Compare commits

..

25 Commits

Author SHA1 Message Date
Simon 11d66f42c4 chore(ext): version 0.1.5 2026-02-11 20:12:57 +08:00
Simon 20e947a1c1 chore(version): bump version to 1.2.0 2026-02-11 20:09:23 +08:00
Simon 7c87c90258 fix(ext): fix multi-thread logic; extensive logging and error handling 2026-02-11 19:51:19 +08:00
Simon fcb9ec4e57 chore: clean up 2026-02-11 19:27:14 +08:00
Simon 1fcf931685 feat(core): add a observe phase in a step; improve abortSignal 2026-02-10 22:20:42 +08:00
Simon c71ffcec1c refactor(ext): tidy up background code 2026-02-10 21:59:58 +08:00
Simon 430be466fc feat(ext): add chat history 2026-02-10 18:43:54 +08:00
Simon 3f284cf2e6 chore(controller): remove link hint 2026-02-10 17:19:02 +08:00
Simon 6f0655478a fix(core): remove trimLines to fix indentation lost 2026-02-10 16:30:03 +08:00
Simon 7d5449967b feat(core): improve observation logic; mark states as private 2026-02-10 16:28:25 +08:00
Simon fd47888d1b chore(core): tidy up PageAgentCore members 2026-02-10 15:45:04 +08:00
Simon 10f470ce7d chore: improve package.json 2026-02-10 15:28:50 +08:00
Simon cb29db1cc4 Merge pull request #138 from alibaba/dependabot/npm_and_yarn/production-dependencies-58bd30f460
chore(deps): bump the production-dependencies group with 2 updates
2026-02-10 14:21:11 +08:00
Simon 8c08f639dd Merge pull request #139 from alibaba/dependabot/npm_and_yarn/development-dependencies-dee6027aab
chore(deps-dev): bump the development-dependencies group with 8 updates
2026-02-10 14:18:18 +08:00
dependabot[bot] 488f080e68 chore(deps-dev): bump the development-dependencies group with 8 updates
Bumps the development-dependencies group with 8 updates:

| Package | From | To |
| --- | --- | --- |
| [@microsoft/api-extractor](https://github.com/microsoft/rushstack/tree/HEAD/apps/api-extractor) | `7.56.0` | `7.56.3` |
| [@types/node](https://github.com/DefinitelyTyped/DefinitelyTyped/tree/HEAD/types/node) | `25.2.0` | `25.2.2` |
| [dotenv](https://github.com/motdotla/dotenv) | `17.2.3` | `17.2.4` |
| [eslint-plugin-react-dom](https://github.com/Rel1cx/eslint-react/tree/HEAD/packages/plugins/eslint-plugin-react-dom) | `2.9.3` | `2.12.2` |
| [eslint-plugin-react-x](https://github.com/Rel1cx/eslint-react/tree/HEAD/packages/plugins/eslint-plugin-react-x) | `2.9.3` | `2.12.2` |
| [typescript-eslint](https://github.com/typescript-eslint/typescript-eslint/tree/HEAD/packages/typescript-eslint) | `8.54.0` | `8.55.0` |
| [@types/react](https://github.com/DefinitelyTyped/DefinitelyTyped/tree/HEAD/types/react) | `19.2.10` | `19.2.13` |
| [wxt](https://github.com/wxt-dev/wxt) | `0.20.13` | `0.20.14` |


Updates `@microsoft/api-extractor` from 7.56.0 to 7.56.3
- [Changelog](https://github.com/microsoft/rushstack/blob/main/apps/api-extractor/CHANGELOG.md)
- [Commits](https://github.com/microsoft/rushstack/commits/@microsoft/api-extractor_v7.56.3/apps/api-extractor)

Updates `@types/node` from 25.2.0 to 25.2.2
- [Release notes](https://github.com/DefinitelyTyped/DefinitelyTyped/releases)
- [Commits](https://github.com/DefinitelyTyped/DefinitelyTyped/commits/HEAD/types/node)

Updates `dotenv` from 17.2.3 to 17.2.4
- [Changelog](https://github.com/motdotla/dotenv/blob/master/CHANGELOG.md)
- [Commits](https://github.com/motdotla/dotenv/commits)

Updates `eslint-plugin-react-dom` from 2.9.3 to 2.12.2
- [Release notes](https://github.com/Rel1cx/eslint-react/releases)
- [Changelog](https://github.com/Rel1cx/eslint-react/blob/main/CHANGELOG.md)
- [Commits](https://github.com/Rel1cx/eslint-react/commits/v2.12.2/packages/plugins/eslint-plugin-react-dom)

Updates `eslint-plugin-react-x` from 2.9.3 to 2.12.2
- [Release notes](https://github.com/Rel1cx/eslint-react/releases)
- [Changelog](https://github.com/Rel1cx/eslint-react/blob/main/CHANGELOG.md)
- [Commits](https://github.com/Rel1cx/eslint-react/commits/v2.12.2/packages/plugins/eslint-plugin-react-x)

Updates `typescript-eslint` from 8.54.0 to 8.55.0
- [Release notes](https://github.com/typescript-eslint/typescript-eslint/releases)
- [Changelog](https://github.com/typescript-eslint/typescript-eslint/blob/main/packages/typescript-eslint/CHANGELOG.md)
- [Commits](https://github.com/typescript-eslint/typescript-eslint/commits/v8.55.0/packages/typescript-eslint)

Updates `@types/react` from 19.2.10 to 19.2.13
- [Release notes](https://github.com/DefinitelyTyped/DefinitelyTyped/releases)
- [Commits](https://github.com/DefinitelyTyped/DefinitelyTyped/commits/HEAD/types/react)

Updates `wxt` from 0.20.13 to 0.20.14
- [Release notes](https://github.com/wxt-dev/wxt/releases)
- [Commits](https://github.com/wxt-dev/wxt/compare/wxt-v0.20.13...wxt-v0.20.14)

---
updated-dependencies:
- dependency-name: "@microsoft/api-extractor"
  dependency-version: 7.56.3
  dependency-type: direct:development
  update-type: version-update:semver-patch
  dependency-group: development-dependencies
- dependency-name: "@types/node"
  dependency-version: 25.2.2
  dependency-type: direct:development
  update-type: version-update:semver-patch
  dependency-group: development-dependencies
- dependency-name: dotenv
  dependency-version: 17.2.4
  dependency-type: direct:development
  update-type: version-update:semver-patch
  dependency-group: development-dependencies
- dependency-name: eslint-plugin-react-dom
  dependency-version: 2.12.2
  dependency-type: direct:development
  update-type: version-update:semver-minor
  dependency-group: development-dependencies
- dependency-name: eslint-plugin-react-x
  dependency-version: 2.12.2
  dependency-type: direct:development
  update-type: version-update:semver-minor
  dependency-group: development-dependencies
- dependency-name: typescript-eslint
  dependency-version: 8.55.0
  dependency-type: direct:development
  update-type: version-update:semver-minor
  dependency-group: development-dependencies
- dependency-name: "@types/react"
  dependency-version: 19.2.13
  dependency-type: direct:development
  update-type: version-update:semver-patch
  dependency-group: development-dependencies
- dependency-name: wxt
  dependency-version: 0.20.14
  dependency-type: direct:development
  update-type: version-update:semver-patch
  dependency-group: development-dependencies
...

Signed-off-by: dependabot[bot] <support@github.com>
2026-02-09 22:25:35 +00:00
dependabot[bot] 6e07af422b chore(deps): bump the production-dependencies group with 2 updates
Bumps the production-dependencies group with 2 updates: [motion](https://github.com/motiondivision/motion) and [simple-icons](https://github.com/simple-icons/simple-icons).


Updates `motion` from 12.30.0 to 12.34.0
- [Changelog](https://github.com/motiondivision/motion/blob/main/CHANGELOG.md)
- [Commits](https://github.com/motiondivision/motion/compare/v12.30.0...v12.34.0)

Updates `simple-icons` from 16.7.0 to 16.8.0
- [Release notes](https://github.com/simple-icons/simple-icons/releases)
- [Commits](https://github.com/simple-icons/simple-icons/compare/16.7.0...16.8.0)

---
updated-dependencies:
- dependency-name: motion
  dependency-version: 12.34.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: production-dependencies
- dependency-name: simple-icons
  dependency-version: 16.8.0
  dependency-type: direct:production
  update-type: version-update:semver-minor
  dependency-group: production-dependencies
...

