Implementation:Mistralai Client python OCRResponse Model

Knowledge Sources	Mistral Client Python Mistral AI OCR
Domains	Document_Processing, OCR
Last Updated	2026-02-15 14:00 GMT

Overview

Concrete tool for extracting structured document content from OCR API responses provided by the OCRResponse and OCRPageObject models.

Description

OCRResponse contains a pages list of OCRPageObject models. Each page has: markdown (extracted text), images (list of OCRImageObject with id, base64 data, and bounding box), tables (list of OCRTableObject), dimensions (OCRPageDimensions with width/height), and index (page number). The top-level response also has usage_info (OCRUsageInfo) and optional document_annotation.

Usage

Access the OCRResponse from client.ocr.process(). Iterate over response.pages and access .markdown, .images, and .tables for each page.

Code Reference

Source Location

• Repository: client-python
• File: src/mistralai/client/models/ocrresponse.py (L1-69), ocrpageobject.py (L1-91)

Signature

class OCRPageObject(BaseModel):
    index: int
    markdown: str
    images: List[OCRImageObject]
    tables: List[OCRTableObject]
    dimensions: OCRPageDimensions

class OCRResponse(BaseModel):
    pages: List[OCRPageObject]
    model: str
    usage_info: OCRUsageInfo
    document_annotation: Optional[str] = None

Import

from mistralai.models import OCRResponse
# Typically received as return value from client.ocr.process()

I/O Contract

Inputs

Name	Type	Required	Description
response	OCRResponse	Yes	Return value from ocr.process()

Outputs

Name	Type	Description
pages	List[OCRPageObject]	Per-page extraction results
page.markdown	str	Full page text as markdown
page.images	List[OCRImageObject]	Extracted images
page.tables	List[OCRTableObject]	Extracted tables
usage_info	OCRUsageInfo	Token consumption

Usage Examples

Process OCR Results

response = client.ocr.process(
    model="mistral-ocr-latest",
    document=DocumentURLChunk(document_url="https://example.com/doc.pdf"),
    include_image_base64=True,
)

# Process each page
for page in response.pages:
    print(f"\n--- Page {page.index} ({page.dimensions.width}x{page.dimensions.height}) ---")

    # Extract text
    print(page.markdown)

    # Process images
    for img in page.images:
        print(f"  Image: {img.id} ({img.top_left_x},{img.top_left_y})")
        if img.image_base64:
            # Save or process base64 image data
            pass

    # Process tables
    for table in page.tables:
        print(f"  Table: {table.markdown}")

# Usage info
print(f"\nTokens: {response.usage_info.pages_processed} pages processed")

Related Pages

Implements Principle

• Principle:Mistralai_Client_python_OCR_Result_Processing