Exploring the Discord API with discord.py and the OpenAI API.
This is mainly used for toying around with a bot user in a single server. I haven't tried anything with multi-server configs yet, but I don't imagine they'd work well for the time being.
- Create a Discord bot app here: https://discord.com/developers/applications/
- Create (or use an existing) OpenAI API key
- Configure
.envlocally (example) - Adjust
initial_prompt.mdas needed (example) poetry run python ./src/app.py
- switch from ChatCompletion to the Assistants API; each server in its own thread with
channel:usernameas the messagenamevalues- store
channel-username: threadidmappings locally; if no thread ID exists, create thread and carry over last (up to) 10 messages in history
- store
- summarize user-attached images via https://platform.openai.com/docs/guides/vision
- add function calling for:
- image generation with https://platform.openai.com/docs/guides/images/usage?context=node
- adding reactions to messages, either with unicode emojis or with server-specific reactions
- making external API requests / web browsing
- handle multi-server bots and settings
- Docker build with helm deployment
- Feedback loop based on 👍👎 reactions to bot messages
- Persistent storage
- Web UI for monitoring context window and some basic overrides/commands
- explore usage with local models capable of using similar function calling methods
- voice channel support
- load user audio https://platform.openai.com/docs/guides/speech-to-text
- emit bot audio https://platform.openai.com/docs/guides/text-to-speech