File Context (OCR) with agents broken

[michnovka]

What happened?

After long struggle with my agent ignoring the File Context (OCR) PDFs I uploaded for context, I decided to debug it and found out the PDFs are totally ignored. I made a reproducer.

Version Information

ghcr.io/danny-avila/librechat-dev latest c9b448016df0 4 days ago 1.05GB

Steps to Reproduce

1. create new agent ColorTest (I used openAI gpt-4.1) with instructions:

You are an agent specilized in saying what color I like the most from given information. if unsure ask.

2. upload this PDF - yellow.pdf - this just says that your fav color is yellow
3. ask what your fav color is

I have OCR_APIKEY specified and I think it works. If I select in convo to upload document as text, and I paste the yellow.pdf file, I get a correct reply:

What browsers are you seeing the problem on?

No response

Relevant log output

N/A

Screenshots

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[danny-avila]

I did some testing here, I couldn't reproduce it on one instance vs. another, and I see the root issue now, will patch this up soon!

[danny-avila]

Working now after this PR is merged: #7814

[michnovka]

Thanks! confirmed