Golang client

Provide a client library for integrating the Mindee API using the Go language. Support for: POST using polling and webhook methods loading files from various sources (path, bytes, File object, etc) PDF splitting before sending Must be able to handle dynamic outputs in a simple way.

Possibility to convert and download extracted data to csv.

Convert a bundle of documents into CSV. To be able to then upload the document in a financial control system

Add the possibility to roll back when I create my model & personalise my data schema

when using the live interface, I find the zoom rate good on mouse but very fast on trackpad.

Any proposed solutions for this ?

Lorsque je me connecte sur Mindee, j'arrive sur mon compte et j'ai cécran noir...

[WEBHOOK] Allow an optional callback_url parameter on the enqueue call, as an alternative to webhook_ids

Today webhooks must be pre-registered per model in the platform UI, and the enqueue call only accepts webhook IDs (webhook_ids). There is no way to give a URL at request time. This does not scale well for us: We have N models and several environments (local, dev, staging, prod), so we have to create and maintain N x M endpoints by hand in the UI, plus one per developer for local work. Each environment then needs its own MINDEE_V2_WEBHOOK_ID value to keep in sync, so configuration lives half in your UI and half in our code. For local development the tunnel URL (ngrok, localtunnel) changes regularly, so a developer has to go back into the UI and edit the endpoint every time.

n8n integration

Hello, I am having a persistent authorization issue while trying to connect to the Invoice OCR API from my n8n.cloud workflow. I have already: Confirmed that the "Invoice OCR" model is active in my Mindee dashboard. Generated a brand new, clean API key from the "API keys" section of my account. Copied this new API key and created a new credential in the official Mindee node in n8n. The node is configured correctly to predict an invoice from a binary file (attachment_0 from a Gmail trigger). Despite these steps, I consistently receive a 401 Unauthorized error with the message "Invalid token provided" every time I execute the node. I have tried generating multiple new keys, and the result is always the same.

Supplier name parsed from logo instead of legal entity

We are seeing incorrect results for the supplier name field. The OCR returns the brand name from the page LOGO instead of the legal entity that issued the document. Example: on a receipt the header logo reads "Make", while the actual supplier (legal entity, tied to the address and US EIN 61-1797223) is "Celonis Inc.". The API returns "Make". What we have already tried, without success: Post-processing with a RAG step Adding explicit field guidance instructing the model to prefer the legal entity near the address and tax ID and to ignore header logos Neither resolves the issue. The model still favors the visually dominant logo over the legal name in the sender block. Questions: Is there a recommended way to make supplier_name prefer the legal entity (near supplier_address and supplier_company_registrations) instead of the header logo? Can this layout be submitted for model review or correction? If we annotate sample documents, will that improve extraction for this case? Link to document - https://app.mindee.com/model/701fa3d9-d54b-4e7c-a506-7efed7a5079e/review?docs=98b124b8-2349-4a9a-82c6-6b7699a4a07c File name - 5dea48c8-35ad-4df5-8543-1aa1164691a2.pdf Happy to share sample documents. Thanks for your help.

In the invoice template the address is read and in the parsed portion of the address mindee keeps leaving the street number and sometimes the street name blank

Add PDF merge to Split Model

Allow a user to apply the split page ranges returned by the API to the local file sent.

Add image extraction to Crop

Allow a user to apply the crop coordinates returned by the API to the local file sent.

Issues with RAG Training Documents

We are currently facing issues while training documents in RAG. The issues are listed below: Trained documents are not being displayed in the list. Pagination is not showing properly. When we upload a document in RAG, it does not appear immediately. After refreshing the page, the uploaded document becomes visible. Once we validate the document, it disappears again from the list. Could you please check this issue and let us know why this is happening?

PDF Content Splitter - split multipage PDF into individual Invoices/Receipts

We need to handle scenario where a single, multi-page PDF is automatically split into separate files, each representing a distinct transaction. Use case: Client uploads a batch scan of invoices or receipts. Each transaction can also be multi-page. Example: 50 page PDF containing a mix of single page and multi-page invoices. Some can have attachments. The resulting split should ALWAYS start with 1st page of invoice. If there are attachments, these do not need to be split apart and can be part of the resulting transaction. In our experience, each invoice starts with Invoice page 1 and hence that can serve as an indicator where to split the file.

Add bounding boxes to extracted fields

It would be helpful to include the coordinates of the bounding boxes for each extracted field in the output. This would allow us to visually highlight extracted data on the original document, improving validation, debugging, and user interface capabilities. When reviewing extracted fields in our UI, having the ability to draw boxes around the relevant text on the document would enhance the user experience and make it easier to verify extraction accuracy.

Extract picture and images

Hi, We would like a feature that allows us to retrieve a picture from a resume that has been sent to analyze via API. The best option would be to have, in the list of extracted data from the “Resume OCR”, a new element : { "image": [ {“type”: “picture|logo|other”, "id": "document_unique_id", '“dimension”: “widthXheight”, “location”:”coordinates” } ] } type : would describe the type of images ; eventually this could also be added as a filter in the OCR request document_unique_id : would be the id of the extracted document, available for download via Mindee API widthXheight : the dimensions of the picture, width by height, for example “150×450” coordinates : the abscisse and ordinates of the pictures in the original documented, for example “ 300;100” The coordinates would be usable by ImageMagick. Thanks

QR / Barcode value decode

When the image or document contains barcodes or QR codes in addition to text or photos, it should interpret them and returns their values, not just the object's coordinates. This would be useful for scanning tickets, boarding passes, and all documents that have barcodes to be read.

Zapier integration

Update the zapier integration to support v2.

API Error

My API key is returning '401 Unauthorized - Invalid token provided' when making parse requests to the InternationalIdV2 endpoint. The client initializes successfully and document creation works, but all parse calls fail with authorization errors. Anything I should Optimized? thanks

Ruby Client Library

Provide a client library for integrating the Mindee API using the Ruby language. Support for: inference using polling and webhook methods loading files from various sources (path, bytes, File object, etc) PDF splitting before sending Must be able to handle dynamic outputs in a simple way.

Page-Specific Data Extraction

I’m working with a 16-page credit score report, but the information I need to extract is located only on page 8. It would be extremely useful to have the ability to target a specific page for parsing and extraction, rather than processing the entire document.

Add confidence score

I need to know when the models are sure of what they extract. I’d need a confidence score associated to each extracted value in the JSON

Adapt model data schema next to live test

Hey! It would be great to see the live test and be able to modify the data schema on the same page. Tedious to do back and forth between the two tabs every time.