Closed (fixed)
Project:
OpenAI Provider
Version:
1.2.x-dev
Component:
Code
Priority:
Normal
Category:
Feature request
Assigned:
Unassigned
Reporter:
Created:
11 Jun 2025 at 13:59 UTC
Updated:
23 Sep 2025 at 12:19 UTC
Jump to comment: Most recent
Comments
Comment #4
bircherComment #5
b_sharpe commentedThis may be more a future option, but it's far more efficient to upload a file and then reference the file rather than send it as filedata. base64-encoded content is treated like raw input. The model must process the file contents as tokens which can get pretty expensive. Also, not very scalable for large files and can even get truncated. I believe it also counts against your context window..
I'd be in favor of looking at how we could use the file upload + thread to enhance this
Comment #6
marcus_johansson commentedComment #7
marcus_johansson commentedTested and working. This is getting merged. Thanks everyone.