Admiral v00044 - v00046: Mesh Fallbacks, Bandwidth Prioritization, and AirDetect
Launching AirDetect spatial analytics, KCP/FEC mesh resiliency, and adaptive bandwidth prioritization.

Following our work hardening storage and ARM64 capabilities inside v00041 through v00043, Admiral v00044 through v00046 introduces our native spatial analytics engine alongside major upgrades to edge resilience over degraded, high-latency network links.
Here is what is new and improved.
1. AirDetect: Spatial Analytics & BLE Refinements
Previewed in our v00041-42 cycle, the AirDetect spatial analytics engine is now live, introducing native Bluetooth dwell calculations and real-time environmental beacon telemetry directly into the core Admiral pipeline.
- AirDetect Engine: Tuned core algorithms for high-accuracy spatial interaction metrics.
- Streamlined BLE Architecture: Converted BLE spec displays to structured objects and added escaping for local Wi-Fi names over BLE to prevent payload corruption.
- Hardware Exposure: Wi-Fi SSIDs and active BLE node counts are now exposed directly via
sysfsfor streamlined hardware reads.
2. Unbreakable Mesh & Bandwidth Prioritization
Building on the SD-WAN and latency optimizations introduced in v00039, operating over metered satellite or overloaded cellular links can stall standard streams. We have rebuilt our connectivity stack to survive the harshest environments:
- KCP + FEC: Moved remote access off WebRTC and onto KCP with Forward Error Correction (FEC) to combat packet loss.
- Adaptive Bandwidth & Mesh Fallbacks: Connectivity is now broken down by priority level. Telemetry, logs, and high-fidelity screenshots are automatically deprioritized on low-bandwidth links, while the Admiral Mesh KCP fallback has been fully revised for extreme stability.
- SD-WAN Resumption: Resolved a critical bug to ensure seamless reconnection and tunnel recovery after a WAN interface drop.
- Double-Safe Recovery Safeguards:
- Bumped the absolute update timeout to 24 hours to accommodate slow-drip connections.
- Added a hard 7-day non-communication watchdog that automatically reboots the unit if it loses contact with the registry backend entirely (complementing our existing 6-hour local IP link rollback boot check from v00039).
3. Fleet Management & Lifecycle APIs
- Dynamic Fleet Migration: Devices can now be programmatically migrated between fleets on the fly via the Device Identity API, expanding on our deterministic naming schemas.
- Unified Hardware Resets: Reorganized the physical GPIO reset flows (leveraging the low-level timing fixes from v00043) to integrate directly with the Identity and Lifecycle controllers.
- Automated Update Windows: Configure predictable local time windows during which system-level updates are permitted to install.
4. Stability, UI, & Micro-Fixes
- Capped Telemetry Buffers: Added strict on-disk size limits for metrics, logs, events, and BLE telemetry to prevent disk exhaustion.
- Atomic Topology Writes: Network and local device topologies now flush atomically to avoid corrupt state files on sudden power loss, matching the transactional configuration commits introduced in v00043.
- Zero-Byte Manifest Fix: Resolved a crash that occurred when parsing an empty system update manifest.
- TUI Image Validation: Fixed a startup hang in the local TUI when initialized with a workload pointing to an invalid image reference.
- Sysctl Cleanups: Fixed formatting bugs (tabs vs. spaces) on differing kernel versions and adjusted kernel module load orders to ensure properties apply correctly on boot.
Daniel Ward
Co-founder & CTO
Daniel runs engineering at Admiral and has a rich background in highly distributed linux systems across infrastructure and finance.