mistralOcr

Runs Mistral OCR on a PDF document reachable via URL. Extracts the full text content, counts the pages, and collects any images found in the document. Images are returned as objects containing S3 paths. Actual API cost is tracked per-run.

Concurrency is capped at 3 simultaneous API calls. Retries up to 5 times with exponential backoff (base 2s, multiplier 2×, cap 32s).

Parameters

Param	Type	Required	Description
file_url	string (URL)	Yes	URL of the PDF document to process
entry	object	Yes	Entry data dict containing dataset_id and entry_id for routing and cost tracking

Output

Field	Type	Description
text	string	The full extracted text from the document
images	ProcessedOCRImage[]	List of processed image objects (see below)
numPages	integer	Number of pages in the document

Each ProcessedOCRImage:

Field	Type	Description
image_url	string?	S3 path to the uploaded image
image_type	string?	MIME type of the image (e.g. image/jpeg)
width	integer?	Image width in pixels
height	integer?	Image height in pixels

note

The images output is intended to be forwarded as-is to a downstream saveEntries node.

Example

{
  "id": "ocrNode",
  "type": "mistralOcr",
  "data": {
    "label": "Mistral OCR",
    "isExecuted": false,
    "handles": ["inputs", "outputs"],
    "schema": {},
    "params": {
      "file_url": { "value": "{{ @downloadEntry.file_url }}", "isExpression": true, "isAttachedToInputNode": false },
      "entry": { "value": "{{ @fetchEntries }}", "isExpression": true, "isAttachedToInputNode": false }
    },
    "inputs": [], "outputs": [], "errors": []
  },
  "position": { "x": 300, "y": 0 },
  "isSelected": false,
  "isDragging": false
}