PDChat: Voice Mode, so a user can ask out loud and hear the answer

[davidnmbond]

Adds Voice Mode to PDChat: a user can ask a question out loud, pause, and hear the answer. This is for Magic Suite MS-26473. The host side, which relays to self-hosted Kyutai speech services, is panoramicdata/MagicSuite#451.

Public API, additive only

• IChatService.VoiceEndpoints is a default interface member returning null, so every existing implementation compiles unchanged and shows no Voice Mode control.
• PDChatVoiceEndpoints(string ListenUrl, string SpeakUrl) is a new record documenting the websocket protocol the host must speak.
• PDChatVoiceState is a new enum, and PDChat.VoiceState / PDChat.VoiceError are new read-only properties.
• No [Parameter] is added or changed, so ComponentDocumentation.md is unaffected.

Behaviour

• Off by default. A "🎙️ Voice" switch in the header. Its title says what it does ("ask out loud and hear the answer"), it carries aria-pressed, and a status line gives the state: Listening, Thinking, Speaking, or an explanation.
• Capture. getUserMedia, resampled to 24 kHz in an AudioWorklet. A fixed-rate AudioContext cannot be connected to a microphone in every browser. Audio goes out in 80 ms frames to the host's listen endpoint, and the host decides when the question has ended.
• One path for spoken and typed questions. A spoken question is sent exactly as a typed one is, so the question and the written answer are both in the transcript.
• Only the finished answer is spoken. That means the first non-Typing reply to a spoken question. Later notifications are never spoken, and nothing is spoken with Voice Mode off. HTML markup is stripped first; the host does the rest.
• Half duplex. The microphone sends nothing while the answer is awaited or spoken, so the assistant never hears itself.
• Playback starts as the first audio arrives, rather than after the whole answer.
• Failures are explained: a refused microphone or a closed connection gets a message.
• The JS module (wwwroot/js/pdchat-voice.js) is imported only when Voice Mode is first switched on.

Verification

• 10 new bUnit tests (PDChatTests.Voice.cs) cover:
  • not offered without endpoints, or where input is not permitted;
  • off by default, and the switch's wording;
  • start, then a spoken turn sent as the user, with the microphone paused;
  • a Typing placeholder not spoken, the finished answer spoken as plain text, then listening resumed;
  • only the first reply spoken, and nothing spoken with Voice Mode off;
  • turning it off; a refused microphone.
• Red-checked: dropping the Typing guard fails the finished-answer test.
• PanoramicData.Blazor.slnx builds with 0 errors. The 2 warnings are the SDK's pre-existing "packaging disabled" notices for non-packable projects.
• All 3,690 tests pass.
• pdchat-voice.js is Prettier-clean and passes node --check.
• Not done: a demo-page check. The demo's DumbChatService has no speech endpoints to offer, so Voice Mode never appears there. Browser behaviour gets its real check once Magic Suite consumes the package.

🤖 Generated with Claude Code

[codacy-production]

Up to standards ✅

🟢 Issues 0 issues

Results:
0 new issues

View in Codacy

🟢 Metrics 82 complexity · 5 duplication

Metric	Results
Complexity	82
Duplication	5

View in Codacy

🟢 Coverage 48.39% diff coverage · -0.10% coverage variation

Metric	Results
Coverage variation	✅ -0.10% coverage variation
Diff coverage	✅ 48.39% diff coverage

View coverage diff in Codacy

Coverage variation details

	Coverable lines	Covered lines	Coverage
Common ancestor commit (470f58c)	15712	15712	100.00%
Head commit (d14aa9b)	15743 (+31)	15727 (+15)	99.90% (-0.10%)

Coverage variation is the difference between the coverage for the head and common ancestor commits of the pull request branch: <coverage of head commit> - <coverage of common ancestor commit>

Diff coverage details

	Coverable lines	Covered lines	Diff coverage
Pull request (#228)	31	15	48.39%

Diff coverage is the percentage of lines that are covered by tests out of the coverable lines that the pull request added or modified: <covered lines added or modified>/<coverable lines added or modified> * 100%

AI Reviewer: run a review on demand. To trigger the first review automatically, go to your organization or repository integration settings. AI can make mistakes. Always validate suggestions.

_{TIP This summary will be updated as you push new changes.}