On-device AI · Vision intake

Now the counter
can see.

Show SudsCo a photo of your laundry pile and the desk reads it — garments, kilos, special items — returns structured JSON, and quotes the price. On supported Chrome the vision model runs on your device: the photo never leaves the tab. That's a pre-consult intake no cloud chatbot can promise.

VISION DETECTING… OUTPUT STRUCTURED JSON PHOTO LEAVES DEVICE NEVER ANALYSES 0
PHOTO INTAKE · PICK OR UPLOAD
Your uploaded photo (stays on this device)
▸ the responseConstraint schema (the automation contract) 

      

    


    

      
VISION → JSON → QUOTE —

      
// pick a sample photo (or upload on a vision-capable
// Chrome) and the analysis appears here.

      

        

        

        
        

      

    

  


  
Honest label: on Chrome with built-in AI (148+), analysis runs via LanguageModel with expectedInputs:[{type:"image"}] and a responseConstraint JSON schema — genuinely on-device, upload enabled, badge flips LIVE. Elsewhere the three sample photos replay pre-computed analyses (labeled SIMULATED) so everyone sees the flow. The quote math is real either way — same price engine as our other SudsCo drops.


  

    
Why "structured" is the whole product

    
A chatbot that chats is a toy; a desk that turns a photo into clean, schema-valid JSON is a system you can wire into bookings, CRMs and pricing engines with zero human retyping. Same pattern works for clinics (pre-consult intake), trades (photo-quote a job), and retail (what-is-this-product) — always with the photo staying on the customer's device.

    Give my counter eyes →