0 &&
+			currentTask
+		) {
+			saveSession({ task: currentTask, history, status }).catch((err) =>
+				console.error('[SidePanel] Failed to save session:', err)
+			)
+		}
+	}, [status, history, currentTask])
+
 	// Auto-scroll to bottom on new events
 	useEffect(() => {
 		if (historyRef.current) {
@@ -56,19 +83,36 @@ export default function App() {
 		}
 	}
 
-	if (showConfig) {
+	// --- View routing ---
+
+	if (view.name === 'config') {
 		return (
 			<ConfigPanel
 				config={config}
 				onSave={async (newConfig) => {
 					await configure(newConfig)
-					setShowConfig(false)
+					setView({ name: 'chat' })
 				}}
-				onClose={() => setShowConfig(false)}
+				onClose={() => setView({ name: 'chat' })}
 			/>
 		)
 	}
 
+	if (view.name === 'history') {
+		return (
+			<HistoryList
+				onSelect={(id) => setView({ name: 'history-detail', sessionId: id })}
+				onBack={() => setView({ name: 'chat' })}
+			/>
+		)
+	}
+
+	if (view.name === 'history-detail') {
+		return <HistoryDetail sessionId={view.sessionId} onBack={() => setView({ name: 'history' })} />
+	}
+
+	// --- Chat view ---
+
 	const isRunning = status === 'running'
 	const showEmptyState = !currentTask && history.length === 0 && !isRunning
 
@@ -85,7 +129,15 @@ export default function App() {
 					<Button
 						variant="ghost"
 						size="icon-sm"
-						onClick={() => setShowConfig(true)}
+						onClick={() => setView({ name: 'history' })}
+						className="cursor-pointer"
+					>
+						<History className="size-3.5" />
+					</Button>
+					<Button
+						variant="ghost"
+						size="icon-sm"
+						onClick={() => setView({ name: 'config' })}
 						className="cursor-pointer"
 					>
 						<Settings className="size-3.5" />

packages/extension/src/entrypoints/sidepanel/components/HistoryDetail.tsx

@@ -0,0 +1,51 @@
+import { ArrowLeft } from 'lucide-react'
+import { useEffect, useState } from 'react'
+
+import { Button } from '@/components/ui/button'
+import { type SessionRecord, getSession } from '@/lib/db'
+
+import { EventCard } from './cards'
+
+export function HistoryDetail({ sessionId, onBack }: { sessionId: string; onBack: () => void }) {
+	const [session, setSession] = useState<SessionRecord | null>(null)
+
+	useEffect(() => {
+		getSession(sessionId).then((s) => setSession(s ?? null))
+	}, [sessionId])
+
+	if (!session) {
+		return (
+			<div className="flex items-center justify-center h-screen text-xs text-muted-foreground">
+				Loading...
+			</div>
+		)
+	}
+
+	return (
+		<div className="flex flex-col h-screen bg-background">
+			{/* Header */}
+			<header className="flex items-center gap-2 border-b px-3 py-2">
+				<Button variant="ghost" size="icon-sm" onClick={onBack} className="cursor-pointer">
+					<ArrowLeft className="size-3.5" />
+				</Button>
+				<span className="text-sm font-medium truncate">History</span>
+			</header>
+
+			{/* Task */}
+			<div className="border-b px-3 py-2 bg-muted/30">
+				<div className="text-[10px] text-muted-foreground uppercase tracking-wide">Task</div>
+				<div className="text-xs font-medium" title={session.task}>
+					{session.task}
+				</div>
+			</div>
+
+			{/* Events (read-only) */}
+			<div className="flex-1 overflow-y-auto p-3 space-y-2">
+				{session.history.map((event, index) => (
+					// eslint-disable-next-line react-x/no-array-index-key
+					<EventCard key={index} event={event} />
+				))}
+			</div>
+		</div>
+	)
+}

packages/extension/src/entrypoints/sidepanel/components/HistoryList.tsx

@@ -0,0 +1,119 @@
+import { ArrowLeft, CheckCircle, Trash2, XCircle } from 'lucide-react'
+import { useCallback, useEffect, useState } from 'react'
+
+import { Button } from '@/components/ui/button'
+import { type SessionRecord, clearSessions, deleteSession, listSessions } from '@/lib/db'
+
+function timeAgo(ts: number): string {
+	const seconds = Math.floor((Date.now() - ts) / 1000)
+	if (seconds < 60) return 'just now'
+	const minutes = Math.floor(seconds / 60)
+	if (minutes < 60) return `${minutes}m ago`
+	const hours = Math.floor(minutes / 60)
+	if (hours < 24) return `${hours}h ago`
+	const days = Math.floor(hours / 24)
+	return `${days}d ago`
+}
+
+export function HistoryList({
+	onSelect,
+	onBack,
+}: {
+	onSelect: (id: string) => void
+	onBack: () => void
+}) {
+	const [sessions, setSessions] = useState<SessionRecord[]>([])
+	const [loading, setLoading] = useState(true)
+
+	const load = useCallback(async () => {
+		setSessions(await listSessions())
+		setLoading(false)
+	}, [])
+
+	useEffect(() => {
+		// eslint-disable-next-line react-hooks/set-state-in-effect
+		load()
+	}, [load])
+
+	const handleDelete = async (e: React.MouseEvent, id: string) => {
+		e.stopPropagation()
+		await deleteSession(id)
+		setSessions((prev) => prev.filter((s) => s.id !== id))
+	}
+
+	return (
+		<div className="flex flex-col h-screen bg-background">
+			{/* Header */}
+			<header className="flex items-center gap-2 border-b px-3 py-2">
+				<Button variant="ghost" size="icon-sm" onClick={onBack} className="cursor-pointer">
+					<ArrowLeft className="size-3.5" />
+				</Button>
+				<span className="text-sm font-medium flex-1">History</span>
+				{sessions.length > 0 && (
+					<Button
+						variant="ghost"
+						size="sm"
+						onClick={async () => {
+							await clearSessions()
+							setSessions([])
+						}}
+						className="text-[10px] text-muted-foreground hover:text-destructive cursor-pointer h-6 px-2"
+					>
+						<Trash2 className="size-3 mr-1" />
+						Clear All
+					</Button>
+				)}
+			</header>
+
+			{/* List */}
+			<div className="flex-1 overflow-y-auto">
+				{loading && (
+					<div className="flex items-center justify-center h-32 text-xs text-muted-foreground">
+						Loading...
+					</div>
+				)}
+
+				{!loading && sessions.length === 0 && (
+					<div className="flex items-center justify-center h-32 text-xs text-muted-foreground">
+						No history yet
+					</div>
+				)}
+
+				{sessions.map((session) => (
+					<div
+						key={session.id}
+						role="button"
+						tabIndex={0}
+						onClick={() => onSelect(session.id)}
+						onKeyDown={(e) => e.key === 'Enter' && onSelect(session.id)}
+						className="w-full text-left px-3 py-2.5 border-b hover:bg-muted/50 transition-colors cursor-pointer flex items-start gap-2 group"
+					>
+						{/* Status icon */}
+						{session.status === 'completed' ? (
+							<CheckCircle className="size-3.5 text-green-500 shrink-0 mt-0.5" />
+						) : (
+							<XCircle className="size-3.5 text-destructive shrink-0 mt-0.5" />
+						)}
+
+						{/* Content */}
+						<div className="flex-1 min-w-0">
+							<p className="text-xs font-medium truncate">{session.task}</p>
+							<p className="text-[10px] text-muted-foreground mt-0.5">
+								{timeAgo(session.createdAt)} · {session.history.length} steps
+							</p>
+						</div>
+
+						{/* Delete */}
+						<button
+							type="button"
+							onClick={(e) => handleDelete(e, session.id)}
+							className="opacity-0 group-hover:opacity-100 transition-opacity p-1 hover:text-destructive cursor-pointer shrink-0"
+						>
+							<Trash2 className="size-3" />
+						</button>
+					</div>
+				))}
+			</div>
+		</div>
+	)
+}

