.__        
  ______ ____  ___________|__|____   
 /  ___// ___\/  _ \_  __ \  \__  \  
 \___ \\  \__(  <_> )  | \/  |/ __ \_
/____  >\___  >____/|__|  |__(____  /
     \/     \/                    \/ 

Human-rock symbiosis

Setup

brew tap twilio/brew && brew install twilio
twilio plugins:install @twilio-labs/plugin-serverless
twilio login

Test

echo OPENAI_API_KEY=XXXXXXX >> .env
echo ACCOUNT_SID=XXXXXXX >> .env
echo AUTH_TOKEN=XXXXXXX >> .env
npm start
curl http://localhost:3000/1 # example command

Deploy

twilio serverless:deploy
twilio serverless:logs # run periodically to get logs

To-do

• add testing environment
• proper error handling
  • to all sync functions
• investigate latency with an upsert operation to prevent errors when creating maps with the same name (or work out better way to do testing)
• check whether async stuff is being used correctly
• prevent skipping (or make transition more natural) when caller talks over the bot
• add metrics for
  • total words detected

Low priority

• add parameter to specify gpt model per prompt
• separate content from code
• composite cody's sounds with voice?
• identify previous callers
• add caller name to notifications to distinguish them more easily

Done

• record from beginning of call rather than 1a
• use GPT3 for responses
• figure out a way to share state between functions or fold all functions together
• create a class to help deal with gpt3 interactions
• make shared utility function for picking random element from array
• record conversations
• remove all static pauses
• record all interactions
• add 'no response' conversation logic so bot doesn't plow ahead
• metrics: length of call

Reference

https://www.twilio.com/docs/runtime/functions-assets-api/quickstart