In this guide you’ll:

1. upload a PDF
2. receive the processing result via webhook
3. optionally retrieve the document later, and
4. optionally send corrections to improve the model.

Prerequisites

• Your x-api-key (add it to all API requests).
• A type/model (find it in the Models section in Invofox) - only required when not using Classifier premium feature.
• A public HTTPS endpoint (if you want to receive webhooks).

1. Upload a document

Use the upload endpoint to send your first PDF.

• Required: your API key and the type, which you can find in the “Models” section in the interface (unless using Classifier premium feature).
• You can upload multiple files or provide URLs.

👉 Direct Upload endpoint 👉 URL Upload endpoint

2. Use webhooks for a better UX

Instead of polling, set up a webhook so Invofox notifies you when processing is finished.

• document.created event is sent when the document is created and immediately enters processing status (document will be empty at this point)
• document.processed event is sent when extraction is complete with all extracted information (document status becomes processed)
• The document may include a publicState field (pendingCorrection, approved, discarded) for workflow management

👉 Webhook reference

3. [OPTIONAL] Retrieve the document with the documentId

At any time, you can fetch the document again using the documentId you received at upload.

• Useful for reporting, integrations, or audits.

👉 Get document info

4. Send corrections - Best Practice

If the extracted data isn’t 100% accurate, you can send corrected values back to Invofox.

• This not only fixes the document in your system but also helps improve the AI model over time.
• This is a recommended best practice for maintaining high accuracy and continuous model improvement.

👉 Update document info