packages/extension/src/lib/db.ts

@@ -0,0 +1,70 @@
+import type { HistoricalEvent } from '@page-agent/core'
+import { type DBSchema, type IDBPDatabase, openDB } from 'idb'
+
+const DB_NAME = 'page-agent-ext'
+const DB_VERSION = 1
+
+export interface SessionRecord {
+	id: string
+	task: string
+	history: HistoricalEvent[]
+	status: 'completed' | 'error'
+	createdAt: number
+}
+
+interface PageAgentDB extends DBSchema {
+	sessions: {
+		key: string
+		value: SessionRecord
+		indexes: { 'by-created': number }
+	}
+}
+
+let dbPromise: Promise<IDBPDatabase<PageAgentDB>> | null = null
+
+function getDB() {
+	if (!dbPromise) {
+		dbPromise = openDB<PageAgentDB>(DB_NAME, DB_VERSION, {
+			upgrade(db) {
+				const store = db.createObjectStore('sessions', { keyPath: 'id' })
+				store.createIndex('by-created', 'createdAt')
+			},
+		})
+	}
+	return dbPromise
+}
+
+export async function saveSession(
+	session: Omit<SessionRecord, 'id' | 'createdAt'>
+): Promise<SessionRecord> {
+	const db = await getDB()
+	const record: SessionRecord = {
+		...session,
+		id: crypto.randomUUID(),
+		createdAt: Date.now(),
+	}
+	await db.put('sessions', record)
+	return record
+}
+
+/** List sessions, newest first */
+export async function listSessions(): Promise<SessionRecord[]> {
+	const db = await getDB()
+	const all = await db.getAllFromIndex('sessions', 'by-created')
+	return all.reverse()
+}
+
+export async function getSession(id: string): Promise<SessionRecord | undefined> {
+	const db = await getDB()
+	return db.get('sessions', id)
+}
+
+export async function deleteSession(id: string): Promise<void> {
+	const db = await getDB()
+	await db.delete('sessions', id)
+}
+
+export async function clearSessions(): Promise<void> {
+	const db = await getDB()
+	await db.clear('sessions')
+}

packages/llms/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@page-agent/llms",
-    "version": "1.1.0",
+    "version": "1.2.0",
     "type": "module",
     "main": "./dist/lib/page-agent-llms.js",
     "module": "./dist/lib/page-agent-llms.js",

packages/llms/src/index.ts

@@ -49,6 +49,9 @@ export class LLM extends EventTarget {
 	): Promise<InvokeResult> {
 		return await withRetry(
 			async () => {
+				// in case user aborted before invoking
+				if (abortSignal.aborted) throw new Error('AbortError')
+
 				const result = await this.client.invoke(messages, tools, abortSignal, options)
 
 				return result

packages/llms/src/utils.ts

@@ -2,7 +2,7 @@
  * Utility functions for LLM integration
  */
 import chalk from 'chalk'
-import { z } from 'zod'
+import * as z from 'zod'
 
 import type { Tool } from './types'
 
@@ -74,6 +74,10 @@ export function modelPatch(body: Record<string, any>) {
 		} else if (modelName.startsWith('gpt-51')) {
 			debug('Applying GPT-51 patch: disable reasoning')
 			body.reasoning_effort = 'none'
+		} else if (modelName.startsWith('gpt-5-mini')) {
+			debug('Applying GPT-5-mini patch: set reasoning effort to low, temperature to 1')
+			body.reasoning_effort = 'low'
+			body.temperature = 1
 		} else if (modelName.startsWith('gpt-5')) {
 			debug('Applying GPT-5 patch: set reasoning effort to low')
 			body.reasoning_effort = 'low'

