@intuned/files

Intuned SDK for file operations. Provides helpers for converting files to markdown and extracting tables from files, backed by the Intuned file APIs. Must be used within Intuned projects and requires the dependency @intuned/runtime.

Installation

yarn add @intuned/files
# @intuned/runtime is an optional peer dependency used when running inside the
# Intuned runtime (for auth and tracking).
# @intuned/browser is an optional peer dependency, only needed for the
# `download` file source (see below).

Usage

Extract markdown

import { extractMarkdownFromFile } from "@intuned/files";

const markdown = await extractMarkdownFromFile({
  type: "pdf",
  url: "https://example.com/report.pdf",
});

Extract tables

extractTablesFromFile returns each table's page number, title, and cell content.

import { extractTablesFromFile } from "@intuned/files";

const tables = await extractTablesFromFile({
  type: "pdf",
  url: "https://example.com/report.pdf",
});

Extract structured data

extractStructuredDataFromFile converts the file to markdown and then extracts data against a schema. It requires the optional @intuned/browser peer dependency (the other operations do not); it throws a clear error at call time if it is missing.

import { extractStructuredDataFromFile } from "@intuned/files";

const data = await extractStructuredDataFromFile(
  { type: "pdf", url: "https://example.com/invoice.pdf" },
  {
    dataSchema: {
      type: "object",
      properties: {
        invoiceNumber: { type: "string" },
        total: { type: "number" },
      },
      required: ["invoiceNumber", "total"],
    },
  },
);

dataSchema accepts a JSON Schema object or a zod schema.

File sources

Provide exactly one source key on the file object:

await extractMarkdownFromFile({
  type: "pdf",
  url: "https://example.com/report.pdf",
});
await extractMarkdownFromFile({ type: "pdf", base64: "<base64 string>" });
await extractMarkdownFromFile({ type: "pdf", buffer: someBuffer });

// `download` accepts the result of downloadFile from @intuned/browser,
// awaited or not. Requires @intuned/browser to be installed.
import { downloadFile } from "@intuned/browser";
await extractMarkdownFromFile({
  type: "pdf",
  download: downloadFile({ page, trigger }),
});

Supported file types

pdf, image, spreadsheet (.xlsx), and document (.docx). Sources can be a url, a base64 string, a Node.js buffer, or a download.

Development

yarn install
yarn tsc      # typecheck
yarn lint     # lint
yarn build    # build to dist/
yarn test     # run unit tests