Auto-transcribe + smart waiting for fragmented messages

Two features work together to make ORYKSA AI feel natural in real WhatsApp conversations.

Auto-Transcribe Audio

When a client sends a voice note, the AI:

1. Downloads the audio.
2. Transcribes it (text shown back to the client with a footer).
3. Generates a written reply based on the transcript.

Toggle the feature on the Dashboard. Choose Received (transcribe what clients send), Sent, or Both.

Smart waiting (debounce)

If a client types in fragments, for example: Hi · I want to know · about availability · tomorrow, the AI waits ~6 seconds after the last message before replying. All fragments are joined into one context so the AI gives a single coherent reply, instead of replying separately to each fragment and re-introducing itself.

This works automatically, there's no toggle.