packages/page-agent/package.json

@@ -1,7 +1,7 @@
 {
     "name": "page-agent",
     "private": false,
-    "version": "1.1.0",
+    "version": "1.2.0",
     "type": "module",
     "main": "./dist/esm/page-agent.js",
     "module": "./dist/esm/page-agent.js",
@@ -44,11 +44,11 @@
         "postpublish": "node -e \"['README.md','LICENSE'].forEach(f=>{try{require('fs').unlinkSync(f)}catch{}})\""
     },
     "dependencies": {
+        "@page-agent/core": "1.2.0",
+        "@page-agent/llms": "1.2.0",
+        "@page-agent/page-controller": "1.2.0",
+        "@page-agent/ui": "1.2.0",
         "chalk": "^5.6.2",
-        "zod": "^4.3.5",
+        "zod": "^4.3.5"
-        "@page-agent/llms": "1.1.0",
-        "@page-agent/page-controller": "1.1.0",
-        "@page-agent/core": "1.1.0",
-        "@page-agent/ui": "1.1.0"
     }
 }

packages/page-agent/vite.iife.config.js

@@ -3,6 +3,7 @@ import { config as dotenvConfig } from 'dotenv'
 import { dirname, resolve } from 'path'
 import { fileURLToPath } from 'url'
 import { defineConfig } from 'vite'
+// import { analyzer } from 'vite-bundle-analyzer'
 import cssInjectedByJsPlugin from 'vite-plugin-css-injected-by-js'
 
 const __dirname = dirname(fileURLToPath(import.meta.url))
@@ -14,8 +15,11 @@ dotenvConfig({ path: resolve(__dirname, '../../.env') })
 // - alias all local packages so that they can be build in
 // - no external
 // - no d.ts. dts does not work with monorepo aliasing
-export default defineConfig(({ mode }) => ({
+export default defineConfig(() => ({
-	plugins: [cssInjectedByJsPlugin({ relativeCSSInjection: true })],
+	plugins: [
+		cssInjectedByJsPlugin({ relativeCSSInjection: true }),
+		// analyzer()
+	],
 	publicDir: false,
 	esbuild: {
 		keepNames: true,

packages/page-controller/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@page-agent/page-controller",
-    "version": "1.1.0",
+    "version": "1.2.0",
     "type": "module",
     "main": "./dist/lib/page-controller.js",
     "module": "./dist/lib/page-controller.js",

packages/page-controller/src/PageController.ts

@@ -248,7 +248,7 @@ export class PageController extends EventTarget {
 			if (element instanceof HTMLAnchorElement && element.target === '_blank') {
 				return {
 					success: true,
-					message: `✅ Clicked element (${elemText ?? index}). ⚠️ Link opens in a new tab. You are not capable of reading new tabs.`,
+					message: `✅ Clicked element (${elemText ?? index}). ⚠️ Link opened in a new tab.`,
 				}
 			}
 

packages/ui/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@page-agent/ui",
-    "version": "1.1.0",
+    "version": "1.2.0",
     "type": "module",
     "main": "./dist/lib/page-agent-ui.js",
     "module": "./dist/lib/page-agent-ui.js",

packages/ui/src/index.ts

@@ -1,3 +1,2 @@
 export { Panel, type PanelConfig } from './panel/Panel'
-export type { AgentActivity, PanelAgentAdapter } from './panel/types'
 export { I18n, type SupportedLanguage, type TranslationKey } from './i18n'

packages/website/package.json

@@ -1,7 +1,7 @@
 {
     "name": "@page-agent/website",
     "private": true,
-    "version": "1.1.0",
+    "version": "1.2.0",
     "type": "module",
     "scripts": {
         "dev": "vite --host 0.0.0.0",
@@ -10,30 +10,26 @@
         "typecheck": "tsc --noEmit"
     },
     "devDependencies": {
-        "@tailwindcss/vite": "^4.1.18",
-        "@types/react": "^19.2.8",
-        "@types/react-dom": "^19.2.1",
-        "@vitejs/plugin-react-swc": "^4.1.0",
-        "react": "^19.2.4",
-        "react-dom": "^19.2.4",
-        "tailwindcss": "^4.1.14",
-        "tw-animate-css": "^1.4.0",
-        "wouter": "^3.9.0"
-    },
-    "dependencies": {
         "@radix-ui/react-icons": "^1.3.2",
         "@radix-ui/react-separator": "^1.1.8",
         "@radix-ui/react-slot": "^1.2.4",
         "@radix-ui/react-switch": "^1.2.6",
         "@radix-ui/react-tooltip": "^1.2.8",
+        "@types/react": "^19.2.13",
+        "@types/react-dom": "^19.2.1",
         "class-variance-authority": "^0.7.1",
         "clsx": "^2.1.1",
         "lucide-react": "^0.563.0",
-        "motion": "^12.30.0",
+        "motion": "^12.34.0",
         "next-themes": "^0.4.6",
+        "react": "^19.2.4",
+        "react-dom": "^19.2.4",
         "rough-notation": "^0.5.1",
-        "simple-icons": "^16.7.0",
+        "simple-icons": "^16.8.0",
         "sonner": "^2.0.7",
-        "tailwind-merge": "^3.4.0"
+        "tailwind-merge": "^3.4.0",
+        "tailwindcss": "^4.1.14",
+        "tw-animate-css": "^1.4.0",
+        "wouter": "^3.9.0"
     }
 }

packages/website/src/constants.ts

@@ -1,8 +1,8 @@
 // Demo build (auto-init with demo LLM, for quick testing)
 export const CDN_DEMO_URL =
-	'https://cdn.jsdelivr.net/npm/page-agent@1.1.0/dist/iife/page-agent.demo.js'
+	'https://cdn.jsdelivr.net/npm/page-agent@1.2.0/dist/iife/page-agent.demo.js'
 export const CDN_DEMO_CN_URL =
-	'https://registry.npmmirror.com/page-agent/1.1.0/files/dist/iife/page-agent.demo.js'
+	'https://registry.npmmirror.com/page-agent/1.2.0/files/dist/iife/page-agent.demo.js'
 
