Describe the bug
Even on local models than have the vision capability, like Qwen 3.5+ models served with Unsloth Sudio in example, the flows brokes anytime an SVG file enter the debate, instead of reading the file as text, Zoo code panic and send it as image.
To Reproduce
Steps to reproduce the behavior:
- Go to Zoo Code
- Ask it to read the content of a SVG
- See error :
Date/time: 2026-08-30T12:56:17.942Z
Extension version: 3.80.1
Provider: openai (proxy)
Model: unsloth/Qwen3.5-9B-MTP-GGUF
400
OpenAI completion error: 400 llama-server error: {"error":{"code":400,"message":"Failed to load image or audio file","type":"invalid_request_error"}}
Expected behavior
Zoo code should consider it as text, I tried to add Rules, but it does not work.
Screenshots
If applicable, add screenshots to help explain your problem.
Video
Add a video of the incident happening, it will help us debug
What version of zoo are you running
Version: 3.80.1 (4b85ff3)
Describe the bug
Even on local models than have the vision capability, like Qwen 3.5+ models served with Unsloth Sudio in example, the flows brokes anytime an SVG file enter the debate, instead of reading the file as text, Zoo code panic and send it as image.
To Reproduce
Steps to reproduce the behavior:
Expected behavior
Zoo code should consider it as text, I tried to add Rules, but it does not work.
Screenshots
If applicable, add screenshots to help explain your problem.
Video
Add a video of the incident happening, it will help us debug
What version of zoo are you running
Version: 3.80.1 (4b85ff3)