Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Omen Vault: AI Omen Oracle

Inspiration

Traditional mobile vaults and secure note managers rely entirely on passive encryption: data sits idle in storage while brute-force or key-tampering attacks happen in the background. We built Omen Vault with the AI Omen Oracle to transform personal storage from a passive target into an active, autonomous cyber-defense system that evaluates threat telemetry and takes immediate defensive action.

What it does

  • Autonomous Threat Evaluation: Analyzes real-time telemetry anomalies (e.g., rapid unauthorized decryption bursts, cryptographic signature tampering).
  • Multi-Turn Function Calling: Powered by Gemini 2.0 Flash, the agent autonomously executes specialized defensive tools:
    • inspectVaultIntegrity: Verifies cryptographic checksums and calculates risk scores.
    • isolateCompromisedNode: Quarantines compromised keys, isolates the node, and revokes active sessions.
    • logSecurityEvent: Commits immutable incident records to the audit ledger.
  • Biometric Edge Gate: Local hardware biometric authentication gating all vault entries.
  • Isolated Cloud Persistence: User-isolated Firebase Cloud Storage rules ensuring zero cross-tenant access.

How we built it

  • Framework: Flutter & Dart for cross-platform edge performance.
  • AI & Agent Architecture: Google Generative AI SDK (google_generative_ai) integrated with gemini-2.0-flash for high-throughput, low-latency autonomous tool calling.
  • Cloud & Auth: Firebase Authentication, Cloud Firestore, and Cloud Storage.
  • Monetization: RevenueCat Firebase Extension for subscription entitlement gating.
  • Mobile-First Pipeline: Built, tested, and committed 100% on mobile using GitHub Codespaces.

Challenges we ran into

  • Implementing deterministic multi-turn tool calling on mobile without blocking the UI thread.
  • Balancing edge-first local encryption with cloud-backed anomaly stream processing.

Accomplishments that we're proud of

  • Creating a fully working autonomous cybersecurity agent loop using Gemini 2.0 Flash function calling directly inside a Flutter client.
  • Executing the entire development and deployment pipeline entirely on mobile.

What we learned

  • How to structure deterministic function declarations and tool schemas with the Gemini 2.0 API.
  • How agentic loops elevate standard apps into proactive autonomous assistants.

What's next for Omen Vault

  • Distributed multi-node peer verification across local networks.
  • Deploying Omen Vault as a foundational cybersecurity learning tool for youth technology initiatives at the Bennett World Youth Center in Panama.

Built With

  • firestore
  • gemini
  • gemini-2.0-flash
  • google-adk
  • google-cloud-run
  • python
  • vertex-ai
Share this project:

Updates

posted an update

Architecture & AI Stack Clarification

Updated model runtime specifications:

  • Orchestration & Reasoning: Gemini 2.5 Pro
  • Low-Latency Dispatch & Telemetry Routing: Gemini 2.5 Flash
  • On-Device Edge Guardrails: Gemma 3
  • Client Framework: Flutter
  • State & Persistence: Firebase / Google Cloud Run

Log in or sign up for Devpost to join the conversation.

Submission history