# StateShift

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Lohitha kayyuru](https://cerebralvalley.ai/u/lohithak)
- **GitHub:** https://github.com/LohithaK25/StateShift
- **Demo video:** https://youtu.be/wQokE-XzC68
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/814

StateShift is an AI-powered network troubleshooting agent built with Meta Muse Spark.

Network failures can be difficult to diagnose because identifying the faulty component is only part of the problem. Engineers must understand the current infrastructure state, choose an appropriate recovery action, and verify that connectivity has actually been restored.

StateShift demonstrates how an AI agent can perform this workflow autonomously in a deterministic network simulator.

The agent follows an Observe -> Diagnose -> Decide -> Repair -> Verify workflow using four structured tools:

inspect_network - inspect network nodes and links.

test_connectivity - check whether a service is reachable.

repair_link - recover a failed network connection.

restart_node - recover an unavailable network node.

The project demonstrates three scenarios:

Broken Link: Identifies and repairs a failed network link, then verifies connectivity.

Failed Router: Detects an unavailable router, restarts it, and independently verifies recovery.

Redundant Path: Recognizes that an alternative route already provides connectivity and avoids unnecessary repairs.

StateShift is designed with constrained tool execution, inspection-first decisions, bounded recovery actions, and independent verification. Its frontend streams the agent's actual tool calls and network state transitions, allowing users to observe what the agent investigates, decides, and changes.

Built with Meta Muse Spark, TypeScript, React, Node.js, and Express, StateShift explores a practical approach to autonomous infrastructure operations.

The central idea: Don't just ask AI what went wrong. Let it investigate, act, and prove the recovery.

## More from Meta Global AI Developer Hackathon

- [Origo Labs](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/811)
- [AdiSaan - TBWIM](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/812)
- [Dealmatch](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/813)
- [Agent Andy](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/815)
- [Tcpcore- Cptlight](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/816)
- [AgentGuardRequestGuard](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/817)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/814. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt