SentinelVega

Autonomous cross-surface physical security orchestrator bridging Ring telemetry, Alexa+ Agentic Reasoning, and Fire TV 10-foot Ambient HUDs.

Comment

InspirationSmart home and enterprise security are paralyzed by modal fragmentation. Ring captures high-fidelity edge events but relies on easily ignored, stateless push notifications. Alexa+ offers profound agentic reasoning via the Model Context Protocol (MCP) but lacks visual presence. Fire TV owns the largest screen in the building but remains blind to the physical perimeter. Our inspiration was simple: exploit the Amazon ecosystem's untapped synergy. We architected SentinelVega to unify these three isolated surfaces, transforming passive alerts into an active, multi-modal, closed-loop ambient security orchestrator.What it doesSentinelVega operates as an autonomous physical security orchestrator. It ingests high-frequency perimeter telemetry from Ring cameras into a sub-millisecond, bounded $O(1)$ sliding window buffer. When anomalous activity triggers (e.g., high-confidence person detection at the gate), an Alexa+ Agent—powered by AWS Bedrock and the Strands SDK—evaluates the threat profile via a mathematically proven MCP 2025-11-25 Streamable HTTP server.Based on deterministic threat classifications, SentinelVega instantly injects an interactive, 10-foot Picture-in-Picture (PiP) Command HUD onto any Fire OS or Vega OS television. This allows users to acknowledge threats, trigger an Alexa+ voice deterrent, or execute a physical lockdown using pure D-pad spatial focus.How we built itWe engineered this pipeline with uncompromising rigor, condensing the entire cross-surface orchestrator into exactly 12 production-grade files.Alexa+ Core (Python): We built a deterministic ASGI gateway fully implementing the 2025-11-25 MCP Streamable HTTP specification, managing cryptographically secure Mcp-Session-Id continuity and JSON-RPC 2.0 tool execution in under 5 milliseconds.AWS Builder Mini Challenge: We integrated the AWS Strands SDK with Amazon Bedrock (Claude 3.7 Sonnet) to synthesize raw Ring event bursts into strict JSON threat models.Spec-Driven Hardening: The architecture is verified against property-based invariants defined in .kiro/specs.yaml using Kiro Crew, mathematically guaranteeing memory-bounded telemetry and zero-latency execution.Fire TV / Vega OS Client: We deployed a resilient WebSocket client with bounded exponential backoff driving a React Native TVOS UI, adhering strictly to Amazon's D-pad spatial focus guidelines.Challenges we ran intoPioneering the newest specifications required navigating undocumented edge cases. We meticulously documented six deep-technical issues in our Friction Logs (see README for the +10% judging criteria), isolating CORS header stripping in the MCP Streamable HTTP spec, D-pad focus state desynchronization during asynchronous WebSocket updates on React Native TVOS, and JSON-schema strictness inside the Bedrock Converse API.Instead of merely reporting these, we engineered production workarounds for all of them—such as writing zero-dependency ASGI middleware to guarantee HTTP state continuity and deterministic spatial controllers for the Fire TV UI.Accomplishments that we're proud ofExtreme Single Responsibility: Every class across the Python and TypeScript codebase strictly enforces a maximum of two functions/methods, ensuring architectural predictability, testability, and zero boilerplate bloat.True MCP 2025-11-25 Compliance: We didn't wrap a legacy API; we built a secure Streamable HTTP transport layer capable of true stateful multi-turn execution.Open Source Mini Challenge Execution: We extracted our ASGI middleware solution for the Mcp-Session-Id CORS stripping issue and submitted it as a clean Pull Request directly to AmazonAppDev/devices-agent-skills, clearing an infrastructure bottleneck for the wider Amazon developer ecosystem.What we learnedThe true leverage of Alexa+ isn't just conversational UI—it is utilizing agentic reasoning as the invisible, intelligent connective tissue between edge physical sensors (Ring) and ambient digital displays (Fire TV). On the client side, we mastered the specific mechanics of managing React Native TVOS spatial focus under asynchronous, high-frequency state mutations.What's next for SentinelVegaSentinelVega is not a hackathon prototype; it is a venture-ready baseline. The immediate roadmap involves adding localized computer-vision bounding box overlays directly to the Fire TV HUD, finalizing our open-source mcp-streamable-http-2025 core transport library, and launching the system as a high-leverage B2B SaaS solution targeting commercial warehousing and enterprise perimeter security.

Built With

Share this project:

Updates

Submission history