main ← fix-respect-persona-s-vision-setting-in-llm-triage-automation
opened 09:29AM - 26 Dec 25 UTC
The triage automation was unconditionally including image data in LLM requests w…henever a post had uploads attached. This happened regardless of whether the persona had vision enabled, leading to unexpected token costs for users who only wanted text analysis.
The fix adds a check for `triage_persona.vision_enabled` before including uploads in the prompt, matching how the Playground already handles this.
Ref - https://meta.discourse.org/t/392143
