WhatsApp Inbound Lab
The Dircom forwards a WhatsApp message they're worried about — a boycott call, an angry screenshot, a critical article link — to Harch's dedicated number. Harch's NLP pipeline analyzes it and returns a risk assessment. This closes the "user-feels-they-put-the-data-in" loop: WhatsApp groups are private, no monitoring tool can reach them, but the Dircom sees the signal at 11pm and can forward it to Harch before the CEO sees it at breakfast.
This simulates what happens when a Dircom forwards a WhatsApp message to Harch's dedicated number. In production, Twilio sends the webhook automatically (POST /api/whatsapp/inbound). Here, the /api/whatsapp/simulate endpoint runs the same NLP pipeline — sentiment + sarcasm + prompt-injection + fakeness + Darija language detection — without needing real Twilio credentials. The verdict appears in the feed on the right within seconds.
Simulate an inbound WhatsApp
Compose a message as if you were the Dircom forwarding something you saw on your phone. Click "Send to webhook" to fire it through the same pipeline as a real Twilio webhook.
If set, the message is treated as an image with the body above as its caption. VLM analysis is queued (manual review pending — vision model integration is on the roadmap).
Inbound feed
No inbound messages yet. Send one from the form on the left, or refresh to load the sample feed.