MistralOCRDocumentConverter

Extract text from documents using Mistral's OCR API, with optional structured annotations.

Key Features

• Extracts text from documents using Mistral's OCR model.
• Accepts local files, URLs, and Mistral file IDs as input.
• Automatically uploads local files to Mistral's file storage.
• Outputs one Haystack Document per source, with all pages concatenated using form feed characters for accurate page-wise splitting.
• Supports optional structured annotations using Pydantic schemas for bounding box and document-level analysis.
• Returns both Haystack Documents and raw Mistral OCR API responses.

Configuration

1. Drag the MistralOCRDocumentConverter component onto the canvas from the Component Library.
2. Click on the component to open the configuration panel.
3. On the General tab:
   1. Create a secret with your Mistral API key. Use MISTRAL_API_KEY as the environment variable name. For instructions, see Create Secrets.
4. Go to the Advanced tab to configure model, include_image_base64, pages, and cleanup_uploaded_files.

Connections

MistralOCRDocumentConverter receives a list of file sources. It outputs a list of Document objects containing the extracted text and a list of raw API responses.

Source Code

To check this component's source code, open ocr_document_converter.py in the Haystack Core Integrations repository.

Usage Examples

Basic Configuration

  MistralOCRDocumentConverter:
    type: haystack_integrations.components.converters.mistral.ocr_document_converter.MistralOCRDocumentConverter
    init_parameters:
      api_key:
        type: env_var
        env_vars:
        - MISTRAL_API_KEY
        strict: false
      model: mistral-ocr-2505
      cleanup_uploaded_files: true

Using the Component in a Pipeline

# haystack-pipeline
components:
  MistralOCRDocumentConverter:
    type: haystack_integrations.components.converters.mistral.ocr_document_converter.MistralOCRDocumentConverter
    init_parameters:
      api_key:
        type: env_var
        env_vars:
        - MISTRAL_API_KEY
        strict: false
      model: mistral-ocr-2505

  document_splitter:
    type: haystack.components.preprocessors.document_splitter.DocumentSplitter
    init_parameters:
      split_by: page
      split_length: 1

connections:
- sender: MistralOCRDocumentConverter.documents
  receiver: document_splitter.documents

max_runs_per_component: 100

metadata: {}

Parameters

Inputs

Parameter	Type	Description
sources	List[Union[str, Path, ByteStream, ...]]	A list of file paths, URLs, ByteStream objects, or Mistral file identifiers to convert. Local files are automatically uploaded.
meta	Optional[Union[Dict[str, Any], List[Dict[str, Any]]]]	Metadata to add to all documents, or a list of metadata dicts (one per source).
bbox_annotation_schema	Optional[Type[BaseModel]]	A Pydantic model schema for structured bounding box region annotations.
document_annotation_schema	Optional[Type[BaseModel]]	A Pydantic model schema for structured document-level annotations.

Outputs

Parameter	Type	Description
documents	List[Document]	The extracted text as Haystack Documents. Pages are separated by form feed characters (\f).
raw_mistral_response	List[Dict[str, Any]]	The raw API responses from the Mistral OCR service for each input file.

Init Parameters

These are the parameters you can configure in Pipeline Builder:

Parameter	Type	Default	Description
api_key	Secret	Secret.from_env_var("MISTRAL_API_KEY")	The Mistral API key.
model	str	mistral-ocr-2505	The Mistral OCR model to use.
include_image_base64	bool	False	Whether to include base64-encoded images in the output.
pages	Optional[List[int]]	None	A list of specific page numbers (0-indexed) to extract from multi-page documents. If not set, all pages are extracted.
image_limit	Optional[int]	None	The maximum number of images to process per document.
image_min_size	Optional[int]	None	The minimum image size (in pixels) to include in processing.
cleanup_uploaded_files	bool	True	Whether to delete uploaded files from Mistral's storage after processing.

Run Method Parameters

These are the parameters you can configure for the component's run() method. This means you can pass these parameters at query time through the API, in Playground, or when running a job. For details, see Modify Pipeline Parameters at Query Time.

Parameter	Type	Default	Description
sources	List		A list of file paths or sources to convert.
meta	Optional[Union[Dict, List[Dict]]]	None	Metadata to add to the resulting documents.
bbox_annotation_schema	Optional[Type[BaseModel]]	None	Pydantic schema for bounding box annotations.
document_annotation_schema	Optional[Type[BaseModel]]	None	Pydantic schema for document annotations.

Related Information

• Pipeline Components