I created an experimental open-source application which allows conversing to multiple AI assistants in a group chat-like environment.
You can find this project on my GitHub Repo: https://github.com/andreicozma1/GroupChat-GPT/
A more comprehensive list of features and screenshots are available on the repo.
This project is code-only for now, which you can easily start up your own local web server using the instructions in the repo.
If you’d like to contribute, feel free to create a pull request and I’ll also add you to the contributors list.
Features & Screenshots
(check GitHub repo for the most up to date list of features & screenshots)
- Powered by the OpenAI GPT-3 API
- Current members:
@davinci,@dalle, and@codex - Assistants are given their own distinct personalities, behaviors, and capabilities through dynamically-created generation prompts.
Response Coordinator
@coordinatoris a special assistant who dictates who should respond to the user's message by tagging other assistants- Tip: Directly tag an assistant (ex:
@davinci) to only have them respond (this skips the coordinator)
Generate Images
@gen_imageis an assistant helper that directly uses the OpenAI DALL·E 2 API, which can generate images from text prompts.@dalle(the chatting assistant) can help you generate these text prompts to generate images as you'd like.- Tip: you can also directly prompt the OpenAI DALL·E 2 API (
<gen_image>A picture of a cat.<gen_image>)
Contexts, Follow-ups, and Ignored Messages
- Hover over a message to visualize the messages that were used to provide context
- Coordinator messages are ignored by default - they won't be used as generation context for other assistants because that would waste tokens
- You can manually ignore messages by clicking the 'eye' icon at the bottom-right of a message
Markdown Formatting
- The UI for the chat message bubbles supports Markdown formatting
