Currently, it looks like the invoice import uses Mistral OCR to extract the contents of the PDF, then a different model to parse the text into structured data.
However, Mistral offers an API where you can provide a PDF and a schema, and it'll extract data into the provided schema, with a single API call: https://docs.mistral.ai/capabilities/document_ai/annotations
Would it be difficult to update account_document_import_llm to support this API?
Posted from my phone - sorry for any typos.
Currently, it looks like the invoice import uses Mistral OCR to extract the contents of the PDF, then a different model to parse the text into structured data.
However, Mistral offers an API where you can provide a PDF and a schema, and it'll extract data into the provided schema, with a single API call: https://docs.mistral.ai/capabilities/document_ai/annotations
Would it be difficult to update account_document_import_llm to support this API?
Posted from my phone - sorry for any typos.