- More error capturing to avoid problems
- block user from submitting blank messages
- check if Dialogflow will filter/censor incoming messages
- stop matched intents with no answers from fucking everything up
- figure out what PaLM does if the current message is OK but history contains messages it wants to filter
- trap blank replies from PaLM before they come in like a wrecking ball
- finish writing todo list
- more fun easter eggs
- "I wish it was snowing!"
- "Something something celebration"
- simulate bot "typing" by outputting a couple characters every few milliseconds instead of whole message at once
- better icons for identifying who the message is front. maybe just the bunny ears for Louise?
- possibly seperate out context and examples for PaLM to another file for easier tweaking
- add more examples to PaLM
- figure out commands to actually tune model instead of only steering the base model's response
- unclusterfuck my google cloud console
- app engine
- cloud functions
- TTS & STT
- cloud storage
- analytics
- cloud build?
- find tool to make writing markdown less of a PITA
- resist urge to add speech-to-text