Transcribe_module and loader_images create document objects in different ways, the former using a custom function not from Langchain and saving a transcription as a pickle file. Although this is similar to langchain's "document object", it does have its drawbacks in terms of the custom approach I created not having a method to serialize into json and/or "bytes."
The pickle file is/was implemented to view a file representing a transcription in the create database tab...would be nice to use langchain's "document object" class for consistency as long as I can figure out a better way to view files that will be put into the vector database.
Also need to determine the best way to open, delete, etc.
Transcribe_module and loader_images create document objects in different ways, the former using a custom function not from Langchain and saving a transcription as a pickle file. Although this is similar to langchain's "document object", it does have its drawbacks in terms of the custom approach I created not having a method to serialize into json and/or "bytes."
The pickle file is/was implemented to view a file representing a transcription in the create database tab...would be nice to use langchain's "document object" class for consistency as long as I can figure out a better way to view files that will be put into the vector database.
Also need to determine the best way to open, delete, etc.