How to Build a Quokkabot
summary
Builds a working quokka-identifying chatbot that accepts images via Twilio WhatsApp, MMS and SendGrid APIs, runs them through Microsoft Custom Vision API (prediction only, currently iteration 13), and returns a percent probability plus an image. Uses Azure Functions to parse WhatsApp URLs or email buffers (classifyImageURL vs classifyImageBuffer), randomly serves one of ~20 quokka photos, and feeds results to a live results page via Sync API. Prototype runs live but can sleep and needs scaling.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 2:15Quokkabot project
- 5:49WhatsApp bot with Twilio
- 13:36Image responses for quokkas
- 17:57Custom Vision integration
- 27:11Email and live results
- 35:36Quokka facts and bug reporting
- 41:01Technical details and costs
- 45:35Q&A
related talks
01h 01mBuilding fun and creative messaging experiences on WhatsApp
28mScrew the Browser! Build ChatGPT via SMS
01h 01mBuilding fun and creative messaging experiences on WhatsApp
54mChatbots: The hype, challenges and opportunities
35mChatbots: The hype, challenges and opportunities
41mHow We Built A Bot For
47mEnd-to-End Chatbot Web App with Your Own Data
27mA primer to write a conversational chatbot
