Some features I've added:
- texting multiple people - I'm the default but she can write "to: dada" or "to: phonenumber" and it'll text the recipient. I made a simple UI for editing the phone number/device id to name mapping.
- A bit of extra paper is printed with "to: sender" at the top to make it easy to chat back and forth.
- GPT support - she can write "to: GPT" and it'll call GPT API and then print out the response. I prompted GPT to make it easy to read for an elementary school child so she can read it herself.
- GPT support for me - I can text to: GPT and it'll text me back rather than printing a message. This has been great for my space-obsessed 3 year old.
- texting other devices - I have another device set up and they can text message each other.
- read outloud - there's a second button that reads whatever text is under the scanner. This is helpful for my 3 year old who can't read.
GPT has been fantastic all around. It's way better at OCR-ing kid handwriting than Tesseract or Google OCR and being able to ask questions and get answers at an age (and reading) appropriate level has been great.
RichardHesketh•8mo ago
alibova•8mo ago
For sending text messages I have a raspberry pi hooked up to an arcade button and an off the shelf document scanner (you can see it in the last pic). When kiddo pushes the button it takes a picture, uploads to s3 and then sends via Twilio API.
For receiving the text message I have a Lambda script that receives the text message (via Twilio webhook) and it publishes it to a Pubnub channel, the raspberry pi monitors the pubnub channel and prints via thermal printer when a new message is received.