Port managed client audio and Windows application

docs/api-dotnet.md

@@ -125,7 +125,7 @@ wrap Opus controls so P/Invoke never calls C varargs. SafeHandle owns every nati
 encoder, decoder, DRED parser/state, and denoiser, including failed initialization.
 
 `OpusOptions` is an immutable record. Supported PCM rates are 8/12/16/24/48 kHz,
-one or two interleaved channels, and integral 10/20/40/60 ms frames. These match the
+one or two interleaved channels, and integral 5/10/20/40/60 ms frames. These match the
 current VoiceCat protocol's integer frame duration; fractional Opus frame durations
 are not exposed. Low-delay application mode requires at most 20 ms. Channel capture
 bandwidth is controlled separately by `MaximumBandwidthHz`; the production audio

docs/porting-to-dotnet.md

@@ -794,6 +794,14 @@ criterion from `roadmap.md`, re-proven against the new server.
 mix cycle; a manual listen test on Windows and macOS with no audible glitching over 10
 minutes.
 
+**Checkpoint (2026-09-16):** `VoiceCat.Audio` now owns local Opus streams, reframing at every
+protocol frame size, VAD/PTT/DTX, DRED → FEC → PLC receive recovery, bounded jitter, per-stream
+controls, RNNoise and stereo mixing. Its normal encode/decode/NR/mix cycle allocates zero
+managed bytes. Capture inputs use bounded non-waiting PCM rings. The Windows implementation
+uses direct C# WASAPI capture, loopback and playback instead of the proposed miniaudio device
+shim; the codec/DSP shim remains the only native component. A real device smoke passed, but
+the required ten-minute Windows/macOS listen test is still a manual release gate.
+
 ---
 
 ### Phase 6 — Client core (est. 3–4 weeks)
@@ -809,6 +817,12 @@ minutes.
 conversation through the C# server**, and a C# `vccli` interoperates with a C++ `vccli` on
 the same server. This is the full M0–M3 criterion re-proven end to end.
 
+**Checkpoint (2026-09-16):** `VoiceCat.Core` implements TOFU-gated TLS, concurrent correlated
+requests, snapshots/events, reconnects, encrypted UDP binding and send/receive stream
+lifecycle. Two managed clients exchange text and decoded PCM through the managed server.
+The remaining Phase 6 item is the managed console client and its explicit C++ CLI
+interoperability scenario.
+
 ---
 
 ### Phase 7 — Windows client (est. 1–2 weeks)
@@ -818,6 +832,14 @@ shape against a real, complete UI before you commit to two rewrites.
 
 **Exit criterion:** feature parity with the current WinForms build, NVDA smoke-tested.
 
+**Checkpoint (2026-09-16):** the shipped WinForms project references `VoiceCat.Managed`, not
+the P/Invoke core. The compatibility facade preserves UI-thread event pumping and stable
+capture IDs while delegating all protocol and audio state to the idiomatic managed projects.
+Channel moves automatically renegotiate active streams. A published build passed real WASAPI
+capture/playback and form-startup smoke tests and contains `voicecat_media.dll` but no
+`voicecat.dll`. Automated text, bidirectional PCM voice, multi-frame audio and stream-move
+tests pass. NVDA and the manual endurance/listen pass remain before the Phase 7 exit criterion.
+
 ---
 
 ### Phase 8 — macOS client (est. 4–5 weeks)