Signed-off-by: dependabot[bot] <support@github.com>
2026-02-09 22:23:28 +00:00
Simon e9d721f6b9 chore(ext): bump extension version to 0.1.4 2026-02-09 21:12:11 +08:00
Simon 840162df8c feat: change observation timing 2026-02-09 21:05:29 +08:00
Simon 1517bc4868 feat: prune prompts 2026-02-09 21:04:39 +08:00
Simon 67f6a56a87 chore(version): bump version to 1.1.2 2026-02-09 17:50:34 +08:00
Simon 6f0ff1fd33 refactor: zod tree-shaking; better error handling in agent steps 2026-02-09 17:49:10 +08:00
Simon 6fdb8d03f2 docs: trouble shooting 2026-02-09 16:32:58 +08:00
Simon d00a8dcc21 feat: improve prompt; update model list 2026-02-09 15:50:20 +08:00
Simon 851e9ef158 docs: model list 2026-02-06 18:49:06 +08:00
Simon 8ea8ce9acd fix(llms): gpt-5-mini temperature 2026-02-06 17:19:06 +08:00
43 changed files with 1726 additions and 842 deletions
+3 -3
View File
@@ -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 |
| China | https://registry.npmmirror.com/page-agent/1.1.0/files/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/1.2.0/files/dist/iife/page-agent.demo.js |
### NPM 安装
+3 -3
View File
@@ -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 |
| China | https://registry.npmmirror.com/page-agent/1.1.0/files/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/1.2.0/files/dist/iife/page-agent.demo.js |
### NPM Installation
+390 -379
View File
File diff suppressed because it is too large Load Diff
+12 -10
View File
@@ -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",
"npm": ">=10.0.0"
"node": "^20.19.0 || ^22.13.0 || >=24"
},
"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"
+3 -3
View File
@@ -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/page-controller": "1.1.0"
"@page-agent/llms": "1.2.0",
"@page-agent/page-controller": "1.2.0"
}
}
+146 -124
View File
@@ -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 } from './utils/assert'
import { assert, normalizeResponse, uid, waitFor } from './utils'
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 }
id = uid()
tools: typeof tools
readonly id = uid()
readonly config: PageAgentConfig & { maxSteps: number }
readonly tools: typeof tools
/** PageController for DOM operations */
readonly pageController: PageController
task = ''
taskId = ''
/** Agent execution status */
#status: AgentStatus = 'idle'
/** History events */
history: HistoricalEvent[] = []
/**
* 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 */
pageController: PageController
/** Runtime states for tracking across steps */
states = {
/** Accumulated wait time in seconds, used by wait tool */
/** internal states during a single task execution */
#states = {
/** Accumulated wait time in seconds */
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.
* This will be visible in <agent_history> and remain in memory across steps.
* Push a observation message to the history event stream.
* 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.#emitHistoryChange()
this.#observations.push(content)
}
async execute(task: string): Promise<ExecutionResult> {
@@ -203,30 +212,25 @@ export class PageAgentCore extends EventTarget {
this.#setStatus('running')
this.#emitHistoryChange()
// Reset states
this.states = {
totalWaitTime: 0,
lastURL: '',
}
// Reset internal states
this.#states = { totalWaitTime: 0, lastURL: '', browserState: null }
try {
let step = 0
let step = 0
while (true) {
await this.#generateObservations(step)
while (true) {
try {
console.group(`step: ${step}`)
await onBeforeStep?.(this, step)
console.group(`step: ${step}`)
// observe
// abort
if (this.#abortController.signal.aborted) throw new Error('AbortError')
console.log(chalk.blue.bold('👀 Observing...'))
// Thinking
console.log(chalk.blue('Thinking...'))
this.#emitActivity({ type: 'thinking' })
this.#states.browserState = await this.pageController.getBrowserState()
await this.#handleObservations(step)
// 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++
if (step > this.config.maxSteps) {
this.#onDone(false)
const result: ExecutionResult = {
success: false,
data: 'Step count exceeded maximum limit',
history: this.history,
}
await onAfterTask?.(this, result)
return result
}
console.groupEnd()
// finish task if done
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)
const errorMessage = String(error)
this.#emitActivity({ type: 'error', message: errorMessage })
this.#onDone(false)
const result: ExecutionResult = {
success: false,
data: errorMessage,
history: this.history,
step++
if (step > this.config.maxSteps) {
this.#onDone(false)
const result: ExecutionResult = {
success: false,
