Summarize PDF documents with AI

🤖/ai/chat sends files from earlier Steps to a large language model along with your prompt, and writes the answer to a result file. This demo passes an uploaded PDF to Claude Sonnet and asks for JSON that matches the schema parameter: a title, the kind of document, a short summary, and a list of key points.

A 🤖/file/filter Step first rejects anything other than a PDF of at most 20 pages, so a large upload can't run up the model bill.

The schema parameter is a string that contains a JSON Schema, so your application can read the result without parsing free-form text. Set format to text instead when you want a plain answer file, for example a summary to show in an email or a CMS field.

PDFs need a model that can read them, such as the Anthropic and Google models. For scanned documents, Word files, or recordings, add a Step that converts or transcribes them first and pass that Step to use. The AI file processing guide has complete examples for invoices, PDF questions, recording summaries, and image descriptions.

This demo sets test_credentials, which uses Transloadit's provider keys and bills the model usage at provider cost plus a markup. In production, store your own provider key as a Template Credential and pass its name to credentials.

Step 1

Handle uploads

We can handle uploads of your users directly. Learn more

bitcoin.pdf

application/pdf · 180 KB · 1273 × 1648

":original": {
  "robot": "/upload/handle"
}

/upload/handle

This bot receives uploads that your users throw at you from browser or apps, or that you throw at us programmatically

Step 2

Filter files

We let you reject, or direct files against different encoding Steps. Learn more

"pdfs": {
  "use": ":original",
  "robot": "/file/filter",
  "accepts": [
    [
      "${file.mime}",
      "==",
      "application/pdf"
    ],
    [
      "${file.meta.page_count}",
      "<=",
      "20"
    ]
  ],
  "condition_type": "and",
  "error_on_decline": true
}

/file/filter

This bot directs files to different encoding Steps based on your conditions

Step 3

Generate AI chat responses

We offer Artificial Intelligence as a service, so you don't have to run your own AI models or install complicated software. Learn more

result.json

application/json · 1.7 KB

"summarized": {
  "use": "pdfs",
  "robot": "/ai/chat",
  "result": true,
  "format": "json",
  "messages": "Summarize the attached PDF for someone deciding whether to read it. Give its title, the kind of document it is, a summary of at most three sentences, and up to five key points. Use only what the document says.",
  "model": "anthropic/claude-sonnet-5-5",
  "reasoning_effort": "low",
  "schema": "{\"type\":\"object\",\"properties\":{\"title\":{\"type\":\"string\"},\"document_type\":{\"type\":\"string\"},\"summary\":{\"type\":\"string\"},\"key_points\":{\"type\":\"array\",\"items\":{\"type\":\"string\"}}},\"required\":[\"title\",\"document_type\",\"summary\",\"key_points\"],\"additionalProperties\":false}",
  "test_credentials": true
}

/ai/chat

This bot analyzes files and prompts with supported LLMs from providers including Anthropic, OpenAI, and Google. Extract structured fields, summarize documents and transcripts, answer questions about PDFs, or draft image descriptions

Step 4

Export files to Amazon S3

We export to the storage platform of your choice. Learn more

:original

https://demos.transloadit.com/8b/b24906f14643289c3749ef0ffeb6df/bitcoin.pdf

summarized

https://demos.transloadit.com/42/c0e4ff5c3d4c21a81aec9e0ce52361/result.json

"exported": {
  "use": [
    "summarized",
    ":original"
  ],
  "robot": "/s3/store",
  "credentials": "demo_s3_credentials",
  "url_prefix": "https://demos.transloadit.com/"
}

/s3/store

This bot exports encoding results to Amazon S3

Live Demo. See for yourself

This live demo is powered by Uppy, our open source file uploader that you can also use without Transloadit, and tus, our open protocol for resumable file uploads that is making uploading more reliable across the world.

Build this in your own language

<!-- This pulls Uppy from our CDN -->
<!-- For smaller self-hosted bundles, install Uppy and plugins manually: -->
<!-- npm i --save @uppy/core @uppy/dashboard @uppy/remote-sources @uppy/transloadit ... -->
<link
  href="https://releases.transloadit.com/uppy/v3.10.0/uppy.min.css"
  rel="stylesheet"
/>
<button id="browse">Select Files</button>
<script type="module">
  import {
    Uppy,
    Dashboard,
    ImageEditor,
    RemoteSources,
    Transloadit,
  } from 'https://releases.transloadit.com/uppy/v3.10.0/uppy.min.mjs'
  const uppy = new Uppy()
    .use(Transloadit, {
      waitForEncoding: true,
      alwaysRunAssembly: true,
      assemblyOptions: {
        params: {
          // It's often better store encoding instructions in your account
          // and use a `template_id` instead of adding these steps inline
          steps: {
            ':original': {
              robot: '/upload/handle',
            },
            pdfs: {
              use: ':original',
              robot: '/file/filter',
              accepts: [['${file.mime}', '==', 'application/pdf'], ['${file.meta.page_count}', '<=', '20']],
              condition_type: 'and',
              error_on_decline: true,
            },
            summarized: {
              use: 'pdfs',
              robot: '/ai/chat',
              result: true,
              format: 'json',
              messages: 'Summarize the attached PDF for someone deciding whether to read it. Give its title, the kind of document it is, a summary of at most three sentences, and up to five key points. Use only what the document says.',
              model: 'anthropic/claude-sonnet-5-5',
              reasoning_effort: 'low',
              schema: '{"type":"object","properties":{"title":{"type":"string"},"document_type":{"type":"string"},"summary":{"type":"string"},"key_points":{"type":"array","items":{"type":"string"}}},"required":["title","document_type","summary","key_points"],"additionalProperties":false}',
              test_credentials: true,
            },
            exported: {
              use: ['summarized', ':original'],
              robot: '/s3/store',
              credentials: 'demo_s3_credentials',
              url_prefix: 'https://demos.transloadit.com/',
            },
          },
        },
      },
    })
    .use(Dashboard, { trigger: '#browse' })
    .use(ImageEditor, { target: Dashboard })
    .use(RemoteSources, {
      companionUrl: 'https://api2.transloadit.com/companion',
    })
    .on('complete', ({ transloadit }) => {
      // Due to `waitForEncoding:true` this is fired after encoding is done.
      // Alternatively, set `waitForEncoding` to `false` and provide a `notify_url`
      console.log(transloadit) // Array of Assembly Statuses
      for (const assembly of transloadit) {
        console.log(assembly.results) // Array of all encoding results
      }
    })
    .on('error', (error) => {
      console.error(error)
    })
</script>

Read docs: Browsers

So many ways to integrate

Transloadit is a service for companies with developers. And there are many ways developers can put Transloadit to good use inside your company to automate media processing.