In previous versions File uploads used to be passed in bytes content only. Now this is being passed as a path only (all my test have None for content so far, maybe I'm missing something).
I use the file upload modality for users to upload documents, which are piped to a vector store. I used the byte content to save the files to a temporary directory, wiped out after processing. Not for multimodal LLM cases.
These files are now persistent in the .files path (Again, am I missing something), which are to be sincere, just bloating my deployment.
Can this behavior be configurable somehow? I didn't see anything in the documentation site and getting byte data from .content made more sense to me.
Python 3.10 and 3.11 Chainlit 1.0.301
In previous versions File uploads used to be passed in bytes content only. Now this is being passed as a path only (all my test have None for content so far, maybe I'm missing something).
I use the file upload modality for users to upload documents, which are piped to a vector store. I used the byte content to save the files to a temporary directory, wiped out after processing. Not for multimodal LLM cases.
These files are now persistent in the .files path (Again, am I missing something), which are to be sincere, just bloating my deployment.
Can this behavior be configurable somehow? I didn't see anything in the documentation site and getting byte data from .content made more sense to me.
Hope this makes sense for the long run.
Best, JC