-
-
Notifications
You must be signed in to change notification settings - Fork 15
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
ChatGPT in Dotabod #472
Comments
I don't know much about how these bots work, is this something possible to do ? Especially protecting the API keys and system prompts are important. |
Hmm it is a great idea. Thanks for the write up. I'm worried though that there are other twitch bots which can already act as ai agents. So I'd be wasting some time writing this feature if it already exists. Do you know? |
I haven't seen any Twitch bots with AI Agent experience, yet. I've seen ChatGPT integrated bots, but they are not really AI Agents, offering just simple API calls for the language models, as far as I know. |
There is a new technology trend: AI Agents.
It is possible to integrate LLM models with external tools or websites using agent framework libraries.
Dota streamer can get his/her own OpenAI / Anthropic / Google API key for their chat language models and activate the !chat command.
For streamer's API key, !api openai <api_key>, !api anthropic <api_key>, !api google <api_key> kind of command, that is only usable by streamer, can be added.
The chat agent can be retricted to dotabuff and/or dota fandom wiki websites. Permissions either can be defined by you, or if you want more comprehensive experience, !permission command can be added, only usable by streamer.
!permission website http://www.some_web_site.com => AI Agent will use that website to generate its responses.
Example usage:
!chat what is the magnus' innate ability?
dotabod: (reads the related wiki page and explains it in twitch chat)
API key and usage rates / pricing is related with the streamer, but you can implement the agentic backend.
One of the popular frameworks for such agentic use case can be Langchain.js.
The system prompt could be modifiable by the streamer but you can define some basic prompts / instructions.
For that custom purpose, !system command only usable by streamer can be added as well.
Overall:
!chat <question, prompt> -> Twitch user asks something to the AI language model
!permission -> AI agent can use that as a source to answer questions or respond to prompts. Only available to Streamer or moderators.
!system -> System prompts that define how language model should behave. Only available to Streamer.
!api <api_key_provider> <api_key> -> Access to chat model using their API. Only available to Streamer.
The text was updated successfully, but these errors were encountered: