ocrbase

TypeScript SDK for ocrbase powered by Eden Treaty.

This SDK waits for job completion over WebSockets (no polling).

Install

bun add ocrbase

npm install ocrbase

Quick start

import { parse } from "ocrbase";

const { text } = await parse("./invoice.pdf");
console.log(text);

parse() and extract() automatically read OCRBASE_API_KEY from process.env when apiKey is not provided.

Structured extraction

import { extract } from "ocrbase";

const { object } = await extract("./invoice.pdf", {
  vendor: "string",
  total: "number",
  date: "date",
});

console.log(object.vendor);

Runtime note

By default, the SDK opens a WebSocket automatically:

• Bun: uses Authorization header
• Other runtimes with WebSocket: uses api_key query parameter

If your runtime does not provide WebSocket, pass a custom socket factory:

import { parse } from "ocrbase";

await parse("./invoice.pdf", {
  createWebSocket: (url, apiKey) =>
    new WebSocket(url, {
      headers: { Authorization: `Bearer ${apiKey}` },
    } as never),
});

Advanced client API

Use createOcrBase() when you want reusable client instances and { data, error } results instead of thrown errors.

import { createOcrBase } from "ocrbase";

const ocr = createOcrBase();
const result = await ocr.parse({ file: "./invoice.pdf" });

if (result.error) throw result.error;
console.log(result.data.text);

Input types

file supports:

• File / Blob / Bun.file(...)
• Buffer, ArrayBuffer, typed arrays
• ReadableStream<Uint8Array>
• local file path string (for example "./invoice.pdf")
• remote URL string (for example "https://example.com/invoice.pdf")

await parse("https://example.com/report.pdf");

Schema adapters

extract() accepts:

• Simple schema objects ("string", "number", "boolean", "integer", "date")
• Elysia t.* / TypeBox schema objects
• Zod schemas
• Raw JSON Schema

React Query hooks

import { createOcrBase } from "ocrbase";
import { useOcrExtract, useOcrParse } from "ocrbase/react";

const ocr = createOcrBase({ apiKey: process.env.NEXT_PUBLIC_API_KEY! });

const parseMutation = useOcrParse(ocr);
const extractMutation = useOcrExtract(ocr);