TODO.txt

Essential for Final Project scope:
================================
√



Not essential for Final Project scope:
================================
Quick: Enforce question/response on the frontend (so no double questions)
Quick: "Alan has joined the room" messages
Quick: "Waiting to connect" messages
Quick: No swears/racism? 
    - Dilemma here, because if that's truly something that shows you're human
        then maybe it should be in there.
    - Could compromise and bleep them out?
Quick: Disconnect functionality

Propogate good responses across similar questions

Time delay before answering as a feature in the vector?

Move into smarter vectorization for conversations (if possible)
    - Questions are not independent, conversations flow.
    - Might violate blockhead property?
    - Neural Net?
    - Sentence n-grams?

Deal with "asdjlkfha;as;dfa;kljsdfhaksd"
    - Right now Alan doesn't think it's similar to "asdfk;jah;dfkal;sdfasdf"

Identify and ignore trouble/adversarial users.

Better robo inquisitor.
    - Start with frequent questions (Cold questions)
    - Then move into ones we've failed (Heating)
    - "Heating" will have to either be controller-side 
        or based on transcript length to keep API stateless

Clean up system - multi-server w/ dns-level apps for api etc.

Https?

Load balancer?

AWS DB?

Group Chat? This should at least be mentioned in the whitepaper.
    - Interesting to see how an AI could take part in a group conversation
    - Punish if singled out, like "/machine Alan"

Hook him up to messaging platforms
    - Facebook Messenger
    - iMessage
    - SMS