GoogleInternal

A Node.js/TypeScript library designed to simplify interaction with Google's internal batchexecute RPC protocol. It handles complex authentication (SAPISIDHASH), declarative spec mapping, and batching.

Features

• Spec Registry: Map obfuscated Google RPC IDs to friendly JSON objects with Zod validation.
• Auth Module: Automatic SAPISIDHASH generation and cookie parsing.
• Transport Layer: Handles XSSI prefix stripping, length-prefixed chunking, and double-JSON encoding.
• Batching: Send multiple RPCs across different services in a single POST request.
• Streaming: Support for length-prefixed streaming responses.
• RPC Scraper: Automatically extract RPC mappings from any Google application by parsing AST structures.

Usage

1. Initialize the Client

The client holds global configuration like cookies and the request origin, which are used to generate the required SAPISIDHASH headers.

import { GoogleInternal } from './src';

const client = new GoogleInternal({
  cookies: 'SAPISID=...; HSID=...; SID=...;',
  origin: 'https://docs.google.com'
});

2. Define Services and Specs

A Service represents a specific Google endpoint (e.g., Drive, Docs). A Spec defines how a specific RPC call within that service works:

• rpcId: The obfuscated 6-character ID (e.g., zx9ptd).
• schema (optional): A Zod schema to validate and type-safe the input data.
• mapArgs: A function that transforms the validated data into the raw array format Google expects.
• mapResult: A function that transforms the raw array response back into a friendly object.

import { z } from 'zod';

// 1. Register the service
const drive = client.registerService('drive', {
  baseUrl: 'https://docs.google.com/_/DrivePicker/data/batchexecute'
});

// 2. Define the input schema
const ListFilesSchema = z.object({
  folderId: z.string().default('root'),
  pageSize: z.number().max(100).default(50)
});

// 3. Register the RPC spec
drive.register('list_files', {
  rpcId: 'zx9ptd',
  schema: ListFilesSchema,
  mapArgs: (data) => [data.folderId, data.pageSize],
  mapResult: (arr) => ({
    files: arr[0].map((f: any) => ({ id: f[0], name: f[1] }))
  })
});

3. Execute Calls

Once registered, you can call services by name. The input will be validated against the Zod schema before being sent.

// Call via the service instance
const { files } = await drive.execute('list_files', { folderId: 'root' });

// Or call via the client
const driveService = client.service('drive');
const result = await driveService.execute('list_files', { pageSize: 10 });

Scraping RPC Mappings

You can automatically extract RPC IDs and their corresponding service methods from any Google application that uses batchexecute.

npm run scrape-rpc <app-url>

Example:

npm run scrape-rpc https://gemini.google.com/app

The script will:

1. Fetch the application's base JavaScript.
2. Dynamically discover and download all registered modules.
3. Perform a 2-phase AST analysis to identify the RPC registration class and extract all mapping definitions.
4. Output a sorted rpc_mappings.txt file.

Credits

Special thanks to bedros-p for sharing the foundational knowledge on Google's module fetching system and for the core logic used in the RPC registration class discovery.

License

MIT