 // Demo LLM for website testing
 export const DEMO_MODEL = 'PAGE-AGENT-FREE-TESTING-RANDOM'

packages/website/src/pages/docs/Layout.tsx

@@ -30,6 +30,10 @@ export default function DocsLayout({ children }: DocsLayoutProps) {
 				{ title: isZh ? '概览' : 'Overview', path: '/introduction/overview' },
 				{ title: isZh ? '快速开始' : 'Quick Start', path: '/introduction/quick-start' },
 				{ title: isZh ? '使用限制' : 'Limitations', path: '/introduction/limitations' },
+				{
+					title: isZh ? '故障排查' : 'Troubleshooting',
+					path: '/introduction/troubleshooting',
+				},
 				{
 					title: '🚧 ' + (isZh ? '最佳实践' : 'Best Practices'),
 					path: '/integration/best-practices',

packages/website/src/pages/docs/advanced/page-agent-core/page.tsx

@@ -376,13 +376,6 @@ const result = await agent.execute('Fill in the form with test data')`}
 								? '执行任务并返回结果。包含 success、data 和 history 字段。'
 								: 'Execute a task and return result. Contains success, data, and history fields.',
 						},
-						{
-							name: 'pushObservation(content: string)',
-							type: 'void',
-							description: isZh
-								? '向历史流推送一个观察事件，会在下一步时被 LLM 看到'
-								: 'Push an observation to history stream, will be seen by LLM in next step',
-						},
 						{
 							name: 'dispose(reason?: string)',
 							type: 'void',

packages/website/src/pages/docs/features/chrome-extension/page.tsx

@@ -152,27 +152,24 @@ localStorage.setItem('PageAgentExtUserAuthToken', '<your-token-from-extension>')
 						</div>
 					</section>
 
-					<h3 className="text-xl font-semibold my-3">
+					<h3 className="text-xl font-semibold my-3">PAGE_AGENT_EXT.execute(task, config)</h3>
-						PAGE_AGENT_EXT.execute(task, llmConfig, hooks?)
-					</h3>
 					<p className="text-gray-600 dark:text-gray-300 mb-4">
 						{isZh
-							? '使用 LLM 配置执行任务。返回一个 Promise，在任务完成时 resolve。可选的 hooks 参数用于监听任务执行过程中的事件。'
+							? '使用配置执行任务。返回一个 Promise，在任务完成时 resolve。config 参数包含 LLM 设置、选项和事件回调。'
-							: 'Execute a task with LLM configuration. Returns a Promise that resolves when the task completes. Optional hooks parameter for listening to events during task execution.'}
+							: 'Execute a task with configuration. Returns a Promise that resolves when the task completes. Config includes LLM settings, options, and event callbacks.'}
 					</p>
 
 					<CodeEditor
 						code={
 							isZh
-								? `// 使用 LLM 配置和 hooks 执行任务
+								? `// 使用配置执行任务
 const result = await window.PAGE_AGENT_EXT.execute(
 	'在 GitHub 上搜索 "page-agent" 并打开第一个结果',
 	{
 		baseURL: 'https://api.openai.com/v1',
 		apiKey: 'your-api-key',
-		model: 'gpt-5-2'
+		model: 'gpt-5-2',
-	},
+		// includeInitialTab: false, // 设为 false 排除初始标签页
-	{
 		onStatusChange: status => console.log('状态变化:', status),
 		onActivity: activity => console.log('活动:', activity),
 		onHistoryUpdate: history => console.log('历史更新:', history),
@@ -181,15 +178,14 @@ const result = await window.PAGE_AGENT_EXT.execute(
 )
 
 console.log(result) // 任务执行结果`
-								: `// Execute a task with LLM configuration and hooks
+								: `// Execute a task with configuration
 const result = await window.PAGE_AGENT_EXT.execute(
 	'Search for "page-agent" on GitHub and open the first result',
 	{
 		baseURL: 'https://api.openai.com/v1',
 		apiKey: 'your-api-key',
-		model: 'gpt-5-2'
+		model: 'gpt-5-2',
-	},
+		// includeInitialTab: false, // Set to false to exclude initial tab
-	{
 		onStatusChange: status => console.log('Status change:', status),
 		onActivity: activity => console.log('Activity:', activity),
 		onHistoryUpdate: history => console.log('History update:', history),
@@ -221,41 +217,26 @@ window.PAGE_AGENT_EXT.dispose()`
 					/>
 				</section>
 
-				{/* LLM Config */}
+				{/* ExecuteConfig */}
 				<section>
-					<h2 className="text-2xl font-bold mb-4">{isZh ? 'LLM 配置' : 'LLM Configuration'}</h2>
+					<h2 className="text-2xl font-bold mb-4">{isZh ? '执行配置' : 'Execute Configuration'}</h2>
-
-					<CodeEditor
-						code={
-							isZh
-								? `interface LLMConfig {
-	baseURL: string   // LLM API 端点
-	apiKey: string    // API 密钥
-	model: string     // 模型名称
-}`
-								: `interface LLMConfig {
-	baseURL: string   // LLM API endpoint
-	apiKey: string    // API key
-	model: string     // Model name
-}`
-						}
-						language="typescript"
-					/>
-				</section>
-
-				{/* Execute Hooks */}
-				<section>
-					<h2 className="text-2xl font-bold mb-4">{isZh ? 'Execute Hooks' : 'Execute Hooks'}</h2>
 					<p className="text-gray-600 dark:text-gray-300 mb-4">
 						{isZh
-							? '通过 hooks 参数，你可以监听任务执行过程中的各种事件，实现实时更新 UI、日志记录等功能。'
+							? 'config 参数包含 LLM 设置、选项和事件回调，用于控制任务执行行为。'
-							: 'With hooks parameter, you can listen to various events during task execution for real-time UI updates, logging, and more.'}
+							: 'The config parameter includes LLM settings, options, and event callbacks to control task execution behavior.'}
 					</p>
 
