jmaynor / dolores Goto Github PK

Pass necessary config variables through using the class constructors.

Make use of SMMRY API
https://smmry.com/api

Takes up unnecessary space in program as is. Would make more sense to store that outside of program itself. And don't want it in the main config file.

LangChain seems like a good tool for this. Can make the calls generic and it will just depend on what is specified in file whether we go to local LLM or commercial one.

Make use of OpenAI since already using it for text gen. Then add self-hosted options. Such as Stable Diffusion or similar.

Store the last couple messages to/from the bot so can have limited memory

Go through and convert the various instances of discord.py so they use pycord. It's kept up-to-date. Also make it so that each command is a "bridge command" in order to make use of Discord's slash command feature.

Build out so that there are generic commands for getting or writing schedule data for both Twitch and Notion

Use the command groups feature to clean up main python file

Currently downloading audio file rather than streaming to prevent weird distortion in beginning of stream. As well as seems to be more consistent. However, not currently cleaning up downloaded files.

Make sure of the copy all (copy .). Will simplify if ever adding more files in the future.

Respond with error messages when a command is used and required info wasn't provided in the config YML file. Rather than silently erroring out and printing to stderr.

Add additional documentation to readme. Go over what each module does and add example config file.

Break down into 3 functions:

• Check for someone mentioning Dolores
• Check for command mistype
• Check for news article auto-summary

Otherwise do normal process_commands

Use proper logging to track command use, errors, etc.

See if possible to use an instance of llama (or some other LLM) as opposed to the chatterbot stuff. Would remove the need to include out-of-date packages in the repo. Would also be more interesting.

• Add option to have playing audio repeat
• Potentially add small GUI, could search for something to play

Useful functionality, too often generates summary that doesn't have anything to do with actual article text. Seems to often be because URL is redirecting user elsewhere, or most often because news websites have tons of non-relevant info on page. (Other articles, breaking headlines and such, random garbage)

Example URL: https://www.cnn.com/2023/11/02/tech/microsoft-ai-news/index.html

Redo in some way. Pull article text first and then summarize. Or investigate other APIs.

Set up continuous deployment so new releases will automatically push to Docker Hub