Ensemble

Inspiration:

• Revolutionize the AI industry by making a multiple-agent systems where agents collaborate to a final product.

How the application works:

• The application segments complex problems into small tasks and assigns them to specialized agents that collaborate to generate a final product, extracting better results from each agent and also lowering cost of production.

How application was built:

• Receive goal from user, then fragment it into multiple small tasks, and assign them to different specialization of agents. These agents then will complete their individual tasks and report to the next level of operation in the hierarchy, if task is not correct agent will be called upon again to fix errors until task passes unit testing, then the framework can proceed to the next task. When the final goal is reached the user should have all application files and the correct solution.

Challenges throughout Hackathon:

• Some challenges we ran into are locally implementing a sandbox in python to support unit testing to ensure correctness of solutions, and also working with AWS and Docker remote services.

What we are proud of:

• We're proud of getting agents to work collaboratively and pick correct tools to arrive at final solutions, making a product that can generate full-stack applications in minutes, and also to do that with only 24 hours.

What have we learned:

• We have learned extensively about Claude artifacts, multi-agents, hierarchal systems, concurrency, AWS, Docker, and several python libraries, such as subprocesses and logging.

What's next for Ensemble?

• Implementing support for more programming languages and other services such as righting word documents, slides, and excel sheets, and ultimately having multiple agents running at the same time waiting to be called upon on a thread and running tasks asynchronously to maximize optimization.

Built With

• amazon-web-services
• docker
• json
• openai
• python