 					<CodeEditor
 						code={
 							isZh
-								? `interface ExecuteHooks {
+								? `interface ExecuteConfig {
+	baseURL: string   // LLM API 端点
+	apiKey: string    // API 密钥
+	model: string     // 模型名称
+
+	// 是否将初始标签页包含在任务中，默认 true
+	includeInitialTab?: boolean
+
 	// Agent 状态变化时调用（idle, running, error, completed 等）
 	onStatusChange?: (status: AgentStatus) => void
 
@@ -268,7 +249,14 @@ window.PAGE_AGENT_EXT.dispose()`
 	// Agent 被停止时调用
 	onDispose?: () => void
 }`
-								: `interface ExecuteHooks {
+								: `interface ExecuteConfig {
+	baseURL: string   // LLM API endpoint
+	apiKey: string    // API key
+	model: string     // Model name
+
+	// Whether to include the initial tab in the task, default true
+	includeInitialTab?: boolean
+
 	// Called when agent status changes (idle, running, error, completed, etc.)
 	onStatusChange?: (status: AgentStatus) => void
 

packages/website/src/pages/docs/features/custom-tools/page.tsx

@@ -25,7 +25,7 @@ export default function CustomTools() {
 					</p>
 
 					<CodeEditor
-						code={`import zod from 'zod'
+						code={`import * as zod from 'zod'
 import { PageAgent, tool } from 'page-agent'
 
 // override internal tool

packages/website/src/pages/docs/features/models/page.tsx

@@ -1,34 +1,49 @@
+import { Fragment } from 'react'
+
 import CodeEditor from '@/components/CodeEditor'
 import { useLanguage } from '@/i18n/context'
 
-// Recommended models: lightweight with excellent tool call capabilities
+const BASELINE = new Set([
-const MODELS = {
+	'gpt-5.1',
-	recommended: [
+	'claude-haiku-4.5',
-		'gpt-4.1-mini',
+	'gemini-3-flash',
-		'claude-haiku-4.5',
+	'deepseek-3.2',
-		'gemini-3-flash',
+	'qwen3-coder-next',
-		'deepseek-3.2',
+])
-		'gpt-5.2',
+
-		'qwen-3-max',
+// Models grouped by brand, newest first
-	],
+const MODEL_GROUPS: Record<string, string[]> = {
-	verified: [
+	OpenAI: ['gpt-5.2', 'gpt-5.1', 'gpt-5', 'gpt-5-mini', 'gpt-4.1', 'gpt-4.1-mini'],
-		'qwen-3-plus',
+	Google: ['gemini-3-pro', 'gemini-3-flash', 'gemini-2.5'],
-		'gpt-4.1',
+	Qwen: ['qwen3-coder-next', 'qwen-3-max', 'qwen-3-plus', 'qwen3:14b (ollama)'],
-		'gpt-5',
+	DeepSeek: ['deepseek-3.2'],
-		'gpt-5-mini',
+	Anthropic: [
-		'grok-4',
+		'claude-opus-4.6',
-		'grok-code-fast',
-		'claude-sonnet-3.5',
-		'claude-sonnet-4.5',
 		'claude-opus-4.5',
-		'gemini-2.5',
+		'claude-sonnet-4.5',
-		'gemini-3-pro',
+		'claude-haiku-4.5',
+		'claude-sonnet-3.5',
 	],
+	xAI: ['grok-4.1-fast', 'grok-4', 'grok-code-fast'],
+	MoonshotAI: ['kimi-k2.5'],
+	'Z.AI': ['glm-4.7'],
 }
 
+const ModelBadge = ({ model, baseline }: { model: string; baseline?: boolean }) => (
+	<div
+		className={`px-3 py-1.5 rounded-md text-xs font-medium font-mono transition-colors ${
+			baseline
+				? 'bg-emerald-500 text-white shadow-sm'
+				: 'bg-white/80 dark:bg-gray-800/80 text-gray-800 dark:text-gray-200 border border-gray-300 dark:border-gray-600'
+		}`}
+	>
+		{model}
+		{baseline && <span className="ml-1">⭐</span>}
+	</div>
+)
+
 export default function Models() {
 	const { isZh } = useLanguage()
-	const allModels = [...MODELS.recommended, ...MODELS.verified]
 
 	return (
 		<div className="max-w-4xl">
@@ -45,28 +60,23 @@ export default function Models() {
 				<p className="text-sm text-gray-600 dark:text-gray-400 mb-4">
 					{isZh
 						? '推荐使用 ToolCall 能力强的轻量级模型。'
-						: 'Recommended: Lightweight models with strong ToolCall capabilities.'}
+						: 'Recommended: Fast, lightweight models with strong ToolCall capabilities.'}
 				</p>
 				<div className="bg-linear-to-br from-emerald-50 to-cyan-50 dark:from-emerald-950/30 dark:to-cyan-950/30 rounded-xl p-6 border border-emerald-200/50 dark:border-emerald-800/50">
-					<div className="flex flex-wrap gap-2">
+					<div className="grid grid-cols-[5rem_1fr] gap-x-3 gap-y-3 items-start">
-						{allModels.map((model) => {
+						{Object.entries(MODEL_GROUPS).map(([brand, models]) => (
-							const isRecommended = MODELS.recommended.includes(model)
+							<Fragment key={brand}>
-							return (
+								<span className="text-xs font-semibold text-gray-500 dark:text-gray-400 pt-2">
-								<div
+									{brand}
-									key={model}
+								</span>
-									className={`px-3 py-1.5 rounded-md text-sm font-medium font-mono transition-colors ${
+								<div className="flex flex-wrap gap-2">
-										isRecommended
+									{models.map((model) => (
-											? 'bg-emerald-500 text-white shadow-sm'
+										<ModelBadge key={model} model={model} baseline={BASELINE.has(model)} />
-											: 'bg-white/80 dark:bg-gray-800/80 text-gray-800 dark:text-gray-200 border border-gray-300 dark:border-gray-600'
+									))}
-									}`}
-								>
-									{model}
-									{isRecommended && <span className="ml-1">⭐</span>}
 								</div>
-							)
+							</Fragment>
-						})}
+						))}
 					</div>
-					<p className="text-xs text-gray-600 dark:text-gray-400 mt-5">⭐ baseline models</p>
 				</div>
 			</section>
 

packages/website/src/pages/docs/index.tsx

@@ -20,6 +20,7 @@ import Limitations from './introduction/limitations/page'
 // Introduction
 import Overview from './introduction/overview/page'
 import QuickStart from './introduction/quick-start/page'
+import Troubleshooting from './introduction/troubleshooting/page'
 
 function DocsPage({ children }: { children: React.ReactNode }) {
 	return (
@@ -51,6 +52,11 @@ export default function DocsRouter() {
 					<Limitations />
 				</DocsPage>
 			</Route>
+			<Route path="/introduction/troubleshooting">
+				<DocsPage>
+					<Troubleshooting />
+				</DocsPage>
+			</Route>
 
