Last updated: January 6

Let's Build: AI invoice parser with Transloadit

Kevin van Zonneveld

Co-founder · Amsterdam, The Netherlands ·

Invoices are a core workflow for finance teams, but extracting fields by hand is slow and error prone. In this Let’s Build, we’ll assemble a small invoice parsing service using Transloadit’s primitives: route files, normalize formats, and ask AI for structured JSON.

What we’re building

Accept any invoice format (PDF, DOCX, image)
Convert non-PDFs to PDF with 🤖 /document/convert
Extract structured fields with 🤖 /ai/chat

Schema first (zod)

We’ll use Zod to define the invoice schema and hand the JSON Schema to the AI step. This keeps the output predictable.

import { z } from 'zod'
import { zodToJsonSchema } from 'zod-to-json-schema'

const invoiceSchema = z.object({
  vendor_name: z.string().optional(),
  invoice_number: z.string().optional(),
  invoice_date: z.string().optional(),
  due_date: z.string().optional(),
  currency: z.string().optional(),
  subtotal: z.number().optional(),
  tax: z.number().optional(),
  total: z.number().optional(),
  line_items: z
    .array(
      z.object({
        description: z.string().optional(),
        quantity: z.number().optional(),
        unit_price: z.number().optional(),
        total: z.number().optional(),
      }),
    )
    .optional(),
})

const toJsonSchema = (schema: z.ZodTypeAny): unknown =>
  typeof (z as { toJSONSchema?: typeof zodToJsonSchema }).toJSONSchema === 'function'
    ? (z as { toJSONSchema: typeof zodToJsonSchema }).toJSONSchema(schema)
    : zodToJsonSchema(schema)

const invoiceJsonSchema = toJsonSchema(invoiceSchema)

Pipeline overview

PDF invoices ──▶ /ai/chat
Other docs  ──▶ /document/convert (pdf) ──▶ /ai/chat

Assembly snippet

const assembly = await transloadit.createAssembly({
  params: {
    steps: {
      pdf_verified: {
        robot: '/file/filter',
        use: ':original',
        accepts: [['${file.mime}', '==', 'application/pdf']],
      },
      non_pdf: {
        robot: '/file/filter',
        use: ':original',
        accepts: [['${file.mime}', '!=', 'application/pdf']],
      },
      pdf_converted: {
        robot: '/document/convert',
        use: 'non_pdf',
        format: 'pdf',
      },
      extract_invoice: {
        robot: '/ai/chat',
        use: ['pdf_verified', 'pdf_converted'],
        model: 'anthropic/claude-4-sonnet-20250514',
        format: 'json',
        schema: JSON.stringify(invoiceJsonSchema),
        messages: 'Extract invoice fields. Omit keys if unknown.',
        result: true,
      },
    },
  },
  files: {
    document: filePath,
  },
  waitForCompletion: true,
})

Try the example app

The runnable version of this post lives at example_apps/invoice-parser/run.ts.

node example_apps/invoice-parser/run.ts

Suggested inputs for visuals

Use the bundled samples or swap in your own invoice PDFs for screenshots:

_assets/demos/inputs/scan.pdf
_assets/demos/inputs/first_document.pdf
_assets/demos/inputs/second_document.pdf

Next steps

Add multi-currency normalization in your application layer.
Store results in your database and add human review queues.
Expand the schema to support multiple vendors and locales.