data: 'Step count exceeded maximum limit',
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
if (toolName !== 'wait') {
this.states.totalWaitTime = 0
// counting wait time
if (toolName === 'wait') {
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
* - URL change detection
* - Too many steps warning
* Generate system observations before each step
* @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()
if (currentURL !== this.states.lastURL) {
const currentURL = this.#states.browserState?.url || ''
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
const remaining = this.config.maxSteps - stepCount
// Remaining steps warning
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>
<user_request>
${this.task}
</user_request>
<step_info>
Step ${stepCount + 1} of ${this.config.maxSteps} max possible steps
Current date and time: ${new Date().toLocaleString()}
</step_info>
</agent_state>
`
prompt += '<agent_state>\n'
prompt += '<user_request>\n'
prompt += `${this.task}\n`
prompt += '</user_request>\n'
prompt += '<step_info>\n'
prompt += `Step ${stepCount + 1} of ${this.config.maxSteps} max possible steps\n`
prompt += `Current time: ${new Date().toLocaleString()}\n`
prompt += '</step_info>\n'
prompt += '</agent_state>\n\n'
// <agent_history>
// - <step_N> for steps
// - <sys> for observations and system messages
prompt += '\n<agent_history>\n'
prompt += '<agent_history>\n'
let stepIndex = 0
for (const event of this.history) {
if (event.type === 'step') {
stepIndex++
prompt += `<step_${stepIndex}>
Evaluation of Previous Step: ${event.reflection.evaluation_previous_goal}
Memory: ${event.reflection.memory}
Next Goal: ${event.reflection.next_goal}
Action Results: ${event.action.output}
</step_${stepIndex}>
`
prompt += `<step_${stepIndex}>\n`
prompt += `Evaluation of Previous Step: ${event.reflection.evaluation_previous_goal}\n`
prompt += `Memory: ${event.reflection.memory}\n`
prompt += `Next Goal: ${event.reflection.next_goal}\n`
prompt += `Action Results: ${event.action.output}\n`
prompt += `</step_${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()
+4 -6
View File
@@ -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."
"action":{"action": {// action-specific parameter}}
"next_goal": "State the next immediate goal and action to achieve it, in one clear sentence.",
"action":{
"Action name": {// Action parameters}
}
}
</output>
+11 -20
View File
@@ -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:
'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.',
description: 'Wait for x seconds. 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:
'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.',
description: 'Scroll the page vertically. Use index for scroll elements (dropdowns/custom UI).',
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
-17
View File
@@ -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)
}
}
+18 -31
View File
@@ -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)
}
}
+22 -21
View File
@@ -1,7 +1,7 @@
{
"name": "@page-agent/ext",
"private": true,
"version": "0.1.3",
"version": "0.1.5",
"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"
}
}
@@ -74,6 +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)
@@ -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),
@@ -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
}
@@ -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 await chrome.runtime.sendMessage({
return 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,
title: await this.currentTabTitle,
url: currentUrl,
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 (preCheckError) {
return { success: false, message: preCheckError }
if (!this.currentTabId) {
return { success: false, message: 'RemotePageController not initialized.' }
}
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 {
@@ -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 || '' }
sendResponse(result)
debug('get_tab_info: success', tab)
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)
})
})
}
+122 -42
View File
@@ -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)
@@ -13,6 +30,8 @@ export class TabsController extends EventTarget {
private windowId: number | null = null
async init(task: string, includeInitialTab: boolean = true) {
debug('init', task, includeInitialTab)
this.task = task
this.tabs = []
this.currentTabId = null
@@ -20,7 +39,7 @@ 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',
})
@@ -33,9 +52,20 @@ export class TabsController extends EventTarget {
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,
})
}