 			{/* Features */}
 			<Route path="/features/custom-tools">

packages/website/src/pages/docs/introduction/troubleshooting/page.tsx

@@ -0,0 +1,424 @@
+import { useEffect, useRef, useState } from 'react'
+import { Link } from 'wouter'
+
+import CodeEditor from '@/components/CodeEditor'
+import { useLanguage } from '@/i18n/context'
+
+// ---------------------------------------------------------------------------
+// Data: each section is a typed object for easy extension
+// ---------------------------------------------------------------------------
+
+interface TroubleshootingSection {
+	id: string
+	title: { en: string; zh: string }
+	symptom: { en: string; zh: string }
+	color: 'red' | 'amber' | 'orange' | 'violet'
+	content: (isZh: boolean) => React.ReactNode
+}
+
+const SECTIONS: TroubleshootingSection[] = [
+	{
+		id: 'format-errors',
+		title: { en: 'Model Response Format Errors', zh: '模型返回格式错误' },
+		symptom: {
+			en: 'The model returns malformed tool calls, plain text, or unexpected JSON instead of structured actions.',
+			zh: '模型返回了格式错误的 tool call、纯文本或非预期的 JSON，而非结构化的操作指令。',
+		},
+		color: 'amber',
+		content: FormatErrorsContent,
+	},
+	{
+		id: 'low-success-rate',
+		title: { en: 'Low Task Success Rate', zh: '任务成功率低' },
+		symptom: {
+			en: 'The agent appears to understand the task but frequently fails to complete it, or produces incorrect results.',
+			zh: 'Agent 似乎理解了任务，但频繁执行失败或产生不正确的结果。',
+		},
+		color: 'amber',
+		content: LowSuccessRateContent,
+	},
+	{
+		id: 'wrong-element',
+		title: { en: "Can't Hit Target Elements", zh: '无法点击目标元素' },
+		symptom: {
+			en: 'The agent repeatedly retries but keeps interacting with the wrong element, or fails to locate the correct one.',
+			zh: 'Agent 反复重试，但始终点击在错误的元素上，或无法定位到正确的目标元素。',
+		},
+		color: 'amber',
+		content: WrongElementContent,
+	},
+	{
+		id: 'api-errors',
+		title: { en: 'API Request Errors', zh: 'API 请求错误' },
+		symptom: {
+			en: 'HTTP 400 Bad Request or similar errors when calling the LLM API.',
+			zh: '调用 LLM API 时出现 HTTP 400 Bad Request 或类似的参数错误。',
+		},
+		color: 'amber',
+		content: ApiErrorsContent,
+	},
+]
+
+// ---------------------------------------------------------------------------
+// Section content components
+// ---------------------------------------------------------------------------
+
+function FormatErrorsContent(isZh: boolean) {
+	return (
+		<ol className="list-decimal pl-5 space-y-4 text-gray-700 dark:text-gray-300">
+			<li>
+				<strong>{isZh ? '确认模型是否支持' : 'Verify model compatibility'}</strong>
+				<p className="mt-1">
+					{isZh
+						? '并非所有模型都能正确处理 page-agent 的 tool 定义。请查看'
+						: 'Not all models can handle page-agent tool definitions correctly. Check the '}
+					<Link
+						href="/features/models"
+						className="text-blue-600 dark:text-blue-400 underline underline-offset-2"
+					>
+						{isZh ? '已测试模型列表' : 'tested models list'}
+					</Link>
+					{isZh ? '。' : '.'}
+				</p>
+			</li>
+			<li>
+				<strong>
+					{isZh ? '检查代理/网关的参数转发' : 'Check proxy/gateway parameter forwarding'}
+				</strong>
+				<p className="mt-1">
+					{isZh
+						? '如果使用了 API 代理或网关，请确保请求中的 '
+						: 'If using an API proxy or gateway, make sure the '}
+					<code>tools</code>
+					{isZh
+						? ' 字段被完整、无修改地转发给模型供应商。部分代理可能会剥离或修改此字段。'
+						: ' parameter is forwarded to the model provider intact. Some proxies may strip or alter this field.'}
+				</p>
+			</li>
+			<li>
+				<strong>{isZh ? '寻求社区帮助' : 'Get community help'}</strong>
+				<p className="mt-1">
+					{isZh ? (
+						<>
+							如果以上步骤无法解决问题，欢迎在{' '}
+							<a
+								href="https://github.com/alibaba/page-agent/discussions"
+								target="_blank"
+								rel="noopener noreferrer"
+								className="text-blue-600 dark:text-blue-400 underline underline-offset-2"
+							>
+								GitHub Discussions
+							</a>{' '}
+							中反馈，附上模型名称和错误信息。
+						</>
+					) : (
+						<>
+							If the above steps don't help, join the{' '}
+							<a
+								href="https://github.com/alibaba/page-agent/discussions"
+								target="_blank"
+								rel="noopener noreferrer"
+								className="text-blue-600 dark:text-blue-400 underline underline-offset-2"
+							>
+								GitHub Discussions
+							</a>{' '}
+							with your model name and error details.
+						</>
+					)}
+				</p>
+			</li>
+		</ol>
+	)
+}
+
+function LowSuccessRateContent(isZh: boolean) {
+	return (
+		<>
+			<p className="text-sm text-gray-500 dark:text-gray-400 mb-4 italic">
+				{isZh
+					? '按以下顺序逐步排查，从最简单的情况开始：'
+					: 'Follow this diagnostic funnel from simplest to most advanced:'}
+			</p>
+			<ol className="list-decimal pl-5 space-y-4 text-gray-700 dark:text-gray-300">
+				<li>
+					<strong>{isZh ? '先从简单指令开始' : 'Start with a simple instruction'}</strong>
+					<p className="mt-1">
+						{isZh
+							? '给一个具体的、单步的简单指令（如"点击登录按钮"），看 Agent 能否完成。如果连简单操作都失败了，问题可能不在模型能力上。'
