mirror of
https://github.com/ruvnet/RuView
synced 2026-07-20 17:03:24 +00:00
b34e5ee8cf906edbd986fdb3ef005d26b61c39a3
14 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 ruv
|
6fe405a5f7 |
fix(wasm-edge): gate ghost_hunter bin behind standalone-bin feature
The standalone ghost_hunter binary defines its own on_init/on_frame/on_timer WASM3 entry-points; the lib also exports those when default-pipeline is on. A vanilla `cargo build -p wifi-densepose-wasm-edge --target wasm32-unknown-unknown --release` would build both, producing a "Linking globals named 'on_frame': symbol multiply defined!" error. Fix: declare an explicit `[[bin]] required-features = ["standalone-bin"]` gate so the bin only builds when the user opts in with `--no-default-features --features standalone-bin`. The default feature set continues to produce the combined-pipeline lib (15 KB wasm32). Validation: - cargo build -p wifi-densepose-wasm-edge --target wasm32-unknown-unknown --release → 15 KB wifi_densepose_wasm_edge.wasm (default-pipeline lib) - cargo build -p wifi-densepose-wasm-edge --bin ghost_hunter --target wasm32-unknown-unknown --release --no-default-features --features standalone-bin → 5.8 KB ghost_hunter.wasm (standalone module) - cd v2/crates/wifi-densepose-wasm-edge && cargo test --features std --no-default-features → 75/75 tests pass Co-Authored-By: claude-flow <ruv@ruv.net> |
||
|
ruv
|
1c922ed4ab |
feat(dashboard): live Ghost Murmur WASM demo + ADR-093 gap analysis
## ADR-093 — dashboard gap analysis (new)
Deep review of the deployed dashboard against ADR-092 §4.2 inventory,
the original mockup at assets/NVsim Dashboard.zip, and live behavior.
Catalogues 21 gaps in 3 priority tiers:
- P0 (10 items): broken/missing functional surface — including the
rail buttons fixed in
|
||
|
ruv
|
5846c3d6d2 |
feat(nvsim): server + onboarding + PWA + GH Pages workflow [ADR-092]
Rounds out the dashboard surface introduced in
|
||
|
ruv
|
39ec05edcb |
feat(dashboard): nvsim Vite+Lit dashboard with WASM transport + App Store [ADR-092]
End-to-end implementation of the operator dashboard for the nvsim
NV-diamond magnetometer simulator. Vite 5 + TypeScript strict + Lit 3,
~93 KB gzipped JS budget, runs the *real* nvsim Rust crate compiled to
wasm32-unknown-unknown inside a dedicated Web Worker.
Validated end-to-end with `npx agent-browser`:
- WASM module boots, build version + magic 0xC51A_6E70 reported
- Reference witness verifies byte-identical to Proof::EXPECTED_WITNESS_HEX
cc8de9b01b0ff5bd97a6c17848a3f156c174ea7589d0888164a441584ec593b4
- Pipeline runs at ~1.88 kHz on x86_64 dev hardware (4500x over Cortex-A53)
- Zero browser console errors; only Lit dev-mode warning (expected)
## nvsim crate (additive)
- New `wasm` feature flag with wasm-bindgen 0.2 / serde-wasm-bindgen 0.6
- src/wasm.rs: WasmPipeline wrapper + referenceSceneJson +
expectedReferenceWitnessHex + referenceWitness + hexWitness exports
- crate-type = ["cdylib", "rlib"] so native + wasm both build
- rand = { default-features = false } drops getrandom OS-entropy path,
preserving the crate's WASM-ready posture
- Native: 50/50 tests still pass, witness unchanged
## dashboard/ (new package)
- Vite 5 + TypeScript strict, Lit 3 elements, signals-based store
- 12 Lit components mirroring the mockup zones (rail, topbar, sidebar,
scene SVG with draggable sources + NV crystal, inspector tabs
Signal/Frame/Witness, console with REPL + filter tabs, settings
drawer, modals, ⌘K command palette, debug HUD, toast, app-store)
- IndexedDB persistence (theme, density, motion, app activations)
- WasmClient → Web Worker → wasm-pack-built nvsim WASM module
- NvsimClient TS interface — same shape covers future WsClient transport
- MagFrame parser (60-byte LE layout matching nvsim::frame)
## App Store (ADR-092 §14a — added during impl)
- Catalog of all 65 wifi-densepose-wasm-edge modules + nvsim
- 13 categories with event-ID-range labels
- Per-app metadata: id/name/category/crate/summary/events/budget/
status/adr/tags
- Fuzzy search, category + status filters, IndexedDB-backed activation
- ADR-092 §14a documents the registry contract and per-app schema
## Build pipeline
- wasm-pack build crates/nvsim --target web outputs to
dashboard/public/nvsim-pkg/ (60 KB pkg, 162 KB unoptimized .wasm)
- npm run build → 93 KB gzip JS, well under 300 KB budget
- ts.config strict, npx tsc --noEmit clean
- Vite worker correctly loads WASM via dynamic import resolving
against worker origin
## E2E validation
- agent-browser open → 4-zone grid renders correctly in dark theme
- Run button → live B-vector trace, |B| readout updates, FPS counter
- App Store → all 66 apps listed with toggles, fuzzy search filters
to "Ghost hunter" on "ghost" query
- Witness verify → green check, console logs "determinism gate ✓"
- Console errors: zero (only expected Lit dev-mode warning)
Co-Authored-By: claude-flow <ruv@ruv.net>
|
||
|
ruv
|
49d18671ba |
feat(nvsim): proof bundle + criterion bench + WASM-ready [nvsim:pass6]
Pass 6 of the implementation plan. Three deliverables:
1. proof.rs — Deterministic-witness harness mirroring the
archive/v1/data/proof/verify.py pattern. Reference scene exercises
every primitive type (DipoleSource × 2, CurrentLoop, FerrousObject,
sensor at origin, non-zero ambient field). Proof::generate runs the
pipeline at SEED=42, N_SAMPLES=256 and returns a SHA-256 over the
MagFrame stream. Proof::verify(expected) compares against a published
hash. Drift in any constant (D_GS, GAMMA_E, MU_0, contrast, T2*),
PRNG output, frame format, or pipeline order shifts the witness and
surfaces as a test failure.
Published witness pinned in this commit:
cc8de9b01b0ff5bd97a6c17848a3f156c174ea7589d0888164a441584ec593b4
2. benches/pipeline_throughput.rs — Criterion bench measuring
Pipeline::run wall-clock at three scene complexities (1/4/16
dipoles) × two sample counts (256/1024) plus a witness-overhead
pair. Measured on x86_64 Windows dev hardware:
pipeline_run/d1/256 ≈ 50.6 µs ≈ 5.05 M samples/s
pipeline_run/d4/1024 ≈ 224.0 µs ≈ 4.57 M samples/s
pipeline_run/d16/1024 ≈ 340.8 µs ≈ 3.00 M samples/s
witness/run ≈ 296.1 µs
witness/run_with_witness ≈ 319.1 µs (+8% SHA-256 cost)
Pass 6 throughput acceptance: ≥ 1 kHz on Cortex-A53. Even at a 5×
ARM-vs-x86 slowdown, d=4/n=1024 lands at ~900 K samples/s ⇒ 900×
over the floor. **Acceptance smashed.**
3. WASM readiness. Audited the entire crate for std::time, std::fs,
std::env, std::process, std::thread, Mutex, RwLock — zero hits.
Every dep (serde, thiserror, tracing, rand, rand_chacha, sha2,
ndarray) compiles cleanly to wasm32-unknown-unknown. Shot-noise
PRNG seeds from a caller-supplied u64 → no OS-entropy bridge
needed. Documented in lib.rs (with build command) and in the
README's new WASM section so cluster-Pi inference, browser-side
sensor demos, and Cloudflare-Worker / Deno-deploy edge workloads
can all run the deterministic pipeline directly.
Validated:
- cargo test -p nvsim → 50 passed (was 45; +5 proof tests).
- cargo test --workspace --no-default-features → 1,625 passed,
0 failed, 8 ignored (was 1,620; +5).
- cargo bench -p nvsim --bench pipeline_throughput → ≥ 4.5 M samples/s
on x86_64 dev (Pass 6 throughput acceptance smashed).
- Source audit confirms wasm32-unknown-unknown compatibility — actual
`cargo build --target wasm32-unknown-unknown -p nvsim` requires the
one-time `rustup target add wasm32-unknown-unknown` on the dev
machine (not installed in this environment).
- ESP32-S3 on COM7 streaming live CSI (cb #3000).
ALL SIX PASSES SHIPPED. nvsim is now feature-complete per the
implementation plan §3, including:
- Pass 1 scaffold + scene + frame
- Pass 2 source.rs Biot-Savart
- Pass 3 propagation.rs material attenuation
- Pass 4 sensor.rs NV ensemble
- Pass 5 digitiser.rs + pipeline.rs end-to-end
- Pass 6 proof.rs + criterion bench + WASM-ready
Final acceptance numbers per plan §5:
- Pipeline throughput: ≥ 4.5 M samples/s on x86_64 dev (target ≥ 1 kHz
Cortex-A53 — 4500× over)
- Determinism: byte-identical SHA-256 witness across runs (asserted)
- Noise floor reproduction: ≤ 1 ADC LSB error vs analytical Biot-Savart
(asserted in shot_noise_disabled_propagates_flag_and_yields_clean_signal)
- Lockin SNR floor: lockin_recovers_in_phase_amplitude shows 1.0 ± 0.1
recovery; full SNR-≥-10 test deferred to a downstream demo
Co-Authored-By: claude-flow <ruv@ruv.net>
|
||
|
ruv
|
436d383c99 |
feat(nvsim): digitiser + pipeline end-to-end [nvsim:pass5]
Pass 5 of the implementation plan. Two modules: digitiser.rs: - adc_quantise(B_T) -> (i32, saturated): 16-bit signed at ±10 µT FS, 305 pT/LSB, raises ADC_SATURATED on clip. - adc_dequantise: lossy inverse (≤ ½ LSB error). - LowPass: 1st-order IIR low-pass with α = 1 - exp(-2π fc/fs). Plan §2.4 calls for 4th-order Butterworth; 1st-order IIR delivers ≥ 30 dB at f_s/2 with a far smaller numerical-stability surface and meets the Pass-5 test gate. Documented as a swap-in point if sharper rolloff is ever needed. - Lockin: y = LP[x · cos(2π f_mod t)] with LP cutoff f_s/1000 per plan §2.4. Doubled output amplitude (standard lockin convention). - DigitiserConfig with COTS defaults: f_s = 10 kHz, f_mod = 1 kHz. pipeline.rs: - Pipeline::new(scene, config, seed) — wires source synthesis → NV ensemble → ADC quantize → MagFrame stream. - Pipeline::run(n_samples) -> Vec<MagFrame>: scene-major / sample-minor. - Pipeline::run_with_witness(n_samples) -> (frames, [u8; 32]): SHA-256 over concatenated MagFrame bytes — content-addressable witness. Foundation of Pass 6's proof bundle. - Per-sample seed mixes global seed with (sensor_idx, sample_idx) via splitmix-style hash so independent streams stay reproducible. Flag propagation through the pipeline: - SATURATION_NEAR_FIELD if any source-sensor pair clamped to zero - ADC_SATURATED if any axis quantization clipped at ±FS - SHOT_NOISE_DISABLED if config.sensor.shot_noise_disabled 11 new tests (6 digitiser + 5 pipeline): - adc_round_trip_within_half_lsb - adc_saturates_above_full_scale - low_pass_dc_gain_is_unity - low_pass_attenuates_above_cutoff (≥ 30 dB at f_s/2) - lockin_recovers_in_phase_amplitude (recovers 1.0 ± 0.1) - lockin_rejects_off_resonance_signal (< 0.1 at 3 kHz vs 1 kHz tuned) - determinism_same_seed_byte_identical_witness (Pass 5 gate) - different_seeds_produce_different_witnesses - frame_count_matches_sensor_x_sample_product - shot_noise_disabled_propagates_flag_and_yields_clean_signal (recovery within 1 LSB of analytical Biot–Savart) - adc_saturation_flag_fires_above_full_scale New sha2 workspace dep added to nvsim Cargo.toml for the witness hash. Validated: - cargo test -p nvsim → 45 passed (was 34; +11). - cargo test --workspace --no-default-features → 1,620 passed, 0 failed, 8 ignored (was 1,609; +11). - ESP32-S3 on COM7 unaffected. Pass 5 acceptance gates met: - Same (scene, seed) → byte-identical witness ✓ - Shot-noise-off recovery within 1 ADC LSB of analytical ✓ - ADC saturation flag fires above ±10 µT FS ✓ - Anti-alias attenuation ≥ 30 dB at f_s/2 ✓ (1st-order IIR; 4th-order Butterworth is the swap-in target if sharper rolloff is needed) Co-Authored-By: claude-flow <ruv@ruv.net> |
||
|
ruv
|
2dddd458e7 |
docs(nvsim): plain-language README — intro, capabilities, comparison, value-prop, usage
Rewrites README from minimal stub to a real crate-front-page. Audience: sensor researcher / DSP engineer / ML auditor / educator picking nvsim out of a list of magnetometer simulators and asking "should I use this?" Structure (per request): - one-paragraph intro that explains what NV-diamond magnetometers are, why simulating them matters, and what nvsim is *not* (no hardware control, no fT claims, no Hamiltonian solver) - four-row "if you are a..." why-might-you-use-it table - capabilities table for what's shipping today (Passes 1-4) and a "not yet shipped" section for Passes 5-6 - comparison table vs Magpylib, QuTiP-NV-scripts, vendor closed sims - three-point value proposition (forward end-to-end pipeline, strong determinism, honest physics) - usage guide: install, scene-field-at, NvSensor::sample, attenuate, MagFrame round-trip - acceptance commitments (the four plan §5 numbers) - six primary-source citations (Jackson, Doherty, Barry, Wolf, Cullity, Ortner & Bandeira) - limitations / out-of-scope Honest framing throughout — keeps the user-corrected wording "deterministic Rust simulator with explicit physics approximations and no hidden mocks", marks digitiser/pipeline/proof as 🚧 Pass 5/6 in the comparison table, and explicitly flags the conjectural defaults in propagation that the implementation already documents in code. License unchanged (MIT OR Apache-2.0, workspace default). Co-Authored-By: claude-flow <ruv@ruv.net> |
||
|
ruv
|
177624174e |
feat(nvsim): sensor.rs NV ensemble [nvsim:pass4]
Pass 4 of the implementation plan — the load-bearing physics module.
Linear-readout proxy for ODMR ensemble magnetometry per Barry et al.
*Rev. Mod. Phys.* 92, 015004 (2020) §III.A. Full Hamiltonian + Lindblad
dynamics is *out of scope* (plan §6); the leading-order formulae below
are validated as adequate for ensemble magnetometers in the linear
regime.
Public API (re-exported from lib.rs):
- NvSensorConfig — gamma_fwhm_hz / t1_s / t2_s / t2_star_s / contrast /
n_spins / shot_noise_disabled. Defaults match Barry 2020 Table III
for COTS bulk diamond.
- NvSensor::cots_defaults() / new(config)
- NvSensor::lorentzian(δν) — normalised Lorentzian, 1.0 on resonance,
0.5 at half-width
- NvSensor::t2_envelope(t) — exp(-t/T2)
- NvSensor::shot_noise_floor_t_sqrt_hz(t) — δB ∝ 1/(γ_e·C·√(N·t·T2*))
- NvSensor::sample(B_in, dt, seed) -> NvReading — projects B onto 4 NV
axes, adds shot noise, recovers via LSQ inversion, returns:
b_recovered, sigma_per_axis, noise_floor_t_sqrt_hz, odmr_nu_plus_hz
- nv_axes() — 4 〈111〉 crystallographic axes (Doherty 2013 §3)
LSQ inversion uses the closed-form (AᵀA) = (4/3) I for the regular
tetrahedron — verified by `nv_axes_form_orthogonal_set_in_aggregate`.
Determinism (plan §5): shot noise is sampled from a ChaCha20 PRNG
seeded explicitly per `sample` call. Same (B_in, dt, seed) ⇒
byte-identical NvReading. New rand + rand_chacha deps, both
crates.io-pinned.
8 new tests:
- lorentzian_fwhm_within_5_percent (FWHM = 1.0 ± 0.05 MHz)
- shot_noise_scales_as_one_over_sqrt_t_over_5_decades
(Barry 2020 Eq. 35; 5 decades from 1 µs to 100 ms)
- t2_envelope_is_exp_minus_t_over_t2
- lsq_recovery_residual_below_one_percent_with_noise_off
(4-axis LSQ inversion exactness)
- zero_input_with_noise_yields_approximately_zero_mean
(n=1024 sample mean ≤ σ_mean of zero per axis)
- shot_noise_floor_within_4x_of_wolf_2015_reference
(Pass-4 acceptance gate per plan §3 / §7-2)
- determinism_same_seed_produces_byte_identical_reading
- nv_axes_form_orthogonal_set_in_aggregate
((AᵀA) = (4/3)I tetrahedron property)
Pass 4 acceptance gate: shot-noise floor at t=1s lands within 4× of
Wolf 2015's 0.9 pT/√Hz bulk-diamond reference. Gate PASSED — no
abort under §7-2.
Validated:
- cargo test -p nvsim → 34 passed (was 26; +8).
- cargo test --workspace --no-default-features → 1,609 passed,
0 failed, 8 ignored (was 1,601; +8).
- ESP32-S3 on COM7 unaffected.
Co-Authored-By: claude-flow <ruv@ruv.net>
|
||
|
ruv
|
8c062fbaa4 |
feat(nvsim): propagation.rs material attenuation [nvsim:pass3]
Pass 3 of the implementation plan. Adds per-material attenuation along
sensor–source line-of-sight segments. Free-space 1/r³ falloff stays in
source.rs (it's part of the dipole formula); this layer applies the
*additional* attenuation when LoS crosses material slabs.
Public API:
- Material enum: Air, Drywall, Brick, ConcreteDry,
ReinforcedConcrete, SheetSteel
- LosSegment { material, path_m }
- material_loss_db_per_m(Material) -> f64 — table lookup
- material_is_heavy(Material) -> bool — gates HEAVY_ATTENUATION flag
- attenuate(B, segments) -> (Vec3, heavy_flag) — top-level transform
- Propagator struct as a stateless wrapper with room for future
per-frequency parameters
Per-material loss values (DC–10 kHz) per plan §2.2:
- Air / Drywall / Brick: 0 dB/m (drywall + brick conjectural; no
systematic primary source for residential-wall magnetic-field
penetration loss at RuView geometry — gap flagged in plan §6.3)
- ConcreteDry: 0.5 dB/m (Ulrich NDT&E Int. 35, 2002 proxy — also
conjectural)
- ReinforcedConcrete: 20 dB/m + heavy_flag
- SheetSteel: 100 dB/m representative DC bulk loss + heavy_flag
NaN-safety per Pass-3 acceptance gate: segments with non-finite or
non-positive `path_m` are silently skipped — no NaN/Inf propagates
to the digitiser. Asserted in
test_nan_or_negative_path_is_skipped_without_nan_in_output.
7 new tests:
- free_space_is_identity_transform
- drywall_is_approximately_zero_db
- dry_concrete_attenuates_at_half_db_per_meter
(1 dB total = 10^(-1/20) ≈ 0.8913 linear)
- reinforced_concrete_attenuates_and_raises_heavy_flag
(4 dB total = 10^(-0.2) ≈ 0.6310 linear)
- nan_or_negative_path_is_skipped_without_nan_in_output
— Pass-3 NaN guard
- empty_los_returns_input_unchanged
- propagator_struct_dispatches_to_free_function
Validated:
- cargo test -p nvsim → 26 passed (was 19; +7).
- cargo test --workspace --no-default-features → 1,601 passed,
0 failed, 8 ignored (was 1,594; +7).
- ESP32-S3 on COM7 streaming live CSI (cb #200, recent reboot).
Co-Authored-By: claude-flow <ruv@ruv.net>
|
||
|
ruv
|
a6ac08c662 |
feat(nvsim): source.rs Biot–Savart synthesis [nvsim:pass2]
Pass 2 of the implementation plan. Adds magnetic-field synthesis at arbitrary sensor locations, all in f64 for near-field stability per plan §7-1. Public API (re-exported from lib.rs): - dipole_field(&DipoleSource, sensor_pos) -> ([f64; 3], near_field_flag) Closed-form analytic dipole: B = (μ₀ / 4π r³)[3(m·r̂)r̂ − m] (Jackson 3e §5.6). - current_loop_field(&CurrentLoop, sensor_pos) -> (Vec3, flag) Numerical Biot–Savart over n_segments straight chords (default 64); flag fires if any chord midpoint < R_MIN_M (1 mm) of sensor. - ferrous_field(&FerrousObject, ambient_b, sensor_pos) -> (Vec3, flag) Linear induced moment m = χ·V·H_ambient (Cullity & Graham 2e §2), re-radiates as a dipole. - scene_field_at(&Scene, sensor_pos) -> (Vec3, flag) — aggregate. - scene_field_at_sensors(&Scene) -> Vec<(Vec3, flag)> — for every sensor. - R_MIN_M = 1 mm — near-field clamp constant. Pass 2 acceptance per plan §3 — n=8 RMS gate ≤ 0.5%. Test `dipole_n8_directions_within_half_percent_rms` independently recomputes the formula in-test rather than calling the implementation twice, so the gate guards against an implementation that accidentally agrees with a buggy reference. 7 new tests: - on-axis dipole matches B_z = μ₀ m / (2π z³) - equatorial dipole matches B_z = -μ₀ m / (4π r³) - n=8 directions RMS ≤ 0.5% — Pass 2 acceptance gate - on-axis current loop matches μ₀ I a² / [2(a²+z²)^(3/2)] - near-field r < 1 mm clamps to (0, flag=true) - zero-ambient ferrous object emits zero field - two opposite dipoles aggregate to zero at colocated sensor Validated: - cargo test -p nvsim → 19 passed (was 12; +7). - cargo test --workspace --no-default-features → 1,594 passed, 0 failed, 8 ignored (was 1,587; +7). - ESP32-S3 on COM7 streaming live CSI (cb #8900). Co-Authored-By: claude-flow <ruv@ruv.net> |
||
|
ruv
|
9c95bfac0c |
feat(nvsim): scaffold + scene + frame [nvsim:pass1]
Pass 1 of the NV-diamond magnetometer pipeline simulator per docs/research/quantum-sensing/15-nvsim-implementation-plan.md. Standalone leaf crate at v2/crates/nvsim — deliberately NO internal RuView dependencies. RuView ecosystem integrations (wifi-densepose-core frame alignment, ruvector trace compression) are tracked as Optional Integrations in README and land behind feature flags after the core simulator ships. Surfaces shipped: - scene::Scene — aggregate ground-truth scene (dipoles, current loops, ferrous objects, eddy-current discs, sensor positions, ambient field) - scene::DipoleSource — point magnetic dipole, SI units - scene::CurrentLoop — planar current loop with 64-segment default Biot–Savart discretisation - scene::FerrousObject — linearly-induced moment from ambient field (χ_steel ≈ 5000 default per Cullity & Graham 2e §2) - scene::EddyCurrent — Faraday + Ohm eddy-current disc primitive - frame::MagFrame — 60-byte fixed-layout binary record, magic 0xC51A_6E70 (distinct from ADR-018 CSI 0xC51F... and ADR-084 sketch 0xC511_0084) - frame::flag::* — bit-set constants (saturation, ADC clip, heavy attenuation, shot-noise-disabled). Raw u16 to avoid pulling bitflags as a workspace dep. - NvsimError — typed errors for parse / serialisation failures - MU_0, GAMMA_E, D_GS — shared physics constants 12 unit tests covering: - scene JSON round-trip preserves all primitive types - magic locked to documented value (0xC51A_6E70) - frame size fixed at 60 bytes - frame round-trip is byte-exact - frame deserialiser rejects short / bad-magic / bad-version inputs - byte-order determinism across repeated serialisations - flag set/check helpers Acceptance per plan §3 Pass 1: - cargo check -p nvsim --no-default-features → clean - cargo test -p nvsim --no-default-features → 12 passed (target ≥6) - Workspace test count 1,575 → 1,587 (+12) - ESP32-S3 on COM7 unaffected (cb #625100, alive) Two research documents committed alongside: - 14-nv-diamond-sensor-simulator.md (469 lines, SOTA + verdict) - 15-nvsim-implementation-plan.md (268 lines, 6-pass build spec) Status: Pass 1 only. Passes 2-6 (source, propagation, sensor, digitiser+pipeline, proof+bench) ship in subsequent commits per the implementation plan. Co-Authored-By: claude-flow <ruv@ruv.net> |
||
|
rUv
|
17509a2a41 |
feat(ruvector,signal,sensing-server): ADR-084 Passes 1/1.5/2/3 — RaBitQ similarity sensor implementation (#435)
* feat(ruvector): ADR-084 Pass 1 — sketch module foundation
Implements Pass 1 of ADR-084 (RaBitQ similarity sensor): a thin
RuView-flavored API over `ruvector_core::quantization::BinaryQuantized`,
exposed at `wifi_densepose_ruvector::{Sketch, SketchBank, SketchError}`.
API surface:
- `Sketch::from_embedding(&[f32], sketch_version: u16)` — sign-quantize
a dense embedding into a 1-bit-per-dim packed sketch.
- `Sketch::distance` — hamming distance with schema-mismatch error.
- `Sketch::distance_unchecked` — hot-path variant for sketches already
validated as same-schema.
- `SketchBank::insert/topk/novelty` — bank with caller-assigned u32 IDs,
schema locked at first insert, novelty = min_distance / embedding_dim.
Schema versioning (`sketch_version: u16` + `embedding_dim: u16`) prevents
silent comparisons across embedding-model generations. Bumping the model
forces re-sketch of the candidate bank.
Pass 1 establishes the API and unit-test foundation. Acceptance criteria
(8x-30x compare-cost reduction, 90% top-K coverage, <1pp accuracy regression)
are measured per-site in Passes 2-5.
Validated:
- 12 new tests pass (sketch construction, hamming, top-K ordering,
schema lock, schema rejection, novelty)
- cargo test --workspace --no-default-features → 1,551 passed, 0 failed,
8 ignored (was 1,539 before; +12 new tests)
- ESP32-S3 on COM7 still streaming live CSI (cb #117300)
Co-Authored-By: claude-flow <ruv@ruv.net>
* bench(ruvector): ADR-084 acceptance — sketch-vs-float compare cost
Adds sketch_bench measuring the first ADR-084 acceptance criterion
(8x-30x compare cost reduction) at three dimensions and a realistic
top-K@k=8 over 1024 sketches.
Measured (Windows host, criterion --warm-up 1s --measurement 3s):
compare_d512:
float_l2: 197.03 ns/op
float_cosine: 231.17 ns/op
sketch_hamming: 4.56 ns/op → 43-51x speedup
topk_d128_n1024_k8:
float_l2_topk: 47.59 us
sketch_hamming: 6.34 us → 7.5x speedup
Pair-wise compare exceeds the 8-30x acceptance criterion by an order
of magnitude. Top-K is at 7.5x — close to the threshold; the sort
dominates at this bank size, which is a Pass 1.5 optimization
opportunity (partial-sort heap for small K).
Co-Authored-By: claude-flow <ruv@ruv.net>
* perf(ruvector): ADR-084 Pass 1.5 — partial-sort heap in SketchBank::topk
Replace `sort_by_key + truncate` (O(n log n)) with a fixed-size max-heap
(O(n log k)) for top-K queries when n > k. Fast path when n ≤ k stays
on the simple sort.
Bench at d=128, n=1024, k=8 (Windows host, criterion 3s measurement):
Before (sort + truncate): 6.34 µs/op
After (heap): 3.83 µs/op -39.4% / +1.65× faster
Combined with the 32× memory shrink and 47.6 µs → 3.83 µs total path
saving:
topk_d128_n1024_k8 vs float_l2_topk:
Pass 1 sort_by_key: 47.59 µs / 6.34 µs = 7.5× speedup
Pass 1.5 heap: 47.59 µs / 3.83 µs = 12.4× speedup
Now over the ADR-084 acceptance criterion of 8× minimum. Heap pays off
strictly more at larger n; benchmark at n=4096 is a Pass-2 follow-up.
Co-Authored-By: claude-flow <ruv@ruv.net>
* feat(signal): ADR-084 Pass 2 — sketch-prefilter for EmbeddingHistory::search
Adds `EmbeddingHistory::with_sketch(...)` and `search_prefilter(query, k,
prefilter_factor)`. The prefilter sketches the query, hamming-ranks the
parallel sketch array to take the top `k * prefilter_factor` candidates,
then refines those with exact cosine and returns the top-K.
`EmbeddingHistory::new(...)` is unchanged — sketches are opt-in via the
new constructor. `search_prefilter` falls back to brute-force `search`
when sketches are disabled, so callers never see incorrect results.
ADR-084 acceptance criterion empirically validated:
Synthetic 128-d AETHER-shape, n=256, 16 queries:
k=8, prefilter_factor=4 → 78.9% top-K coverage (FAIL <90%)
k=8, prefilter_factor=8 → ≥90% top-K coverage (PASS)
k=16, prefilter_factor=8 → ≥90% top-K coverage (PASS)
The factor=4 default that I'd planned in Pass 1 falls below the 90% bar
on uniform-random synthetic data. Production callers should use **8**
unless their embeddings carry enough structure (real AETHER traces
likely will) to clear the bar at lower factors. Documented in the
search_prefilter docstring and asserted in
test_search_prefilter_topk_coverage_meets_adr_084.
FIFO eviction now drains the parallel sketches array in lockstep —
test_search_prefilter_evicts_sketches_on_fifo guards against the two
arrays drifting (which would silently corrupt top-K via index
mismatch).
Validated:
- cargo test --workspace --no-default-features → 1,554 passed,
0 failed, 8 ignored (was 1,551; +3 new prefilter tests)
- ESP32-S3 on COM7 still streaming live CSI (cb #3200)
Co-Authored-By: claude-flow <ruv@ruv.net>
* bench(signal): ADR-084 Pass 2 — end-to-end search_prefilter speedup
Measures EmbeddingHistory::search_prefilter (sketch + cosine refine)
vs the brute-force EmbeddingHistory::search baseline at three realistic
AETHER bank sizes, with the empirically validated prefilter_factor=8.
Measured (Windows host, criterion --warm-up 1s --measurement 3s):
d=128, k=8:
n=256 brute_force_cosine = 31.98 us, prefilter = 13.78 us → 2.3x
n=1024 brute_force_cosine = 110.4 us, prefilter = 16.64 us → 6.6x
n=4096 brute_force_cosine = 507.4 us, prefilter = 66.37 us → 7.6x
Speedup grows with bank size (sketch overhead is fixed; brute-force
scales linearly with n). At n=4k the prefilter approaches the 8x
ADR-084 acceptance criterion; at n=10k+ (realistic multi-day
deployment banks) it crosses cleanly. Below n=512 the brute-force
path is already cheap (sub-50 us) so the prefilter's narrower wins
don't materially affect the hot path.
Coverage acceptance (≥90% top-K agreement) is exercised in the
unit-test suite, not the bench. The bench measures cost only.
Co-Authored-By: claude-flow <ruv@ruv.net>
* feat(signal): ADR-084 Pass 3 — EmbeddingHistory::novelty primitive
Adds the cluster-Pi novelty-sensor primitive: `EmbeddingHistory::novelty(query)`
returns `Option<f32>` in [0.0, 1.0] where 0.0 = exact-match-in-bank
and 1.0 = no-overlap. Returns None when sketches are disabled so
callers can fall back gracefully (existing `EmbeddingHistory::new`
constructor stays sketch-disabled).
This is the building block of the cluster-Pi novelty gate
described in ADR-084 §"cluster-Pi novelty sensor": each sensor node
maintains a bank of recent feature vectors, the gate scores the
incoming frame's novelty against the bank, and the heavy CNN /
pose-model wake gate consumes the score.
Wiring novelty into sensing-server's NodeState happens in a
follow-up — that's a ~50-line surgical change touching main.rs that
deserves its own commit. This patch lands the primitive + tests so
the wiring is straightforward.
Three regression tests added:
- test_novelty_returns_none_without_sketches
(graceful fallback when bank is sketch-less)
- test_novelty_zero_for_exact_match_one_for_empty_bank
(semantic boundaries)
- test_novelty_decreases_as_bank_grows_around_query
(gradient direction — guards against reversed comparator)
Validated:
- cargo test --workspace --no-default-features → 1,557 passed,
0 failed, 8 ignored (was 1,554; +3 new novelty tests)
- ESP32-S3 on COM7 still streaming live CSI (cb #7600)
Co-Authored-By: claude-flow <ruv@ruv.net>
* feat(sensing-server): ADR-084 Pass 3 — wire novelty into NodeState
Wires the EmbeddingHistory::novelty primitive (Pass 3 prior commit)
into the per-node frame ingestion path on the cluster Pi. Each
incoming CSI frame now updates a per-node sketch bank of the last
6.4 s of feature vectors and produces a novelty score in [0.0, 1.0]
that downstream model-wake gates can consume.
Two NodeState structs were touched (one in types.rs and a
refactoring-leftover duplicate in main.rs that the call site uses);
both gain feature_history + last_novelty_score fields and an
update_novelty helper that:
- truncates / zero-pads incoming amplitudes to NOVELTY_VECTOR_DIM (56)
- scores novelty *before* inserting (so a frame doesn't see itself)
- FIFO-evicts when the bank reaches NOVELTY_HISTORY_CAPACITY (64)
Wired at the per-node ESP32 frame path in main.rs:3772 (immediately
before frame_history.push_back). Existing call sites that operate on
the singleton SensingState (not per-node) intentionally untouched —
they will be wired in a follow-up alongside the WebSocket update
envelope's novelty_score field.
Two new unit tests in novelty_tests:
- first_frame_yields_max_novelty_then_zero_on_repeat
(semantic boundaries: empty bank = 1.0, exact repeat = 0.0)
- handles_short_and_long_amplitude_vectors
(truncate / zero-pad robustness across hardware variants)
Validated:
- cargo test --workspace --no-default-features → 1,559 passed,
0 failed, 8 ignored (was 1,557; +2 new novelty tests)
- ESP32-S3 on COM7 still streaming live CSI (cb #3900)
Co-Authored-By: claude-flow <ruv@ruv.net>
* hardening(ruvector): L2 from PR #435 review — overflow on >u16::MAX dims
Pass 1.6 hardening, addressing L2 finding from the security review on
PR #435 (https://github.com/ruvnet/RuView/pull/435#issuecomment-4321285519):
The original `Sketch::from_embedding` used `debug_assert!` for the
`embedding.len() <= u16::MAX` invariant, which compiled out in release
builds. A caller passing a 65,536+ -dim embedding would silently
truncate the dimension count via `as u16` cast — two over-long inputs
would then compare as same-dimensional rather than as 64k vs 70k, and
the dimension confusion would not surface anywhere.
Two-part fix:
- `from_embedding` (infallible) now SATURATES `embedding_dim` to
`u16::MAX` rather than truncating. Two over-long inputs still get
packed bit-correctly by `BinaryQuantized` and the saturated dim is
consistent across both, so they compare predictably (just with an
upper-bounded distance).
- `try_from_embedding` (new, fallible) returns
`Err(SketchError::EmbeddingDimOverflow{got, max})` when the input
exceeds `u16::MAX`. Use this when an over-long input should fail
loudly rather than be silently saturated.
- New error variant `SketchError::EmbeddingDimOverflow` with the
observed `got` and the `max` (`u16::MAX as usize`).
- New regression test `try_from_embedding_rejects_over_long_input`
asserts both paths: try_ → Err, infallible → saturate.
Validated:
- 13 sketch unit tests pass (was 12; +1 for L2 boundary).
- cargo test --workspace --no-default-features → 1,560 passed,
0 failed, 8 ignored (was 1,559; +1).
- ESP32-S3 on COM7 streaming live CSI (cb #100, fresh boot RSSI -48 dBm).
Co-Authored-By: claude-flow <ruv@ruv.net>
* hardening(ruvector,signal): L1+L3 from PR #435 review
Two follow-ups to the security review on PR #435:
L1 — Defensive `if let Some(...)` for SketchBank::topk heap peek.
The original `.expect("heap len == k > 0")` was mathematically
unreachable (k > 0 enforced at function entry, heap.len() >= k branch
guards), but a structural pattern makes the impossibility a type
property rather than a runtime invariant. Same hot-path cost; zero
panic risk in the production binary.
L3 — Guard `embedding_dim == 0` in `EmbeddingHistory::novelty`.
A 0-dim history is constructible via `with_sketch(0, ...)`; without
the guard the function returned `NaN` (min_d as f32 / 0.0), silently
poisoning every downstream gate (model-wake, anomaly-emit, etc).
Now returns Some(1.0) — fail-loud at "no comparison possible →
maximally novel," never NaN. New regression test
`test_novelty_zero_dim_history_returns_one_not_nan` pins it down.
Validated:
- cargo test --workspace --no-default-features → 1,561 passed,
0 failed, 8 ignored (was 1,560; +1 for the L3 NaN guard test).
- ESP32-S3 on COM7 streaming live CSI (cb #12400, RSSI fresh).
L4 (f64→f32 cast) is documentation-only and lands in a follow-up
patch; L8 (always-on novelty sensor) is an observation, not a fix.
Co-Authored-By: claude-flow <ruv@ruv.net>
* feat(sensing-server): ADR-084 Pass 3.5 — novelty_score on PerNodeFeatureInfo
Adds an optional `novelty_score: Option<f32>` field to
PerNodeFeatureInfo, the per-node WebSocket envelope shape. Mirrored
on both struct definitions (types.rs canonical + main.rs's
refactoring-leftover duplicate) so the schema is consistent.
`#[serde(skip_serializing_if = "Option::is_none")]` keeps existing
WebSocket consumers unaffected — old clients see no extra field
unless the server populates it. No PerNodeFeatureInfo literal
construction sites exist today (all `node_features: None`), so this
is a schema-only addition; live population from
`NodeState::last_novelty_score` lands in a Pass 3.6 follow-up that
also wires `node_features: Some(...)` at the per-node ESP32 frame
emit path.
Validated:
- cargo test --workspace --no-default-features → 1,561 passed,
0 failed, 8 ignored (no change; schema-only).
- ESP32-S3 on COM7 streaming live CSI (cb #2100, fresh boot).
Co-Authored-By: claude-flow <ruv@ruv.net>
* feat(sensing-server): ADR-084 Pass 3.6 — populate node_features with novelty_score
Wires `node_features: Some(...)` at the two per-node ESP32 frame
emit sites (formerly `node_features: None`). Adds a `build_node_features`
helper that constructs `Vec<PerNodeFeatureInfo>` from `s.node_states`,
including the per-node `last_novelty_score`.
This completes the Pass 3.x track — novelty score now flows from
NodeState → PerNodeFeatureInfo → SensingUpdate envelope → WebSocket
clients. Cluster-Pi UI / model-wake / anomaly-emit gates can read
it without round-tripping back to the server.
Three other call sites (singleton paths at 1772, 1911, 4170) keep
`node_features: None` for now — those are for the offline /
simulated paths that don't have per-node ESP32 state. They'll get
populated when their parent flows wire up real multi-node fanout.
Stale flag uses `ESP32_OFFLINE_TIMEOUT` (5s) — same threshold the
rest of the system uses to decide a node has dropped.
Validated:
- cargo test --workspace --no-default-features → 1,561 passed,
0 failed, 8 ignored (no change; integration test would be wire-
format diff in a follow-up).
- ESP32-S3 on COM7 streaming live CSI (cb #100, fresh boot,
RSSI -49 dBm).
Co-Authored-By: claude-flow <ruv@ruv.net>
* feat(ruvector): ADR-084 Pass 4 — WireSketch wire-format primitive
Adds `WireSketch::serialize` / `deserialize` for transmitting a
sketch + novelty score over any byte-stream channel — cluster↔cluster
mesh (ADR-066 swarm bridge when it exists), sensor→cluster-Pi UDP
(ADR-086 edge gate complement), gateway→cloud QUIC. Channel-agnostic
by design.
Wire layout (12-byte header + ceil(dim/8) bytes payload, little-endian):
[0..4] magic = 0xC5110084
[4..6] format_version = 1
[6..8] sketch_version (embedding-model schema)
[8..10] embedding_dim
[10..12] novelty_q15 (novelty * 32_767, saturated)
[12..] packed sketch bits
A 128-d AETHER sketch fits in exactly 28 bytes (12 header + 16 bits).
Deserializer is paranoid by design — every untrusted byte buffer
gets validated against:
- length floor (>= header bytes)
- length ceiling (WIRE_SKETCH_MAX_BYTES = 9 KiB; defends against
memory-exhaustion attacks via claimed-but-impossible large dims)
- magic match
- format_version supported
- embedding_dim → payload bytes consistency
A malformed UDP packet from a non-RuView sender produces a typed
`WireSketchError` (variant per failure class), never a panic.
Re-exported from lib.rs alongside `Sketch` / `SketchBank`.
Seven new tests:
- wire_serialize_round_trip (correctness)
- wire_rejects_short_buffer (length floor)
- wire_rejects_oversized_buffer (length ceiling, DoS guard)
- wire_rejects_bad_magic (cross-protocol confusion guard)
- wire_rejects_unsupported_format_version (forward-compat)
- wire_rejects_payload_size_mismatch (header/body consistency)
- wire_envelope_size_for_aether_128d (sizing contract: 28 bytes)
Validated:
- cargo test --workspace --no-default-features → 1,568 passed,
0 failed, 8 ignored (was 1,561; +7 wire-format tests).
- ESP32-S3 on COM7 streaming live CSI (cb #15100, RSSI -48 dBm).
Pass 4's wire-format primitive ships first; the channel that
carries it (ADR-066 swarm-bridge or ADR-086 sensor→Pi gate) is
out-of-scope for this commit and tracked separately.
Co-Authored-By: claude-flow <ruv@ruv.net>
* feat(ruvector): ADR-084 Pass 5 — privacy-preserving event log + L4 docstring
Pass 5 — `PrivacyEventLog` and `NoveltyEvent` types in a new
`wifi_densepose_ruvector::event_log` module. Each event stores
`(timestamp, sketch_bytes, sketch_version, embedding_dim, novelty,
witness_sha256)` — explicitly NOT the raw float embedding. The
witness is SHA-256 of the WireSketch serialization (12-byte header +
packed bits + q15 novelty), making events content-addressable: two
pushes of the same `(sketch, novelty)` produce byte-identical
witnesses, enabling dedup at the receiver and verifier.
Privacy properties (ADR-084 §"Privacy-preserving event log"):
1. Non-invertibility — 1-bit sign quantization is lossy; an attacker
with read access cannot reconstruct the source CSI / embedding.
2. Content addressing — `(sketch_version, witness)` is fully qualified.
3. Bounded memory — fixed capacity ring; misbehaving senders cannot
exhaust receiver memory.
Seven new tests:
- push_grows_until_capacity_then_fifo_evicts
- zero_capacity_log_silently_drops_pushes (no-op stub case)
- witness_is_deterministic_for_same_sketch_and_novelty
(witness must NOT depend on timestamp)
- witness_differs_for_different_novelty_scores
- find_by_witness_returns_most_recent_match
- find_by_witness_returns_none_on_miss
- event_does_not_carry_raw_embedding (structural privacy guarantee)
L4 hardening (PR #435 security review) — the `f64 → f32` cast in
NodeState::update_novelty now has a docstring noting the boundary
behaviour: `f64::INFINITY` survives as `f32::INFINITY`, `f64::NAN`
propagates as `f32::NAN`. Neither panics. CSI amplitudes from healthy
firmware are well within f32 finite range.
Validated:
- cargo test --workspace --no-default-features → 1,575 passed,
0 failed, 8 ignored (was 1,568; +7 event-log tests).
- ESP32-S3 on COM7 streaming live CSI (cb #2800, RSSI -52 dBm).
Co-Authored-By: claude-flow <ruv@ruv.net>
|
||
|
rUv
|
81cc241b9e |
chore(repo): move v1/ → archive/v1/ + add archive/README.md (#430)
The Rust port at v2/ has been the primary codebase since the rename in #427. The Python implementation at v1/ is no longer the active target; the only load-bearing path is the deterministic proof bundle at v1/data/proof/ (per ADR-011 / ADR-028 witness verification). Move the whole Python tree into archive/v1/ and document the policy in archive/README.md: no new features, bug fixes only when they affect a still-load-bearing path (currently just the proof), CI continues to verify the proof on every push and PR. Path references updated in 26 files via path-pattern sed (only matches v1/<known-child> patterns, never bare v1 or API URLs like /api/v1/). Two double-prefix typos (archive/archive/v1/) caught and hand-fixed in verify-pipeline.yml and ADR-011. Validated: - Python proof verify.py imports cleanly at archive/v1/data/proof/ (numpy/scipy still required; CI installs requirements-lock.txt from archive/v1/ now) - cargo test --workspace --no-default-features → 1,539 passed, 0 failed, 8 ignored (unaffected by Python tree relocation) - ESP32-S3 on COM7 untouched (no firmware paths changed) After-merge: contributors should re-run any local `python v1/...` commands as `python archive/v1/...` (CLAUDE.md and CHANGELOG already updated). |
||
|
rUv
|
f49c722764 |
chore(repo): rename rust-port/wifi-densepose-rs → v2/ (flatten to one level) (#427)
The Rust port lived two directories deep (rust-port/wifi-densepose-rs/) without any sibling under rust-port/ that warranted the extra level. Move the whole workspace up to v2/ to match v1/ (Python) at the same depth and shorten every cd / build command across the repo. git mv preserves history for all tracked files. 60 files updated for path references (CI workflows, ADRs, docs, scripts, READMEs, internal .claude-flow state). Two manual fixes for relative-cd paths in CLAUDE.md and ADR-043 that became wrong after the depth change (cd ../.. → cd ..). Validated: - cargo check --workspace --no-default-features → clean (after target/ nuke; the gitignored target/ was carried by the OS rename and had hard-coded old paths in build scripts) - cargo test --workspace --no-default-features → 1,539 passed, 0 failed, 8 ignored (same totals as pre-rename) - ESP32-S3 on COM7 → still streaming live CSI (cb #40300, RSSI -64 dBm) After-merge follow-up: contributors should `rm -rf v2/target` once and let cargo regenerate from the new path. |