BlueEye Tracker

Android Bluetooth/BLE situational-awareness app focused on local-first collection, explainable evidence and conservative decisions.

• observes nearby BLE and Classic Bluetooth devices;
• keeps device, signal and session evidence locally;
• supports watchlist returns and cautious Follow-Me/identity analysis;
• reduces protocol noise before it reaches the main Radar view;
• keeps technical/raw Bluetooth data inspectable without making it the default UI;
• can export structured session evidence for deeper review.

BlueEye reports observations and evidence. It does not claim to identify a person, infer malicious intent from Bluetooth presence, turn RSSI into precise distance, or treat the phone's GPS observation point as the exact location of another device.

• Source of truth: main.
• Phase 3 scanner/ingest field reacceptance: closed / accepted.
• Compact Radar cards and lightweight Radar projection: accepted.
• Details progressive disclosure and collapsed Technical Details: implemented.
• Recent field work reduced false positives/protocol flooding and added the Live Nearby signal view.
• Active product work: field validation, Radar/Details quality and the deterministic analysis pipeline.
• Planned developer workstream: optional Drive + Gmail telemetry/AI feedback bridge.

Current plans:

• UI/UX and analysis plan
• Telemetry & AI feedback bridge plan

Bluetooth observations
  -> normalize / parse
  -> bounded ingest
  -> local persistence
  -> identity / classification / Follow-Me analysis
  -> evidence + alerts
  -> Radar / Details / history
  -> optional reduced external analysis

Modules:

app            entry point, navigation, Hilt wiring
core:model     shared domain data types
core:domain    contracts and use cases
core:data      Room, scanning, service, classification, sessions/alerts
core:decoders  Bluetooth decoders
core:ui        shared Compose design system
feature:*      Radar, Details, Settings, Watchlist UI/ViewModels

See Current Architecture and Detection Model.

Scanning, persistence, deterministic classification and baseline alerts must work without cloud/AI access. Optional external analysis receives bounded structured evidence and remains separate from the local verdict.

Requirements: Android SDK and JDK 21. Generated Android/JVM bytecode remains targeted at JVM 17.

./gradlew qualityCheck
./gradlew :app:assembleDebug
git diff --check

When available:

gitleaks git --config .gitleaks.toml --redact --verbose

The canonical networked build is GitHub Actions. Physical BLE/runtime behavior is verified separately on a real Android device.

• Latest tester release
• All releases

Tester APKs are engineering/debug artifacts unless a release explicitly says otherwise.

Start at docs/README.md. The repository intentionally keeps only current documentation plus one concise history file; detailed old phase audits/handoffs remain recoverable from Git history instead of living beside active instructions.

Key documents:

• Product Goal
• Current Architecture
• Detection Model
• Quality Gate
• Engineering Execution Flow
• Field Session Checklist

Read CONTRIBUTING.md before substantial changes. Never commit private field telemetry such as exact GPS, MAC addresses, Room/WAL/SHM databases, HCI snoops or bugreports.

Security reporting guidance: SECURITY.md.

No license file is currently present; reuse rights are therefore not explicitly granted.