+							: 'Give a concrete, single-step instruction (e.g. "click the login button"). If even simple actions fail, the issue is likely not model capability.'}
+					</p>
+				</li>
+				<li>
+					<strong>{isZh ? '尝试最强模型' : 'Try the strongest model available'}</strong>
+					<p className="mt-1">
+						{isZh
+							? '切换到你能获取到的最先进、最大的模型，以排除是否是模型智能水平不足导致的问题。'
+							: "Switch to the most capable model you have access to, to isolate whether it's a model intelligence issue."}
+					</p>
+				</li>
+				<li>
+					<strong>{isZh ? '优化指令质量' : 'Improve instruction quality'}</strong>
+					<p className="mt-1">
+						{isZh
+							? '给出尽可能具体的指令。对于复杂任务，建议使用另一个 LLM 来预先拆分和细化用户的需求，然后逐步执行。'
+							: "Be as specific as possible. For complex tasks, consider using another LLM to decompose and refine the user's request before execution."}
+					</p>
+				</li>
+				<li>
+					<strong>{isZh ? '提供充足的上下文' : 'Provide sufficient context'}</strong>
+					<p className="mt-1">
+						{isZh
+							? '通过 instructions 配置注入网站背景描述、关键术语解释等上下文信息，帮助 Agent 更好地理解页面。'
+							: 'Use the instructions config to inject website descriptions, key terminology, and background context to help the agent understand the page.'}
+					</p>
+				</li>
+				<li>
+					<strong>{isZh ? '检查 HTML 清洗结果' : 'Check HTML sanitization output'}</strong>
+					<p className="mt-1">
+						{isZh
+							? '使用开发者工具检查清洗后的 HTML，确认关键信息、文本和可操作元素是否被正确保留。'
+							: 'Inspect the sanitized HTML in dev tools to confirm that key information, text, and interactive elements are preserved correctly.'}
+					</p>
+				</li>
+			</ol>
+		</>
+	)
+}
+
+function WrongElementContent(isZh: boolean) {
+	return (
+		<ol className="list-decimal pl-5 space-y-4 text-gray-700 dark:text-gray-300">
+			<li>
+				<strong>{isZh ? '了解现实局限' : 'Understand the reality'}</strong>
+				<p className="mt-1">
+					{isZh
+						? '并非所有网站都提供了完善的语义化 HTML 和 accessibility 标签。对于此类网站，DOM 清洗可能无法产出足够好的结果。'
+						: 'Not all websites provide proper semantic HTML and accessibility labels. For such sites, DOM sanitization may not produce good enough results.'}
+				</p>
+			</li>
+			<li>
+				<strong>{isZh ? '检查目标元素类型' : 'Check target element type'}</strong>
+				<p className="mt-1">
+					{isZh
+						? '确认目标元素是否为图片、Canvas、或需要复杂交互（如拖拽、基于坐标的点击）的元素。这些本身就超出了当前的能力范围。'
+						: 'Verify if the target is an image, Canvas, or requires complex interactions (drag-and-drop, coordinate-based clicking). These are beyond current capabilities.'}
+				</p>
+			</li>
+			<li>
+				<strong>{isZh ? '检查清洗后的 HTML' : 'Inspect sanitized HTML'}</strong>
+				<p className="mt-1">
+					{isZh
+						? '检查清洗结果中是否存在关键信息丢失、可操作元素未被编号等问题。'
+						: 'Look for missing key information or unnumbered interactive elements in the sanitized output.'}
+				</p>
+			</li>
+			<li>
+				<strong>{isZh ? '注入 accessibility 增强' : 'Inject accessibility improvements'}</strong>
+				<p className="mt-1">
+					{isZh
+						? '通过注入脚本为网站添加 aria-label、语义化标签等 accessibility 属性，改善 DOM 清洗质量。'
+						: 'Inject scripts to add aria-labels, semantic attributes, and other a11y improvements to enhance DOM sanitization quality.'}
+				</p>
+			</li>
+			<li>
+				<strong>{isZh ? '开发专用 Tool' : 'Build a custom Tool'}</strong>
+				<p className="mt-1">
+					{isZh ? (
+						<>
+							对于特定的、持续难以操作的元素，考虑开发{' '}
+							<Link
+								href="/features/custom-tools"
+								className="text-blue-600 dark:text-blue-400 underline underline-offset-2"
+							>
+								自定义 Tool
+							</Link>{' '}
+							来直接操作这些元素。
+						</>
+					) : (
+						<>
+							For consistently difficult elements, consider building a{' '}
+							<Link
+								href="/features/custom-tools"
+								className="text-blue-600 dark:text-blue-400 underline underline-offset-2"
+							>
+								custom Tool
+							</Link>{' '}
+							to interact with them directly.
+						</>
+					)}
+				</p>
+			</li>
+		</ol>
+	)
+}
+
+function ApiErrorsContent(isZh: boolean) {
+	return (
+		<div className="space-y-4 text-gray-700 dark:text-gray-300">
+			<p>
+				{isZh
+					? '一些 LLM 供应商使用了与 OpenAI 不完全兼容的参数格式，导致请求参数校验失败。'
+					: 'Some LLM providers use parameter formats that are not fully compatible with the OpenAI spec, causing request validation failures.'}
+			</p>
+			<div className="bg-gray-50 dark:bg-gray-800/50 rounded-lg p-4">
+				<p className="font-medium mb-2">
+					{isZh ? '解决方案：使用 customFetch' : 'Solution: use customFetch'}
+				</p>
+				<p className="text-sm mb-3">
+					{isZh
+						? '通过 customFetch 配置拦截请求，在发送前调整参数格式以适配目标供应商的要求。'
+						: 'Use the customFetch config to intercept requests and adapt parameters before sending them to the target provider.'}
+				</p>
+				<CodeEditor
+					code={`const agent = new PageAgent({
+  // ...
+  customFetch: async (url, init) => {
+    const body = JSON.parse(init.body)
+    // Adapt parameters for your provider