@@ -70,6 +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
}
}
}
@@ -80,8 +118,10 @@ export class TabsController extends EventTarget {
})
}
async openNewTab(url: string): Promise<{ success: boolean; tabId: number; message: string }> {
const result = await chrome.runtime.sendMessage({
async openNewTab(url: string): Promise<string> {
debug('openNewTab', url)
const result = await sendMessage({
type: 'TAB_CONTROL',
action: 'open_new_tab',
payload: { url },
@@ -104,7 +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 },
@@ -118,7 +158,7 @@ export class TabsController extends EventTarget {
this.tabGroupId = groupId
await chrome.runtime.sendMessage({
await sendMessage({
type: 'TAB_CONTROL',
action: 'update_tab_group',
payload: {
@@ -131,57 +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
// @todo
await new Promise((resolve) => setTimeout(resolve, 500))
await this.waitUntilTabLoaded(tabId)
return {
success: true,
tabId,
message: `Opened new tab ID ${tabId} with URL ${url}`,
}
return `✅ 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 {
success: false,
message: `Tab ID ${tabId} not found in tab list.`,
}
throw new Error(`Tab ID ${tabId} not found in tab list.`)
}
await this.updateCurrentTabId(tabId)
return {
success: true,
message: `Switched to tab ID ${tabId}.`,
}
return `✅ 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 {
success: false,
message: `Tab ID ${tabId} not found in tab list.`,
}
throw new Error(`Tab ID ${tabId} not found in tab list.`)
}
if (targetTab.isInitial) {
return {
success: false,
message: `Cannot close the initial tab ID ${tabId}.`,
}
throw new Error(`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 },
@@ -198,29 +224,39 @@ export class TabsController extends EventTarget {
}
}
return {
success: true,
message: `Closed tab ID ${tabId}.`,
}
return `✅ Closed tab ID ${tabId}.`
} else {
return {
success: false,
message: `Failed to close tab ID ${tabId}: ${result.error}`,
}
throw new Error(`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
}
@@ -239,6 +275,17 @@ export class TabsController extends EventTarget {
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'))
}
@@ -257,6 +304,9 @@ export type TabAction =
interface TabMeta {
id: number
isInitial: boolean
url?: string
title?: string
status?: 'loading' | 'unloaded' | 'complete'
}
const TAB_GROUP_COLORS = [
@@ -275,3 +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)
})
}
@@ -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."
"action":{"action_name": {// action-specific parameter}}
"next_goal": "State the next immediate goal and action to achieve it, in one clear sentence.",
"action":{
"Action name": {// Action parameters}
}
}
</output>
+16 -5
View File
@@ -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)
return result.message
try {
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)}`
}
},
},
}
+1 -1
View File
@@ -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)
@@ -1,24 +1,13 @@
import { handlePageControlMessage } from '@/agent/RemotePageController.background'
import { handleTabControlMessage } 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
// ============================================================================
import { handleTabControlMessage, setupTabChangeEvents } from '@/agent/TabsController.background'
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 },
})
})
})
@@ -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}`)
@@ -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" />
@@ -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>
)
}
@@ -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>
)
}
+70
View File
@@ -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')
}
+1 -1
View File
@@ -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",
+3
View File
@@ -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
+5 -1
View File
@@ -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'
+6 -6
View File
@@ -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",
"@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"
"zod": "^4.3.5"
}
}
+6 -2
View File
@@ -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 }) => ({
plugins: [cssInjectedByJsPlugin({ relativeCSSInjection: true })],
export default defineConfig(() => ({
plugins: [
cssInjectedByJsPlugin({ relativeCSSInjection: true }),
// analyzer()
],
publicDir: false,
esbuild: {
keepNames: true,
+1 -1
View File
@@ -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",
@@ -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.`,
}
}
+1 -1
View File
@@ -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",
