RoundTable

• 

  Thumbnail

• 

  Example of the program

• 

  Jupyter Notebook sample prompt and responce

• 

  Postman testing

Inspiration

When single AI tools tackle complex problems, they may produce biased or incomplete solutions. This let us to develop RoundTable. Inspired by the dynamics of human teams collaborating on challenging tasks, RoundTable introduces the idea of assigning specific roles to various AI members, similar to how humans divide tasks based on individual strengths and expertise. By leveraging the unique capabilities and combining the strengths of different AI models, RoundTable aims to provide more balanced and comprehensive outcomes.

What it does

Prompted RoundTable is a highly customizable AI tool that supports interactions between multiple different AI models. AIs will work together on tasks, generating solutions with better quality in terms of efficiency and accuracy to problems than single AI-supported tools. Users can edit their own "problem-solving team" by assigning specific roles like "researcher" or "programmer" to an AI team member, allowing them to focus on designated tasks, releasing the full potential of AI.

How we built it

Just like the concept of RoundTable, we divided tasks to each team member based on our individual strength. Separately we finished the designing and coding to our frontend, AI power and logic, and databases. With the help of autogen api, we were able to accomplish the conversations between AI team members with different logical roles. Mongodb provided the server for our database and the API we need to access it. RoundTable is the product of every team member’s hardworking throughout the event period.

Challenges we ran into

Autogen is a very new resource to our team, we spent a long time in the first half of the event to learn and understand it. The most challenging part is the transportation of messages between backend and frontend. We were having trouble displaying the chat messages after it’s generated after the backend receives user input. We eventually fixed it after testing different parts of the code using postman tests.

Accomplishments that we're proud of

We successfully built an interacting tool between the user and multiple AIs’. Greatly increased the accuracy of AI solution out puts.

What we learned

Relatively inexperienced team members learned about database building, data communication between frontends and backends and api testing. Every team member learned about the autogen API.

What's next for RoundTable

In the future, we are expecting more AI models to be supported. Users will be able to customize not only the roles of AI models but also the number of AI models in the team. We will also feature API learning in AI models. AIs’ will be able to call the correct API after given the key, this will be an important update as it strongly enhance the programming efficiency of RoundTable

Built With

• bootstrap
• fastapi
• mongodb
• python
• react
• tailwind