feature: meeting transcriber

.kiro/specs/meeting-transcription/design.md

@@ -0,0 +1,65 @@
+# Meeting Transcription Mode — Implementation Plan
+
+## Overview
+Add a new "Meeting Mode" to Wispr that:
+1. Shows a floating square window with recording controls and live transcript
+2. Captures both system audio (what others say in meetings) and microphone audio (what you say)
+3. Separates speakers into "You" vs "Others" based on audio source
+4. Displays a scrolling live transcript in the window
+5. Allows copying/exporting the transcript for notes
+
+## Architecture Decisions
+
+### System Audio Capture
+macOS requires `ScreenCaptureKit` (macOS 13+) to capture system audio. This is the only sanctioned API — `AVAudioEngine` can only capture microphone input. We'll use `SCStreamConfiguration` with `capturesAudio = true` and `excludesCurrentProcessAudio = true`.
+
+This requires the **Screen Recording** permission (user grants in System Settings > Privacy & Security > Screen Recording).
+
+### Speaker Separation Strategy
+Instead of ML-based diarization (complex, heavy), we use a simple but effective approach:
+- **Microphone audio** → labeled as "You"
+- **System audio** → labeled as "Others"
+
+This works perfectly for meetings because system audio = remote participants, mic = you.
+
+### Dual Audio Engine
+Create a new `MeetingAudioEngine` actor that runs two capture pipelines in parallel:
+1. `AVAudioEngine` for microphone (existing approach)
+2. `SCStreamConfiguration` for system audio
+
+Both streams are resampled to 16kHz mono Float32 and fed to separate transcription instances.
+
+### Transcription Approach
+Run two parallel transcription sessions:
+- One for mic audio chunks → "You:" prefix
+- One for system audio chunks → "Others:" prefix
+
+Use chunked transcription (process every ~5-10 seconds of audio) for near-real-time results.
+
+## Implementation Tasks
+
+### Phase 1: Core Infrastructure
+- [x] 1.1 Create `MeetingTranscript` model (timestamped entries with speaker labels)
+- [x] 1.2 Create `MeetingAudioEngine` actor (dual capture: mic + system audio via ScreenCaptureKit)
+- [x] 1.3 Create `MeetingStateManager` (orchestrates meeting mode state machine)
+- [x] 1.4 Add Screen Recording permission handling to `PermissionManager`
+
+### Phase 2: Meeting Mode UI
+- [x] 2.1 Create `MeetingTranscriptView` (scrolling transcript with speaker labels)
+- [x] 2.2 Create `MeetingWindowPanel` (floating square NSPanel with controls)
+- [x] 2.3 Add "Meeting Mode" menu item to `MenuBarController`
+- [x] 2.4 Wire meeting window visibility to `MeetingStateManager`
+
+### Phase 3: Integration
+- [x] 3.1 Add meeting mode settings to `SettingsStore` (not needed for MVP — uses existing language settings)
+- [x] 3.2 Wire up `WisprAppDelegate` to bootstrap meeting mode services
+- [x] 3.3 Add transcript export (copy to clipboard / save as text file)
+
+## File Plan
+```
+wispr/Models/MeetingTranscript.swift          — transcript data model
+wispr/Services/MeetingAudioEngine.swift       — dual audio capture
+wispr/Services/MeetingStateManager.swift      — meeting mode coordinator
+wispr/UI/Meeting/MeetingTranscriptView.swift  — transcript UI
+wispr/UI/Meeting/MeetingWindowPanel.swift     — floating window
+```

wispr/Models/MeetingTranscript.swift

@@ -0,0 +1,63 @@
+//
+//  MeetingTranscript.swift
+//  wispr
+//
+//  Data model for meeting transcription entries with speaker labels.
+//
+
+import Foundation
+
+/// Identifies the audio source / speaker in a meeting transcript.
+enum MeetingSpeaker: String, Sendable, Equatable, Hashable {
+    case you = "You"
+    case others = "Others"
+}
+
+/// A single timestamped entry in a meeting transcript.
+struct MeetingTranscriptEntry: Identifiable, Sendable, Equatable {
+    let id: UUID
+    let speaker: MeetingSpeaker
+    let text: String
+    let timestamp: Date
+
+    init(speaker: MeetingSpeaker, text: String, timestamp: Date = Date()) {
+        self.id = UUID()
+        self.speaker = speaker
+        self.text = text
+        self.timestamp = timestamp
+    }
+}
+
+/// The full transcript of a meeting session.
+struct MeetingTranscript: Sendable, Equatable {
+    var entries: [MeetingTranscriptEntry] = []
+    let startTime: Date
+
+    init(startTime: Date = Date()) {
+        self.startTime = startTime
+    }
+
+    /// Formats the entire transcript as plain text for export.
+    func asPlainText() -> String {
+        let formatter = DateFormatter()
+        formatter.dateFormat = "HH:mm:ss"
+
+        return entries.map { entry in
+            let time = formatter.string(from: entry.timestamp)
+            return "[\(time)] \(entry.speaker.rawValue): \(entry.text)"
+        }.joined(separator: "\n")
+    }
+
+    /// Duration of the meeting so far.
+    var duration: TimeInterval {
+        Date().timeIntervalSince(startTime)
+    }
+
+    /// Formatted duration string (e.g. "12:34").
+    var formattedDuration: String {
+        let total = Int(duration)
+        let minutes = total / 60
+        let seconds = total % 60
+        return String(format: "%d:%02d", minutes, seconds)
+    }
+}