🤝 Top Collaborators
.png&w=640&q=75)
.png&w=640&q=75)
.png&w=640&q=75)
🤓 Latest Submissions
EventOS
Live events rarely fail because of the plan itself. They fail in the handoffs between vendors, staff, catering, registration, schedules, resources, and communications. When something changes, event teams must manually detect the issue, investigate its impact, coordinate a response, communicate with external parties, update operational systems, and follow up. EventOS transforms this reactive process into an agent-driven operational loop. Its runtime Event Manager Agent, EVA, continuously works with event state to detect incidents, investigate relevant operational data, assess risks, create recovery plans, execute authorized internal actions, verify outcomes, and maintain an audit trail. EventOS is designed around controlled autonomy. EVA can perform low-risk internal actions such as creating tasks, assigning resources, updating operational information, and notifying internal teams. Protected actions remain under human authority. Our primary demonstration is a catering vendor delay. Catering is scheduled for 13:15, but the vendor ETA changes to 13:30. EVA detects the conflict, investigates the timeline, vendor information, catering coverage, assigned staff, and affected activities, then creates a recovery package. Because contacting an external vendor crosses an authorization boundary, EVA generates a grounded email draft and places the incident in WAITING_FOR_APPROVAL. The organizer can inspect the exact recipient, subject, message and reason before selecting Approve & Send. Once approved, EventOS executes the communication through Google OAuth 2.0 and the Gmail API. The resulting communication status is recorded alongside the incident lifecycle and EVA activity, creating a traceable path from detection to real-world action. EventOS also includes Firestore synchronization, communications tracking, operational audit history, incident lifecycle management, and deterministic safety fallback behavior when the configured AI service is unavailable. The prototype demonstrates a model for autonomous operations where AI handles operational complexity while humans retain authority over consequential actions.
27 Sep 2026
