Not all implementations of DocumentReader allow adding metadata to a document during the document loader step. Only TextReader and JsonReader allow it, using different methods.
I have added support for custom metadata in the following implementations, using the same structure as in the TextReader implementation:
- PagePdfDocumentReader
- ParagraphPdfDocumentReader
- TikaDocumentReader
I have updated the tests as well. However, the ParagraphPdfDocumentReader currently only checks for a negative case because I don't have any PDF with the correct license and structure to create a positive case.