@pageindex/sdk

TypeScript SDK for PageIndex document processing.

Get your API Key at dash.pageindex.ai. For full API documentation, see docs.pageindex.ai.

Installation

pnpm add @pageindex/sdk

Requires Node.js >= 18.0.0

Quick Start

import { PageIndexClient } from '@pageindex/sdk';

const client = new PageIndexClient({
  apiKey: 'your-api-key',
});

// Upload a document
const { doc_id } = await client.api.submitDocument(fileBuffer, 'report.pdf');

// List recent documents
const recent = await client.tools.recentDocuments();

// Extract document structure
const structure = await client.tools.getDocumentStructure({ docName: 'report.pdf' });

// Extract page content
const pages = await client.tools.getPageContent({ docName: 'report.pdf', pages: '1-5' });

With explicit resource management (TypeScript 5.2+):

await using client = new PageIndexClient({ apiKey: 'your-api-key' });
const recent = await client.tools.recentDocuments();
// connection closed automatically when scope exits

API

Client

const client = new PageIndexClient({
  apiKey: 'your-api-key',
  folderScope: 'folder-id', // optional, restricts all operations to this folder
});

When folderScope is set, all operations are restricted to the specified folder and its descendants. Per-call folderId can narrow within the scope (e.g. target a subfolder) but cannot access folders outside the boundary. Change it at runtime via client.setFolderScope(id).

Tools

All methods via client.tools:

| Method | Description | | ------------------------------------------------------------------------- | ------------------------ | | recentDocuments({ folderId?, cursor?, limit? }) | List recent uploads | | findRelevantDocuments({ query?, limit?, folderId?, cursor? }) | Search documents | | getDocument({ docName, waitForCompletion?, folderId? }) | Get document details | | getDocumentStructure({ docName, part?, waitForCompletion?, folderId? }) | Extract document outline | | getPageContent({ docName, pages, waitForCompletion?, folderId? }) | Extract page content | | getDocumentImage({ imagePath }) | Retrieve embedded image | | removeDocument({ docNames, folderId? }) | Delete documents | | createFolder({ name, description?, parentFolderId? }) | Create folder | | listFolders({ parentFolderId? }) | List folders |

Page specification formats: "5", "3,7,10", "5-10", "1-3,7,9-12"

API

All methods via client.api:

// Submit a document
const result = await client.api.submitDocument(file, 'document.pdf');

// Get document metadata
const doc = await client.api.getDocument(docId);

// List all documents
const docs = await client.api.listDocuments({ limit: 20, offset: 0 });

// Delete a document
await client.api.deleteDocument(docId);

// Chat completions
const chat = await client.api.chatCompletions({
  messages: [{ role: 'user', content: 'Summarize the document' }],
  doc_id: docId,
});

Error Handling

import { PageIndexError } from '@pageindex/sdk';

try {
  await client.tools.getDocument({ docName: 'xxx' });
} catch (e) {
  if (e instanceof PageIndexError) {
    // e.code: 'NOT_FOUND' | 'UNAUTHORIZED' | 'RATE_LIMITED' | 'USAGE_LIMIT_REACHED' | ...
    // e.statusCode: HTTP status code
  }
}

Documentation

• API Quickstart — Get started with document processing
• API Endpoints — Full REST API reference
• Python SDK — Python client (tree, chat, OCR)
• MCP Integration — Use PageIndex with AI agents

Examples

See examples/chat-demo for Next.js + AI SDK integration.

License

MIT