reverse-engineer

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Reverse Engineering Perplexity AI APIs

逆向工程Perplexity AI APIs

Discover, document, and implement undocumented Perplexity AI endpoints for pplx-sdk.

发现、记录并为pplx-sdk实现未公开的Perplexity AI接口。

When to use

适用场景

Use this skill when:

Discovering new API endpoints from perplexity.ai browser traffic
Decoding request/response payloads from captured cURL commands
Mapping SSE event types from streaming endpoints
Extracting auth flows (cookies, tokens, session management)
Adding new endpoint support to the SDK

在以下场景使用本技能：

从perplexity.ai浏览器流量中发现新的API接口
解码捕获的cURL命令中的请求/响应负载
映射流式接口的SSE事件类型
提取认证流程（Cookie、令牌、会话管理）
为SDK添加新接口支持

Instructions

操作步骤

Step 1: Capture Traffic

步骤1：捕获流量

DevTools (F12) → Network → XHR/Fetch
1. Perform action on perplexity.ai
2. Right-click request → Copy as cURL
3. Paste cURL for analysis

开发者工具（F12）→ 网络 → XHR/Fetch
1. 在perplexity.ai上执行操作
2. 右键点击请求 → 复制为cURL
3. 粘贴cURL进行分析

Step 2: Decode Payload

步骤2：解码负载

Extract from the captured request:

URL path (e.g.,
```
/rest/sse/perplexity.ask
```
)
Method (GET, POST, PUT, DELETE)
Auth (Bearer token from cookie
```
pplx.session-id
```
)
Request body (JSON payload with field names and types)
Response format (JSON, SSE stream, etc.)

从捕获的请求中提取：

URL路径（例如：
```
/rest/sse/perplexity.ask
```
）
请求方法（GET、POST、PUT、DELETE）
认证信息（来自Cookie
```
pplx.session-id
```
的Bearer令牌）
请求体（包含字段名和类型的JSON负载）
响应格式（JSON、SSE流等）

Step 3: Document Schema

步骤3：记录Schema

For each field in the payload:

Name, type, required/optional, default value
What it controls (e.g.,
```
mode: "concise"
```
= shorter answers)
Discovered values (e.g.,
```
model_preference
```
= "pplx-70b-chat" | "pplx-70b-deep")

针对负载中的每个字段：

名称、类型、必填/可选、默认值
功能说明（例如：
```
mode: "concise"
```
= 更简短的回答）
已发现的取值（例如：
```
model_preference
```
= "pplx-70b-chat" | "pplx-70b-deep"）

Step 4: Map to SDK Architecture

步骤4：映射到SDK架构

Discovery	Target	Action
New endpoint	`transport/`	Add endpoint constant and method
Request fields	`domain/models.py`	Create Pydantic request model
Response fields	`domain/models.py`	Create Pydantic response model
SSE events	`transport/sse.py`	Add event type to parser
Auth variation	`shared/auth.py`	Add extraction method
Error codes	`core/exceptions.py`	Map to exception type

研究成果	目标文件	操作
新接口	`transport/`	添加接口常量和方法
请求字段	`domain/models.py`	创建Pydantic请求模型
响应字段	`domain/models.py`	创建Pydantic响应模型
SSE事件	`transport/sse.py`	为解析器添加事件类型
认证变体	`shared/auth.py`	添加提取方法
错误码	`core/exceptions.py`	映射到异常类型

Step 5: Implement

步骤5：实现

Create the SDK implementation following the layered architecture:

Model in
```
domain/models.py
```
Service in
```
domain/<service>.py
```
Tests in
```
tests/test_<service>.py
```

遵循分层架构创建SDK实现：

在
```
domain/models.py
```
中定义模型
在
```
domain/<service>.py
```
中编写服务逻辑
在
```
tests/test_<service>.py
```
中编写测试用例

Known API Surface

已知API范围

Implemented

已实现

```
POST /rest/sse/perplexity.ask
```
— Main query endpoint (SSE streaming)

```
POST /rest/sse/perplexity.ask
```
— 主要查询接口（SSE流式传输）

Discovered (not yet implemented)

已发现（未实现）

Thread management (
```
/rest/threads/
```
)
Collections (
```
/rest/collections/
```
)
User profile (
```
/rest/user/
```
)
File upload (
```
/rest/upload
```
)
Sharing (
```
/rest/share/
```
)

线程管理（
```
/rest/threads/
```
）
收藏集（
```
/rest/collections/
```
）
用户资料（
```
/rest/user/
```
）
文件上传（
```
/rest/upload
```
）
分享功能（
```
/rest/share/
```
）

Anti-Detection Notes

反检测注意事项

Use realistic
```
User-Agent
```
headers matching Chrome/Safari

Include

Referer: https://www.perplexity.ai/

and

Origin

headers

Consider TLS fingerprinting via curl_cffi with
```
impersonate="chrome120"
```
Rate limit requests to avoid detection

使用符合Chrome/Safari的真实
```
User-Agent
```
请求头

包含

Referer: https://www.perplexity.ai/

和

Origin

请求头

考虑通过curl_cffi使用
```
impersonate="chrome120"
```
进行TLS指纹模拟
限制请求速率以避免被检测到