Manual Integration
Manual integration - reference only.
We do not recommend manually integrating, and cannot guarantee full support in this case.
Recommended integrations:
Also take a look at the Integration Overview if you are setting up your integration for the first time.
If you would like to download the OpenAPI specification directly, it is available here: https://api-v2.mindee.net/openapi.json
Send a file to the asynchronous processing queue for an inference.
Results can be retrieved in two ways:
Loop on the
polling_urlendpoint in the response until the status isProcessedorFailed(polling)Use a webhook to be notified when the inference is finished. Webhooks can be configured on the platform on a per-model basis.
Inferences taking longer than 1200 seconds will time out.
Form to enqueue an inference.
Model ID to use for the inference.
Upload a file as bytes. Required if url is not provided.
Download the file from a URL, must be secure (HTTPS) and publicly available. Required if file is not provided.
Upload a file as a base 64 string. Used as an alternative to file. Not recommended, for specific use only.
Base64String==Webhook IDs to call after all processing is finished. If empty, no webhooks will be used.
Extract the entire text from the document as strings, and fill the raw_text attribute.
Calculate bounding box polygons for values, and fill the locations attribute of fields
Calculate confidence scores for values, and fill the confidence attribute of fields. Useful for automation.
Use Retrieval-Augmented Generation during inference.
Use an alias to link the file to your own DB. If empty, no alias will be used.
my_file_aliasAdditional text context used by the model during inference.
This document is a receipt from a grocery store.Successful Response
Unprocessable Content
Too Many Requests
POST /v2/inferences/enqueue HTTP/1.1
Host: api-v2.mindee.net
Authorization: YOUR_API_KEY
Content-Type: multipart/form-data
Accept: */*
Content-Length: 267
{
"model_id": "text",
"file": "binary",
"url": "https://example.com",
"file_base64": "Base64String==",
"webhook_ids": [
"text"
],
"raw_text": true,
"polygon": true,
"confidence": true,
"rag": true,
"alias": "my_file_alias",
"text_context": "This document is a receipt from a grocery store."
}{
"job": {
"id": "text",
"model_id": "text",
"filename": "file.jpg",
"alias": "text",
"created_at": "2025-11-20T10:25:26.026Z",
"status": "Processing",
"polling_url": "https://example.com",
"result_url": "https://example.com",
"webhooks": [
{
"id": "text",
"created_at": "2025-11-20T10:25:26.026Z",
"status": "Processing",
"error": {
"status": 404,
"detail": "One or more fields failed validation.",
"title": "Invalid fields in form",
"code": "422-001",
"errors": [
{
"pointer": "#/model_id",
"detail": "'model_id' must be a valid UUID format"
}
]
}
}
],
"error": {
"status": 404,
"detail": "One or more fields failed validation.",
"title": "Invalid fields in form",
"code": "422-001",
"errors": [
{
"pointer": "#/model_id",
"detail": "'model_id' must be a valid UUID format"
}
]
}
}
}Here is a working example of the built-in test client:
Get the status of an inference that was previously enqueued.
If you are using webhooks there is no need to call this route.
When the job status is Processing or Failed the return code will be HTTP 200.
When the job status is Processed the return code will be HTTP 302. The processing result can then be retrieved in two ways:
Recommended: Using a GET on the URL specified in the
result_urlfield of the responseAllowing your client to redirect on the
Locationheader of the response
UUID of the job to retrieve
Successful Response
Found
Not Found
Unprocessable Content
Too Many Requests
GET /v2/jobs/{job_id} HTTP/1.1
Host: api-v2.mindee.net
Authorization: YOUR_API_KEY
Accept: */*
{
"job": {
"id": "text",
"model_id": "text",
"filename": "file.jpg",
"alias": "text",
"created_at": "2025-11-20T10:25:26.026Z",
"status": "Processing",
"polling_url": "https://example.com",
"result_url": "https://example.com",
"webhooks": [
{
"id": "text",
"created_at": "2025-11-20T10:25:26.026Z",
"status": "Processing",
"error": {
"status": 404,
"detail": "One or more fields failed validation.",
"title": "Invalid fields in form",
"code": "422-001",
"errors": [
{
"pointer": "#/model_id",
"detail": "'model_id' must be a valid UUID format"
}
]
}
}
],
"error": {
"status": 404,
"detail": "One or more fields failed validation.",
"title": "Invalid fields in form",
"code": "422-001",
"errors": [
{
"pointer": "#/model_id",
"detail": "'model_id' must be a valid UUID format"
}
]
}
}
}Get the result of an inference that was previously enqueued.
The inference will only be available after it has finished processing.
If you are using webhooks there is no need to call this route. However, the JSON payload described here is what will be sent to each specified webhook's URL.
If you activated the "Delete extracted data when fetched" option for the model, the inference will be deleted after this route completes. Any further requests to the same inference ID will return a 404.
UUID of the inference to retrieve
Successful Response
Not Found
Unprocessable Content
GET /v2/inferences/{inference_id} HTTP/1.1
Host: api-v2.mindee.net
Authorization: YOUR_API_KEY
Accept: */*
{
"inference": {
"id": "text",
"model": {
"id": "text"
},
"file": {
"name": "text",
"alias": "text",
"page_count": 1,
"mime_type": "text"
},
"result": {
"fields": {
"ANY_ADDITIONAL_PROPERTY": {
"locations": [
{
"polygon": [
[]
],
"page": 1
}
],
"confidence": "Certain",
"value": true
}
},
"raw_text": {
"pages": [
{
"content": "text"
}
]
},
"rag": {
"retrieved_document_id": "text"
}
},
"active_options": {
"raw_text": true,
"polygon": true,
"confidence": true,
"rag": true
}
}
}When an inference is processed, this payload is sent to the specified webhooks.
Webhooks are specified in the webhooks_ids parameter when enqueuing an inference.
Response for an inference.
Successful Response
Validation Error
Payload
{
"inference": {
"id": "text",
"model": {
"id": "text"
},
"file": {
"name": "text",
"alias": "text",
"page_count": 1,
"mime_type": "text"
},
"result": {
"fields": {
"ANY_ADDITIONAL_PROPERTY": {
"locations": [
{
"polygon": [
[]
],
"page": 1
}
],
"confidence": "Certain",
"value": true
}
},
"raw_text": {
"pages": [
{
"content": "text"
}
]
},
"rag": {
"retrieved_document_id": "text"
}
},
"active_options": {
"raw_text": true,
"polygon": true,
"confidence": true,
"rag": true
}
}
}Last updated
Was this helpful?