add discord bot "text to 3dstreet" using open ai API about discord-bots HOT 6 CLOSED

Also relevant, using function call to generate guaranteed JSON response

https://yonom.substack.com/p/native-json-output-from-gpt-4

from discord-bots.

Another option could be to host on "edge server" like google cloud functions or cloudflare using discord's web hook. Essentially to register the application hooks for discord, then the event can send a webhook to the cloud function which fires the actual logic to fetch a response from open ai api server. Here is example docs for discord webhook w/cloudflare worker:
https://discord.com/developers/docs/tutorials/hosting-on-cloudflare-workers

from discord-bots.

WIP https://github.com/Algorush/text-to-street/blob/main/README.md

from discord-bots.

hosting and access questions

1. how to host the bot and respond to inquiries: for now Discord seems like the easiest solution since there is no mechanism to add to app at the moment. suggest using the discord bot cloudflare example but adapt to firebase: https://discord.com/developers/docs/tutorials/hosting-on-cloudflare-workers

2. how to access the results - brainstorm of ideas

• create new street json file and add new entry in firestore database under a robot user -- might be easiest with existing infrastructure. then to provide the URL of scene to access

suggestions for ease of deployment:

• convert to node / javascript instead of python like existing firebase functions
• use existing config/.env and add openai and discord keys to that and/or gcp secret where .env is stored too

from discord-bots.

After thinking about how we will support this, here are the next steps:

for sep 28 live demo - keep in python?

• new repo or just fork into 3dstreet?
• make it easier to see output scene. ie downloadable or copy/pasteable (or even create file)
• debug existing performance - create a few example test cases with expected output
• consider prompt change / other functional improvements?
• Output: opens browser tab with new file? title is prompt from user?
• later v2 idea: basic ui to open input dialog in browser, then a processing icon, then show the scene with title

for deployment - convert to JS

• convert to javascript instead of python function
• hook up to discord api to register function per cloudflare example
• openai and discord api keys, add to existing development .env file and reference somehow, or use separate gcp secret that the firebase function can access?

from discord-bots.

follow more in this repo 3DStreet/text-to-street#1

from discord-bots.