pi-sense

Pi extension that gives text-only models media understanding — describes images, transcribes local audio, routes local videos through native understanding for what-questions or frames+local ASR for when-questions, then feeds text to the active model