refactor: decompose sensing-server main.rs (3,765 lines → 14 modules) — ADR-051

[ruvnet]

Problem

sensing-server/src/main.rs is 3,765 lines with cyclomatic complexity ~65. It contains 12 structs, 60+ functions, and a 37-field AppStateInner god object. This was identified as Finding #2 in the quality engineering analysis (#170) and is tracked as Sprint 2 in ADR-050.

Plan

Decompose into 14 focused modules across 6 phases. See ADR-051 for the full extraction plan.

Module Breakdown

Module	Contents	~Lines
cli.rs	Args struct, CLI parsing	100
state.rs	AppStateInner + all DTOs	220
signal.rs	Feature extraction, classification	350
smoothing.rs	EMA smoothing, vital sign filtering	180
routes_health.rs	Health/readiness/info endpoints	350
routes_model.rs	Model management endpoints	180
routes_recording.rs	Recording endpoints	210
routes_training.rs	Training + adaptive endpoints	120
routes_sensing.rs	Vitals, WASM, SONA endpoints	150
routes_pose.rs	Pose estimation + endpoints	280
websocket.rs	WS upgrade + client handlers	170
udp_receiver.rs	UDP CSI frame receiver	170
data_sources.rs	Windows WiFi + simulated data	400
router.rs	Route assembly	80

Result: main.rs drops to ~250 lines (startup only).

Extraction Phases

1. cli.rs + state.rs (types only, zero behavioral change)
2. signal.rs + smoothing.rs (pure functions)
3. routes_health.rs + routes_model.rs + routes_recording.rs
4. routes_training.rs + routes_sensing.rs + routes_pose.rs
5. websocket.rs + udp_receiver.rs + data_sources.rs
6. router.rs + final cleanup

Acceptance Criteria

• No file over 500 lines
• cargo check and cargo test pass after each phase
• All existing endpoints behave identically
• Each new module has at least smoke tests
• AppStateInner refactored into domain sub-states
• Add criterion benchmarks for CSI parsing, signal pipeline, WS fanout

References

• ADR-051: Sensing Server Decomposition
• ADR-050: Quality Engineering Response
• Issue Quality Engineering Analysis: 10-Report Assessment (Security, Performance, Testing, Architecture) #170: Quality Engineering Analysis