churro_ocr.templates

Template helpers for model-specific OCR input rendering.

churro_ocr.templates.build_ocr_conversation(template, page)

Build an OCR conversation from a template or template callable.

Parameters:

• template (OCRPromptTemplate | Callable[[DocumentPage], list[dict[str, Any]]]) – Prompt template object or callable.

• page (DocumentPage) – Page to convert into a conversation.

Returns:

Structured OCR conversation for page.

Return type:

list[dict[str, Any]]

class churro_ocr.templates.HFChatTemplate

Bases: object

Template for processor/tokenizer chat-template OCR models.

Parameters:

• system_message – Optional system message prepended to the conversation.

• user_prompt – Optional user-side text prompt appended with the image.

• include_image – Whether to include the page image in the user message.

__init__(system_message=None, user_prompt=None, include_image=True)

Parameters:

• system_message (str | None)

• user_prompt (str | None)

• include_image (bool)

Return type:

None

build_conversation(page)

Build a structured multimodal conversation for one OCR page.

Parameters:

page (DocumentPage) – Page to represent in the conversation.

Returns:

Conversation payload suitable for chat-template OCR models.

Return type:

list[dict[str, Any]]

class churro_ocr.templates.OCRPromptTemplate

Bases: Protocol

Protocol for OCR templates that build model conversations.

__init__(*args, **kwargs)

build_conversation(page)

Build a model conversation for one page.

Parameters:

page (DocumentPage) – Page to convert into a model-specific prompt payload.

Returns:

Structured conversation ready for backend-specific rendering.

Return type:

list[dict[str, Any]]