feat(ext): use PAGE_AGENT_EXT namespace; add viber instructions

This commit is contained in:
Simon
2026-02-03 19:09:37 +08:00
parent 648a0c1bda
commit 71ca554108
7 changed files with 618 additions and 19 deletions

View File

@@ -0,0 +1,283 @@
# Page Agent Extension API
This document describes how to integrate the Page Agent browser extension into your web application.
## Installation
### 1. Install the browser extension
Install the Page Agent extension from the Chrome Web Store.
### 2. Install type definitions (recommended)
```bash
npm install @page-agent/core --save-dev
```
### 3. Set up authentication
The extension only injects APIs when it detects a valid token in `localStorage`.
1. Open the extension's side panel to get your authorization token
2. Set the token in your page:
```typescript
localStorage.setItem('PageAgentExtUserAuthToken', 'your-token')
```
## Quick Start
```typescript
import type {
AgentActivity,
AgentStatus,
ExecutionResult,
HistoricalEvent,
LLMConfig,
} from '@page-agent/core'
// Wait for extension injection (up to 1 second)
async function waitForExtension(timeout = 1000): Promise<boolean> {
const start = Date.now()
while (Date.now() - start < timeout) {
if (window.PAGE_AGENT_EXT) return true
await new Promise((r) => setTimeout(r, 100))
}
return false
}
// Usage
if (await waitForExtension()) {
const result = await window.PAGE_AGENT_EXT!.execute(
'Click the login button',
{
baseURL: 'https://api.openai.com/v1',
apiKey: 'your-api-key',
model: 'gpt-5.2',
},
{
onStatusChange: (status) => console.log('Status:', status),
onActivity: (activity) => console.log('Activity:', activity),
}
)
console.log('Result:', result)
}
```
## Global API
The extension injects the following APIs into the `window` object:
### `window.PAGE_AGENT_EXT_VERSION`
Extension version string (e.g., `"1.0.0"`). This is exposed separately to allow version checking before accessing the main API object.
### `window.PAGE_AGENT_EXT`
Main API namespace object containing:
#### `PAGE_AGENT_EXT.execute(task, llmConfig, hooks?)`
Execute an agent task.
**Parameters:**
| Name | Type | Required | Description |
|------|------|----------|-------------|
| `task` | `string` | Yes | Task description |
| `llmConfig` | `LLMConfig` | Yes | LLM configuration |
| `hooks` | `ExecuteHooks` | No | Event callbacks |
**Returns:** `Promise<ExecutionResult>`
#### `PAGE_AGENT_EXT.dispose()`
Stop and destroy the current running agent.
## Types
Install `@page-agent/core` for full type definitions:
```typescript
import type {
AgentActivity,
AgentStatus,
ExecutionResult,
HistoricalEvent,
LLMConfig,
} from '@page-agent/core'
export interface ExecuteHooks {
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>
```
### AgentStatus
```typescript
type AgentStatus = 'idle' | 'running' | 'completed' | 'error'
```
| Status | Description |
|--------|-------------|
| `idle` | Agent is idle, ready to execute |
| `running` | Agent is executing a task |
| `completed` | Task completed successfully |
| `error` | Task failed with an error |
### AgentActivity
```typescript
type AgentActivity =
| { type: 'thinking' }
| { type: 'executing'; tool: string; input: unknown }
| { type: 'executed'; tool: string; input: unknown; output: string; duration: number }
| { type: 'retrying'; attempt: number; maxAttempts: number }
| { type: 'error'; message: string }
```
| Type | Description |
|------|-------------|
| `thinking` | Agent is analyzing the page and planning |
| `executing` | Agent is executing a tool action |
| `executed` | Tool execution completed |
| `retrying` | Retrying after a failure |
| `error` | An error occurred |
### HistoricalEvent
```typescript
type HistoricalEvent =
| { type: 'step'; stepIndex: number; reflection: AgentReflection; action: Action }
| { type: 'observation'; content: string }
| { type: 'user_takeover' }
| { type: 'retry'; message: string; attempt: number; maxAttempts: number }
| { 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
interface ExecutionResult {
success: boolean
data: string
history: HistoricalEvent[]
}
```
## Usage Examples
### Basic Execution
```typescript
const result = await window.PAGE_AGENT_EXT!.execute(
'Fill in the email field with test@example.com and click Submit',
{
baseURL: 'https://api.openai.com/v1',
apiKey: process.env.OPENAI_API_KEY!,
model: 'gpt-5.2',
}
)
if (result.success) {
console.log('Task completed:', result.data)
} else {
console.error('Task failed')
}
```
### With Event Hooks
```typescript
await window.PAGE_AGENT_EXT!.execute(
'Navigate to the settings page',
llmConfig,
{
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)
// Later, stop it
window.PAGE_AGENT_EXT!.dispose()
```
## Window Type Declaration
If not using `@page-agent/core`, add this to your project:
```typescript
import type {
AgentActivity,
AgentStatus,
ExecutionResult,
HistoricalEvent,
LLMConfig,
} from '@page-agent/core'
declare global {
interface Window {
PAGE_AGENT_EXT_VERSION?: string
PAGE_AGENT_EXT?: {
version: string
execute: (
task: string,
llmConfig: LLMConfig,
hooks?: {
onStatusChange?: (status: AgentStatus) => void
onActivity?: (activity: AgentActivity) => void
onHistoryUpdate?: (history: HistoricalEvent[]) => void
onDispose?: () => void
}
) => Promise<ExecutionResult>
dispose: () => void
}
}
}
```

View File

@@ -0,0 +1,283 @@
# Page Agent 浏览器插件 API
本文档介绍如何在网页应用中接入 Page Agent 浏览器插件。
## 安装
### 1. 安装浏览器插件
从 Chrome 应用商店安装 Page Agent 插件。
### 2. 安装类型定义(推荐)
```bash
npm install @page-agent/core --save-dev
```
### 3. 配置认证
插件在页面加载后检测 `localStorage` 中的 token匹配时才会注入 API。
1. 打开插件的侧边栏面板,获取授权 token
2. 在页面中设置 token
```typescript
localStorage.setItem('PageAgentExtUserAuthToken', 'your-token')
```
## 快速开始
```typescript
import type {
AgentActivity,
AgentStatus,
ExecutionResult,
HistoricalEvent,
LLMConfig,
} from '@page-agent/core'
// 等待插件注入(最多 1 秒)
async function waitForExtension(timeout = 1000): Promise<boolean> {
const start = Date.now()
while (Date.now() - start < timeout) {
if (window.PAGE_AGENT_EXT) return true
await new Promise((r) => setTimeout(r, 100))
}
return false
}
// 使用
if (await waitForExtension()) {
const result = await window.PAGE_AGENT_EXT!.execute(
'点击登录按钮',
{
baseURL: 'https://api.openai.com/v1',
apiKey: 'your-api-key',
model: 'gpt-5.2',
},
{
onStatusChange: (status) => console.log('状态:', status),
onActivity: (activity) => console.log('活动:', activity),
}
)
console.log('结果:', result)
}
```
## 全局 API
插件在 `window` 对象上注入以下 API
### `window.PAGE_AGENT_EXT_VERSION`
插件版本号字符串(例如 `"1.0.0"`)。单独暴露版本号,方便在访问主 API 对象前进行版本检查。
### `window.PAGE_AGENT_EXT`
主 API 命名空间对象,包含:
#### `PAGE_AGENT_EXT.execute(task, llmConfig, hooks?)`
执行 Agent 任务。
**参数:**
| 名称 | 类型 | 必填 | 说明 |
|------|------|------|------|
| `task` | `string` | 是 | 任务描述 |
| `llmConfig` | `LLMConfig` | 是 | LLM 配置 |
| `hooks` | `ExecuteHooks` | 否 | 事件回调 |
**返回:** `Promise<ExecutionResult>`
#### `PAGE_AGENT_EXT.dispose()`
停止并销毁当前运行的 Agent。
## 类型定义
安装 `@page-agent/core` 获取完整类型:
```typescript
import type {
AgentActivity,
AgentStatus,
ExecutionResult,
HistoricalEvent,
LLMConfig,
} from '@page-agent/core'
export interface ExecuteHooks {
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>
```
### AgentStatus
```typescript
type AgentStatus = 'idle' | 'running' | 'completed' | 'error'
```
| 状态 | 说明 |
|------|------|
| `idle` | 空闲,准备执行 |
| `running` | 正在执行任务 |
| `completed` | 任务成功完成 |
| `error` | 任务执行失败 |
### AgentActivity
```typescript
type AgentActivity =
| { type: 'thinking' }
| { type: 'executing'; tool: string; input: unknown }
| { type: 'executed'; tool: string; input: unknown; output: string; duration: number }
| { type: 'retrying'; attempt: number; maxAttempts: number }
| { type: 'error'; message: string }
```
| 类型 | 说明 |
|------|------|
| `thinking` | Agent 正在分析页面并规划 |
| `executing` | 正在执行工具操作 |
| `executed` | 工具执行完成 |
| `retrying` | 失败后重试 |
| `error` | 发生错误 |
### HistoricalEvent
```typescript
type HistoricalEvent =
| { type: 'step'; stepIndex: number; reflection: AgentReflection; action: Action }
| { type: 'observation'; content: string }
| { type: 'user_takeover' }
| { type: 'retry'; message: string; attempt: number; maxAttempts: number }
| { 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
interface ExecutionResult {
success: boolean
data: string
history: HistoricalEvent[]
}
```
## 使用示例
### 基础执行
```typescript
const result = await window.PAGE_AGENT_EXT!.execute(
'在邮箱输入框填入 test@example.com 然后点击提交',
{
baseURL: 'https://api.openai.com/v1',
apiKey: process.env.OPENAI_API_KEY!,
model: 'gpt-5.2',
}
)
if (result.success) {
console.log('任务完成:', result.data)
} else {
console.error('任务失败')
}
```
### 使用事件回调
```typescript
await window.PAGE_AGENT_EXT!.execute(
'导航到设置页面',
llmConfig,
{
onStatusChange: (status) => {
updateUI({ agentStatus: status })
},
onActivity: (activity) => {
switch (activity.type) {
case 'thinking':
showSpinner('Agent 正在思考...')
break
case 'executing':
showSpinner(`正在执行: ${activity.tool}`)
break
case 'executed':
log(`${activity.tool} 完成,耗时 ${activity.duration}ms`)
break
case 'error':
showError(activity.message)
break
}
},
onHistoryUpdate: (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)
// 稍后停止
window.PAGE_AGENT_EXT!.dispose()
```
## Window 类型声明
如果不使用 `@page-agent/core`,可以添加以下声明:
```typescript
import type {
AgentActivity,
AgentStatus,
ExecutionResult,
HistoricalEvent,
LLMConfig,
} from '@page-agent/core'
declare global {
interface Window {
PAGE_AGENT_EXT_VERSION?: string
PAGE_AGENT_EXT?: {
version: string
execute: (
task: string,
llmConfig: LLMConfig,
hooks?: {
onStatusChange?: (status: AgentStatus) => void
onActivity?: (activity: AgentActivity) => void
onHistoryUpdate?: (history: HistoricalEvent[]) => void
onDispose?: () => void
}
) => Promise<ExecutionResult>
dispose: () => void
}
}
}
```