Main Functionalities

[MPieke]

• Builds on AutoGen, but can also create python scripts based on multi-agent conversations. Similar to AutoGen, these scripts can be iteratively improved.

• Once the script has been cleared by the head agent, the script is created. In later iterations, this script can be imported by new scripts in development.

• Main value: creating a modular, multi-script application using autonomous agents. Going from idea to actual app

[MPieke]

Ideally, both front and backend integration should also be automatically testable. But the true user-experience cannot be tested by an LLM agent, only whether it works or not.

[MPieke]

Would also be nice if the agents can work in separate branches to test out certain things. E.g. one branch is fully autonomous, the other has human in the loop etc. Merging branches should happen manually for now, but would be nice if that can happen automatically in the future.

[MPieke]

Possibly create a community of LLM agents that pose as consumers to trial run the app