enricoros / big-AGI

Generative AI suite powered by state-of-the-art models and providing advanced AI/AGI functions. It features AI personas, AGI functions, multi-model chats, text-to-image, voice, response streaming, code highlighting and execution, PDF import, presets for developers, much more. Deploy on-prem or in the cloud.

Home Page:https://big-agi.com

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

BEAM - feature thread

enricoros opened this issue · comments

Big-AGI 1.15.0 introduces BEAM, a chat mode that lets you find a better answer. Beam is a user-centered multi-model chat that allows the user to select the best responses, and fuse them together in different ways.

Released on March 28, 2024 to the first round of Early Adopters.

Released publicly on April 1, 2024.


Former description from the "Best-of N" naming:
In this mode, the assistant will generate N (2..10) completions, which will be shown to the user on a modal popup dialog. When all of them are done, they become active and scrollable. At this point either the user can select one of them to append to the conversation that started it all, or can use a "Merge"function specifying the criteria.

  • initiate from the "Retry" or "Restart from here" menu options
  • also initiate as a special "Mode" for the Chat
  • Beams (aka alternate responses) are generated from: Chat up to that point, submitted N times in parallel
  • Merge happens by applying: [merge system, prompt, optional initial system prompt, history, last user message, N alternatives, merge user message]

While this is an interesting feature, why is the chat configured on Beam mode by default?
I use the keyboard a lot, and having everytime to click on chat with the mouse is a bit annoying.. it would be nice if users could have the flexibility to setup their own preference. Personally, I'd use Beam very rarely, so I would prefer to have the Chat button set up by default when I hit "Enter"