What changed?

Adds Local Mesh Discovery (Settings → Local Mesh Discovery) — an automated multi-preset RF scanner that cycles through LoRa modem presets to audit the local mesh environment and recommend optimal radio configuration using on-device AI.

Scan Engine (DiscoveryScanEngine)

@MainActor @Observable state machine with states: idle → shifting → reconnecting → dwell → analysis → complete (plus paused and restoring).

• Sends AdminMessage setConfig.lora to switch the connected radio to each selected preset
• Handles radio reboot + BLE reconnect automatically (60-second timeout)
• Collects packets per-preset during configurable dwell windows (default 15 min)
• 2-Packet Rule: requires ≥2 DeviceMetrics to compute Δ airtime rate
• Graceful stop: halts mid-scan, saves partial results, restores the user's home preset
• Alerts and disables scan start when primary channel uses the default key (Alert and disable start scan when primary channel does not use default key #1706)

Data Model (3 new SwiftData @Model types)

• DiscoverySessionEntity — session-level aggregates: timestamp, presets scanned, total unique nodes, average channel utilization, message/sensor counts, furthest distance, AI summary, user location, completion status
• DiscoveryPresetResultEntity — per-preset metrics: node counts (direct/mesh/infrastructure), message/sensor counts, channel utilization, airtime rate, packet success/failure rates, raw LocalStats fields (TX/RX/bad/dupe/relay/relay-canceled/online/total/uptime)
• DiscoveredNodeEntity — per-node observations: short/long name, neighbor type (direct vs. mesh), position, distance, hop count, SNR/RSSI, message/sensor counts, infrastructure flag, computed iconName (social → person.2.fill, sensor → thermometer.medium)

Views

PDF Export (DiscoverySummaryPDF)

• UIGraphicsPDFRenderer + MKMapSnapshotter for map image
• FileDocument conformance for share sheet integration
• Includes session metadata, per-preset metrics, and map snapshot

Integration Points

• Navigation: SettingsNavigationState.localMeshDiscovery case, NavigationLink in Settings view
• Tips: DiscoveryScanTip (TipKit) explains the feature on first use
• FoundationModels: LanguageModelSession (iOS 26+) generates natural-language preset recommendations; gated with #if canImport(FoundationModels)
• 2.4 GHz Gating: LORA_24 preset hidden for hardware without SX1280/SX1281 support

Why did it change?

There was no existing way to understand mesh activity across different LoRa presets in a given area. Users had to manually switch presets, wait, observe, and compare — a tedious process. This feature automates that workflow and uses on-device AI (FoundationModels, iOS 26+; no internet required) to surface which preset is optimal for a location based on node count, channel utilization, and traffic patterns.

How is this tested?

• DiscoveryScanEngineTests (Swift Testing) — initial state, isScanning property, default dwell duration, state transitions
• DiscoveryModelTests (Swift Testing) — entity relationships, computed properties (icon classification based on message vs. sensor counts)
• Manual testing: single-preset scan with 15-minute dwell against a live radio, verified node collection, map rendering, BLE reconnect after preset change, PDF export, and session persistence across app restart

Screenshots/Videos (when applicable)

Checklist

• My code adheres to the project's coding and style guidelines.
• I have conducted a self-review of my code.
• I have commented my code, particularly in complex areas.
• I have verified whether these changes require an update to existing documentation or if new documentation is needed, and created an issue in the docs repo if applicable.
• I have tested the change to ensure that it works as intended.