Passing Document/Image buffer to Orchestrator

[JonBernardHealia]

I'm using bedrock to analyze and pull structured data from documents and images. I'd like to use the orchestrator to preprocess the files and route the request to specialized bedrock agents based on document type.

It looks like I can only pass a userInput string to orchestrator.routeRequest, but I also need to pass a file buffer of the s3 file to get the file's contents.

Any ideas how I can achieve this in the most efficient way? Maybe I pass a text array of s3 file keys and write a tool spec/handler to pull the files from s3?