+    delete body.stream_options
+    return fetch(url, { ...init, body: JSON.stringify(body) })
+  },
+})`}
+				/>
+			</div>
+			<p className="text-sm">
+				{isZh ? '参见 ' : 'See '}
+				<Link
+					href="/advanced/page-agent-core"
+					className="text-blue-600 dark:text-blue-400 underline underline-offset-2"
+				>
+					PageAgentCore API
+				</Link>
+				{isZh ? ' 了解 customFetch 的完整用法。' : ' for full customFetch documentation.'}
+			</p>
+		</div>
+	)
+}
+
+// ---------------------------------------------------------------------------
+// Color mapping for symptom callouts
+// ---------------------------------------------------------------------------
+
+const SYMPTOM_COLORS = {
+	red: 'border-red-400 bg-red-50 dark:bg-red-900/15 text-red-800 dark:text-red-200',
+	amber: 'border-amber-400 bg-amber-50 dark:bg-amber-900/15 text-amber-800 dark:text-amber-200',
+	orange:
+		'border-orange-400 bg-orange-50 dark:bg-orange-900/15 text-orange-800 dark:text-orange-200',
+	violet:
+		'border-violet-400 bg-violet-50 dark:bg-violet-900/15 text-violet-800 dark:text-violet-200',
+} as const
+
+// ---------------------------------------------------------------------------
+// Right-side TOC with IntersectionObserver
+// ---------------------------------------------------------------------------
+
+function useActiveSection(ids: string[]) {
+	const [activeId, setActiveId] = useState(ids[0])
+	const observerRef = useRef<IntersectionObserver | null>(null)
+
+	useEffect(() => {
+		observerRef.current?.disconnect()
+
+		const visibleEntries = new Map<string, number>()
+
+		observerRef.current = new IntersectionObserver(
+			(entries) => {
+				for (const entry of entries) {
+					if (entry.isIntersecting) {
+						visibleEntries.set(entry.target.id, entry.intersectionRatio)
+					} else {
+						visibleEntries.delete(entry.target.id)
+					}
+				}
+				// Pick the first visible section in document order
+				const firstVisible = ids.find((id) => visibleEntries.has(id))
+				if (firstVisible) setActiveId(firstVisible)
+			},
+			{ rootMargin: '-80px 0px -60% 0px', threshold: [0, 0.25] }
+		)
+
+		for (const id of ids) {
+			const el = document.getElementById(id)
+			if (el) observerRef.current.observe(el)
+		}
+
+		return () => observerRef.current?.disconnect()
+	}, [ids])
+
+	return activeId
+}
+
+// ---------------------------------------------------------------------------
+// Page component
+// ---------------------------------------------------------------------------
+
+export default function TroubleshootingPage() {
+	const { isZh } = useLanguage()
+	const sectionIds = SECTIONS.map((s) => s.id)
+	const activeId = useActiveSection(sectionIds)
+
+	return (
+		<div className="max-w-5xl mx-auto">
+			{/* Header */}
+			<div className="mb-10">
+				<h1 className="text-4xl font-bold mb-4 text-gray-900 dark:text-white">Troubleshooting</h1>
+			</div>
+
+			{/* Two-column: content + TOC */}
+			<div className="flex gap-8">
+				{/* Main content */}
+				<div className="flex-1 min-w-0 space-y-12">
+					{SECTIONS.map((section) => (
+						<section key={section.id} id={section.id} className="scroll-mt-24">
+							<h2 className="text-2xl font-bold mb-4 text-gray-900 dark:text-white">
+								{isZh ? section.title.zh : section.title.en}
+							</h2>
+
+							{/* Symptom callout */}
+							<div
+								className={`border-l-4 px-4 py-3 rounded-r-lg mb-6 ${SYMPTOM_COLORS[section.color]}`}
+							>
+								<span className="text-xs font-semibold uppercase tracking-wider opacity-70">
+									{isZh ? '症状' : 'Symptom'}
+								</span>
+								<p className="mt-1 text-sm">{isZh ? section.symptom.zh : section.symptom.en}</p>
+							</div>
+
+							{/* Diagnostic steps */}
+							<div className="prose-sm">{section.content(isZh)}</div>
+						</section>
+					))}
+				</div>
+
+				{/* Right TOC */}
+				<aside className="hidden lg:block w-48 shrink-0">
+					<div className="sticky top-24">
+						<h4 className="text-xs font-semibold text-gray-400 dark:text-gray-500 uppercase tracking-wider mb-3">
+							{isZh ? '目录' : 'On this page'}
+						</h4>
+						<nav className="space-y-1">
+							{SECTIONS.map((section) => (
+								<button
+									key={section.id}
+									type="button"
+									onClick={() =>
+										document
+											.getElementById(section.id)
+											?.scrollIntoView({ behavior: 'smooth', block: 'start' })
+									}
+									className={`block text-sm py-1 text-left transition-colors ${
+										activeId === section.id
+											? 'text-blue-600 dark:text-blue-400 font-medium'
+											: 'text-gray-500 dark:text-gray-400 hover:text-gray-900 dark:hover:text-gray-200'
+									}`}
+								>
+									{isZh ? section.title.zh : section.title.en}
+								</button>
+							))}
+						</nav>
+					</div>
+				</aside>
+			</div>
+		</div>
+	)
+}