epic: Jan Agents and Modal switching while Inferencing
takeerdevs opened this issue · comments
Motivation
Designs
I don't have a design prepared, but I have an idea on how it can be drafted, hopefully someone else may dive in!. On left pane where there is Jan Logo, we have Threads icon, which can allow user to chat with LLM, below that we have Hub, but before Hub, there should be an Icon for Agents i.e chatting with agents. In settings, we can have a settings section for Agents, like Creating an Agent that will allow creation of either AssistantAgent or UserProxyAgent etc., each agent can be set to specific LLM, which can be very useful since LLM's perform differently in different areas.
Jan server should be able to switch models i.e turning off current model then start another model that has been requested by specific agent. Jan normal chat can be turned off while chat with agents is active.
What will this Unlock?
Jan has enabled chatting with LLM easily, but If AutoGen feature is integrated, Jan will be a powerful multi professional tool and can allow Jan to have access to internet with the help of AutoGen.
Use cases:
- Ability to chat with Jan that has access to internet
- Ability to create more than one agents, including an agent that can execute code, run code in docker containers and so providing accurate results, once errors found in code execution, agents can chat back and forth to fix errors and finally give a correct code.
Creation of Agents can be delegated to users so they can control the flow of agents communication, code execution etc.. including making sure that docker is installed in the host machine if code execution will be involved.
This is just an Idea, If anyone else has this vision in mind, you can contribute where I haven't covered.. including defining Tasklist, GUI, technical advice etc.
+1 for agents/swarm. Some of these repos might help: https://github.com/kyrolabs/awesome-agents