o ưi@sddZddlmZmZmZddlZddlmZddlm Z m Z m Z m Z ddl mZGddde ZdS) z, Mistral OCR transformation implementation. )AnyDictOptionalN)verbose_logger) BaseOCRConfig DocumentTypeOCRRequestData OCRResponse)get_secret_strc seZdZdZdfdd ZdedefddZd ed ededefd d Z   dd e dede ede ede ede f ddZ dde eded ede edef ddZ deded ed edef ddZdedejdedefddZZS)MistralOCRConfigzl Mistral OCR transformation configuration. Reference: https://docs.mistral.ai/api/#tag/ocr returnNcstdSN)super__init__)self __class__^/home/app/Keep/.python/lib/python3.10/site-packages/litellm/llms/mistral/ocr/transformation.pyrszMistralOCRConfig.__init__modelcCsgdS)a Get supported OCR parameters for Mistral OCR. Mistral OCR supports: - pages: List of page numbers to process - include_image_base64: Whether to include base64 encoded images - image_limit: Maximum number of images to return - image_min_size: Minimum size of images to include - bbox_annotation_format: Format for bounding box annotations - document_annotation_format: Format for document annotations )pagesZinclude_image_base64Z image_limitZimage_min_sizeZbbox_annotation_formatZdocument_annotation_formatr)rrrrrget_supported_ocr_paramss z)MistralOCRConfig.get_supported_ocr_paramsnon_default_paramsoptional_paramscCs6|j|d}i}|D] \}}||vr|||<q |S)z Map OCR parameters to Mistral-specific format. Mistral accepts these parameters directly, so no transformation needed. Just filter out unsupported params. )r)ritems)rrrrZsupported_paramsZ mapped_paramsparamvaluerrrmap_ocr_params1s zMistralOCRConfig.map_ocr_paramsheadersapi_keyapi_baselitellm_paramscKs6|durtd}|durtddd|i|}|S)zJ Validate environment and return headers for Mistral OCR. NZMISTRAL_API_KEYz}Missing Mistral API Key - A call is being made to Mistral but no key is set either in the environment variables or via params AuthorizationzBearer )r ValueError)rrrrr r!kwargsrrrvalidate_environmentGs  z%MistralOCRConfig.validate_environmentcKs4|durd}|d}|dr|dS|dS)zt Get complete URL for Mistral OCR endpoint. Returns: https://api.mistral.ai/v1/ocr Nzhttps://api.mistral.ai/v1/z/v1z/ocrz/v1/ocr)rstripendswith)rr rrr!r$rrrget_complete_urlgs      z!MistralOCRConfig.get_complete_urldocumentcKsLtd|t|tstdt|||d}||t|ddS)a Transform OCR request to Mistral-specific format. Mistral OCR API accepts: { "model": "mistral-ocr-latest", "document": { "type": "document_url", "document_url": "" }, "pages": [0], # optional "include_image_base64": false, # optional ... } Args: model: Model name (e.g., "mistral-ocr-latest") document: Document dict from user (Mistral format) - already validated in main.py optional_params: Already mapped optional parameters headers: Request headers Returns: OCRRequestData with JSON data z+Mistral OCR transform_ocr_request - model: zExpected document dict, got )rr*N)datafiles)rdebug isinstancedictr#typeupdater)rrr*rrr$r+rrrtransform_ocr_requests   z&MistralOCRConfig.transform_ocr_request raw_response logging_objc Kszz&|}td|t|dg|d||d|dddWSty<}z td||d }~ww) a Return Mistral OCR response in native format. Mistral OCR is the standard format for LiteLLM OCR responses. No transformation needed - return native response. Mistral OCR returns: { "pages": [ { "index": 0, "markdown": "extracted text content", "images": [...], "dimensions": {...} }, ... ], "model": "mistral-ocr-2505-completion", "document_annotation": null, "usage_info": {...} } zMistral OCR response keys: rrdocument_annotation usage_infoZocr)rrr5r6objectz$Error parsing Mistral OCR response: N)jsonrr-keysr get Exceptionerror)rrr3r4r$Z response_jsonerrrtransform_ocr_responses  z'MistralOCRConfig.transform_ocr_response)r N)NNNr )__name__ __module__ __qualname____doc__rstrlistrr/rrrr%r)rrr2httpxResponserr r> __classcell__rrrrr sz  %  3r )rBtypingrrrrEZlitellm._loggingrZ(litellm.llms.base_llm.ocr.transformationrrrr Zlitellm.secret_managers.mainr r rrrrs