-1
View File
@@ -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'
+11 -15
View File
@@ -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"
}
}
+2 -2
View File
@@ -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'
@@ -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',
@@ -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',
@@ -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
@@ -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 MODELS = {
recommended: [
'gpt-4.1-mini',
'claude-haiku-4.5',
'gemini-3-flash',
'deepseek-3.2',
'gpt-5.2',
'qwen-3-max',
],
verified: [
'qwen-3-plus',
'gpt-4.1',
'gpt-5',
'gpt-5-mini',
'grok-4',
'grok-code-fast',
'claude-sonnet-3.5',
'claude-sonnet-4.5',
const BASELINE = new Set([
'gpt-5.1',
'claude-haiku-4.5',
'gemini-3-flash',
'deepseek-3.2',
'qwen3-coder-next',
])
// Models grouped by brand, newest first
const MODEL_GROUPS: Record<string, string[]> = {
OpenAI: ['gpt-5.2', 'gpt-5.1', 'gpt-5', 'gpt-5-mini', 'gpt-4.1', 'gpt-4.1-mini'],
Google: ['gemini-3-pro', 'gemini-3-flash', 'gemini-2.5'],
Qwen: ['qwen3-coder-next', 'qwen-3-max', 'qwen-3-plus', 'qwen3:14b (ollama)'],
DeepSeek: ['deepseek-3.2'],
Anthropic: [
'claude-opus-4.6',
'claude-opus-4.5',
'gemini-2.5',
'gemini-3-pro',
'claude-sonnet-4.5',
'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">
{allModels.map((model) => {
const isRecommended = MODELS.recommended.includes(model)
return (
<div
key={model}
className={`px-3 py-1.5 rounded-md text-sm font-medium font-mono transition-colors ${
isRecommended
? '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}
{isRecommended && <span className="ml-1"></span>}
<div className="grid grid-cols-[5rem_1fr] gap-x-3 gap-y-3 items-start">
{Object.entries(MODEL_GROUPS).map(([brand, models]) => (
<Fragment key={brand}>
<span className="text-xs font-semibold text-gray-500 dark:text-gray-400 pt-2">
{brand}
</span>
<div className="flex flex-wrap gap-2">
{models.map((model) => (
<ModelBadge key={model} model={model} baseline={BASELINE.has(model)} />
))}
</div>
)
})}
</Fragment>
))}
</div>
<p className="text-xs text-gray-600 dark:text-gray-400 mt-5"> baseline models</p>
</div>
</section>
@@ -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">
@@ -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 ? (
<>
{' '}
<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>{' '}
</>
) : (
<>
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
? ''
: '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 ? '' : 'Start with a simple instruction'}</strong>
<p className="mt-1">
{isZh
? '"点击登录按钮" Agent '
: '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 ? '' : 'Try the strongest model available'}</strong>
<p className="mt-1">
{isZh
? ''
: "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 ? '' : 'Provide sufficient context'}</strong>
<p className="mt-1">
{isZh
? ' instructions Agent '
: '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 ? ' HTML ' : 'Check HTML sanitization output'}</strong>
<p className="mt-1">
{isZh
? '使 HTML'
: '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 ? '' : 'Understand the reality'}</strong>
<p className="mt-1">
{isZh
? ' HTML accessibility DOM '
: '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 ? '' : 'Check target element type'}</strong>
<p className="mt-1">
{isZh
? 'Canvas'
: '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 ? ' HTML' : 'Inspect sanitized HTML'}</strong>
<p className="mt-1">
{isZh
? ''
: 'Look for missing key information or unnumbered interactive elements in the sanitized output.'}
</p>
</li>
<li>
<strong>{isZh ? ' accessibility ' : 'Inject accessibility improvements'}</strong>
<p className="mt-1">
{isZh
? ' aria-label accessibility DOM '
: 'Inject scripts to add aria-labels, semantic attributes, and other a11y improvements to enhance DOM sanitization quality.'}
</p>
</li>
<li>
<strong>{isZh ? ' 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
? ' LLM 使 OpenAI '
: '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 ? '使 customFetch' : 'Solution: use customFetch'}
</p>
<p className="text-sm mb-3">
{isZh
? ' customFetch '
: '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 ? ' ' : 'See '}
<Link
href="/advanced/page-agent-core"
className="text-blue-600 dark:text-blue-400 underline underline-offset-2"
>
PageAgentCore API
</Link>
{isZh ? ' customFetch ' : ' 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 ? '' : '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 ? '' : '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>
)
}