Race Condition is an open-source, deployable reference architecture built by Google Cloud developer advocates that demonstrates orchestrating, scaling, and securing autonomous AI agents using Gemini Enterprise Agent Platform. The project simulates a marathon through Las Vegas with 1,000 concurrent runner agents, showcasing key patterns: agent-to-agent (A2A) protocol for dynamic agent discovery via an agent registry instead of hard-coded REST APIs, agent-to-user interface (A2UI) protocol for AI-generated structured UI components rendered in Angular, and a polyglot backend using Go for high-concurrency WebSocket messaging and Python for the AI ecosystem. The architecture incorporates dead reckoning from game development for state synchronization, a fan-out pattern for high-volume event processing, long-term memory via Memory Bank and LODB for vector search, and a zero-trust security model with per-agent IAM identities enforced at the Agent Gateway. The full codebase is open source on GitHub and can be deployed to Google Cloud with free credits.
Sort: