Pull request overview

This PR extends AetherSDR’s KiwiSDR integration so that when a Kiwi RX antenna/profile is selected, the existing RX surfaces (RxApplet, VFO controls, spectrum overlay, automation snapshot, TCI, shortcuts/controllers) operate the Kiwi “external receive replacement” AGC + SQL behavior, while preserving normal Flex slice AGC/SQL behavior.

Changes:

• Adds independent Kiwi vs Flex AGC/SQL state in SliceModel, with new signals and “receive*” accessors used by UI/control surfaces.
• Implements KiwiSDR server-compatible AGC + squelch command formatting, plus SND squelch-flag parsing and local audio gating on squelch.
• Updates spectrum overlay rendering and auto-squelch routing for Flex vs Kiwi, and adds regression tests for protocol/model behavior.

Reviewed changes

Copilot reviewed 27 out of 27 changed files in this pull request and generated 2 comments.

Thanks for this @rfoust — it's a large change but a genuinely careful one. The Flex/Kiwi state separation in SliceModel (mirrored getters, distinct externalReceive* signals, independent overlay state in SpectrumWidget) is clean, the clean-room source attribution is thorough, and the protocol-level helpers (formatSquelchCommand/formatAgcCommand/squelchSliderLevelToMarginDb/agcDecayMsForMode) are nicely factored out and well-covered by kiwi_sdr_protocol_test. CI is green across build / macOS / Windows / CodeQL / a11y.

A few notes:

Copilot findings — both already addressed in the current revision (false positives against HEAD):

• The "Auto SQL doesn't survive slice switches" comment Copilot flagged is gone — connectSlice now reads "Flex keeps the applet-wide Auto mode across active Flex slice switches, while Kiwi replacement receive keeps Auto as per-slice external receive state," which matches the m_flexSqlMode else-if branch. ✅
• The squelchThresholdDb field comment in KiwiSdrClient.h now documents both meanings ("non-NBFM uses signed dB offset from median RSSI floor, while NBFM uses the raw 1-99 noise squelch level"), exactly what Copilot asked for. ✅

Conventions / safety — looks good:

• AppSettings used throughout (no stray QSettings); AutoSqlMarginDb/LastManualSquelchLevel reuse the existing keys.
• m_kiwiSdrSquelchMeterSamples is bounded (removeFirst() past kServerRssiMedianSamples) — no unbounded growth.
• Local SND squelch gating (pcm.fill(0) on the 0x40 flag) and ignoring SLC/LEVEL for Kiwi-assigned slices both line up with the documented server behavior.
• AgcTCalibrator::startAutoSweep() and the calibrate context-menu action are correctly disabled under externalReceiveReplacementActive().

One minor thing to consider (not blocking):

• TciProtocol::cmdAgcGain now returns receiveAgcThreshold(), which for a Kiwi-assigned slice is on the -160..0 dB scale rather than the conventional TCI agc_gain 0..100 range. A TCI client polling agc_gain on a Kiwi-replacement slice would see a negative value. This only matters in the TCI + Kiwi-replacement combination and is arguably the intended "report the active surface" behavior, but worth a deliberate decision (and maybe a one-line comment) since other TCI clients may assume the 0–100 contract.

Nice work — this is solid, and the regression tests give good confidence in the protocol math.

_{🤖 aethersdr-agent · cost: $9.6327 · model: claude-opus-4-8}

Approving — with one provenance commit pushed on top.

The code is strong and well-vetted (Flex/Kiwi state cleanly separated in SliceModel, protocol helpers factored + covered by kiwi_sdr_protocol_test, SND squelch gating + bounded RSSI-median buffer correct, both prior reviews' findings addressed, CI green).

Pushed 4c4c7091 to close the one gap I found: THIRD_PARTY_LICENSES was still scoped (from #3716) to "only … MSG keys and numeric badp denial codes" and cited only LGPL, but this PR also derives the SET squelch/SET agc command shapes (LGPL server source) and the AGC/squelch numeric ranges from CuteSDR (Simplified BSD) — a distinct, previously-unrecorded license. The commit broadens the scope line, lists the consulted LGPL files, and adds CuteSDR (BSD-2-Clause) as its own sub-component, pointing to docs/kiwisdr-cleanroom-design.md for the per-fact detail. Facts only; no code incorporated. The canonical license file now matches the shipped code.

One deliberate decision worth recording (non-blocking): TciProtocol::cmdAgcGain now returns receiveAgcThreshold() (−160..0 dB) for a Kiwi-replacement slice, vs the conventional TCI agc_gain 0..100 contract — so a TCI client polling that on a Kiwi slice sees a negative value. Reasonable as "report the active surface," but a one-line comment (or a follow-up to map it) would help TCI clients that assume 0–100.

Thanks @rfoust — careful work, and the clean-room attribution discipline is exactly right; I just brought the license file in sync.