Feature Request: Custom classifier/splitter based on file metadata

Currently document AI custom classifier/splitter classifies the documents based on the file contents. But the file metadata can also include several important keys to help in classification. For example user can upload files with the name "W2-2024" which can easily be used to increase the weight of the file being classified as a W-2. This would be beneficial to boost the score of ambiguous documents.

For the implementation perspective, I would like the request to inlclude an additional optional field of metadata which and on the model, this can be implemented as a feature flag where if the metadata input is enabled, the models result will also be weighted based on the metadata input.

Currently I am trying to implement this manually on my server but if the source could return results directly based on the metadata then it would be a huge win.