StarFlow

• 

  Homepage: Meet StarFlow: one shared API, multiple AI agents, and fair access, brought to life with an animated space theme.

• 

  Agent selection: Choose 3–4 AI agents from 20 services to build your team and share Gemini API capacity fairly.

• 

  Dashboard: Watch live requests, queues, and traffic as StarFlow gives each agent a fair turn to use Gemini.

Inspiration StarFlow was inspired by an acequia, a traditional olden day community irrigation system that shares limited water between farms. Each farm gets an agreed share, and unused water can flow to farms that need more. We saw a similar problem with AI tools. Several agents can share one API, but one busy agent can use up the available capacity. We built StarFlow to bring the acequia’s idea of fair sharing to AI.

How does it work StarFlow is a traffic controller for AI agents sharing Google’s Gemini API. It decides which requests can go through each second, gives every agent a fair share, and puts extra requests in a queue. If one agent is quiet, others can use its unused capacity. This keeps one busy agent from taking over.

What we built StarFlow helps several AI agents share Google’s Gemini API fairly. Each agent has its own request queue and an assigned share. When an agent does not need its full share, the remaining capacity goes to other agents with waiting requests. Users can choose their agents, adjust their shares, and watch requests move through a space-themed dashboard. Gemini answers questions, summarizes text, and labels text by category.

How we built it We built the frontend with React and the backend with Python and FastAPI. The backend scheduler controls when requests reach Gemini, while a WebSocket sends live updates to the dashboard.

What we learned We learned that building useful AI tools also means managing how they share resources. We learned about request queues, fair-share scheduling, live updates, and connecting a frontend to a working backend.We also learned to balance visual detail with speed, so the animations support the experience while the numbers stay readable.

Challenges we faced One challenge was keeping a busy agent from taking over the shared capacity while still making good use of any unused shares. Another was keeping the dashboard connected as the backend changed. We had to update the data handling, use secure connections, and show connection problems clearly. The visual work also took care: all agents needed properly connected paths, smooth animations, and clear labels. StarFlow brings an old community idea into a modern setting: share a limited resource fairly, and let unused capacity help someone else.

Built With

• css
• fastapi
• geminiapi
• github
• html
• javascript
• postgresql
• python
• railway
• react
• typescript