Decouple transcripts from note text; end-to-end watch transfer acks

Schema v3: transcription attempts accumulate on AudioInfo (text, locale,
confidence, forced flag) — latest attempt is displayed, payload.text is
purely user writing, v2 notes migrate at read time. Editor shows the
transcript in its own live-updating section with insert-into-note; search
matches transcripts weighted by confidence. Watch recordings are now
deleted only on the phone's durable-ingest ack (transferUserInfo), closing
every phone-side loss window; failed sidecar writes unstage instead of
orphaning, and unreadable inbox recordings are surfaced in logs.

Claude-Session: https://claude.ai/code/session_014esDWi42URLEC6Cj17hGQ3

Notes/Model/Note.swift

@@ -40,10 +40,13 @@ struct CaptureContext: Codable, Sendable, Equatable {
 /// One note = one JSON file in the iCloud container, filed under
 /// `Records/YYYY/MM/<ULID>.json` (UTC-bucketed by the ULID's timestamp).
 struct Note: SyncDocument, VersionedDocument, Equatable {
-    /// v2 added `Payload.audio` for voice notes. Because it is optional, a v1
-    /// file decodes unchanged (missing key ⇒ nil) and `migrateIfNeeded` only
-    /// bumps the stamped version — no field back-fill is needed.
-    static let currentSchemaVersion = 2
+    /// v2 added `Payload.audio` for voice notes. v3 moved transcripts out of
+    /// `payload.text` into `AudioInfo.attempts`: text is now purely the user's
+    /// own writing, and every transcription attempt is preserved. v1/v2 files
+    /// decode unchanged (missing keys ⇒ nil/empty); `migrateIfNeeded` back-fills
+    /// v2 voice notes by moving a completed transcript from `text` into
+    /// `attempts` at read time.
+    static let currentSchemaVersion = 3
 
     struct Meta: Codable, Sendable, Equatable {
         var id: ULID
@@ -66,6 +69,24 @@ struct Note: SyncDocument, VersionedDocument, Equatable {
         case failed
     }
 
+    /// One finished transcription of a voice note. Attempts accumulate — a
+    /// re-transcription appends rather than overwrites, so no machine output is
+    /// ever lost and the user can be shown alternatives.
+    struct TranscriptionAttempt: Codable, Sendable, Equatable {
+        var text: String
+        /// BCP-47 locale the transcript was produced in.
+        var localeID: String
+        /// Character-weighted mean confidence 0…1, or nil when the engine
+        /// emitted none (`DictationTranscriber` locales need not report it).
+        /// Confidences are not comparable across locales/engines — display and
+        /// search-weighting metadata, never a cross-attempt selector.
+        var confidence: Double?
+        var transcribedAt: Date
+        /// Whether the user forced this language (re-transcribe menu) rather
+        /// than auto-detection picking it.
+        var languageWasForced: Bool
+    }
+
     /// The audio-note sidecar metadata. Present only on voice notes; the audio
     /// bytes themselves live in a `.m4a` next to the note JSON (path derived,
     /// never stored — see `audioRelativePath`).
@@ -79,6 +100,45 @@ struct Note: SyncDocument, VersionedDocument, Equatable {
         /// BCP-47 locale the transcript was produced in; nil while pending.
         var transcriptionLocaleID: String?
         var transcribedAt: Date?
+        /// Every finished transcription, oldest first. The **latest** attempt is
+        /// the displayed one — a forced re-transcribe is a statement of intent,
+        /// so recency wins, not confidence (which isn't cross-locale comparable).
+        var attempts: [TranscriptionAttempt] = []
+
+        /// The attempt currently shown for this note.
+        var latestAttempt: TranscriptionAttempt? { attempts.last }
+
+        private enum CodingKeys: String, CodingKey {
+            case duration, transcriberDeviceID, transcriptionStatus,
+                 transcriptionLocaleID, transcribedAt, attempts
+        }
+
+        init(
+            duration: TimeInterval,
+            transcriberDeviceID: String,
+            transcriptionStatus: TranscriptionStatus,
+            transcriptionLocaleID: String? = nil,
+            transcribedAt: Date? = nil,
+            attempts: [TranscriptionAttempt] = []
+        ) {
+            self.duration = duration
+            self.transcriberDeviceID = transcriberDeviceID
+            self.transcriptionStatus = transcriptionStatus
+            self.transcriptionLocaleID = transcriptionLocaleID
+            self.transcribedAt = transcribedAt
+            self.attempts = attempts
+        }
+
+        init(from decoder: any Decoder) throws {
+            let container = try decoder.container(keyedBy: CodingKeys.self)
+            duration = try container.decode(TimeInterval.self, forKey: .duration)
+            transcriberDeviceID = try container.decode(String.self, forKey: .transcriberDeviceID)
+            transcriptionStatus = try container.decode(TranscriptionStatus.self, forKey: .transcriptionStatus)
+            transcriptionLocaleID = try container.decodeIfPresent(String.self, forKey: .transcriptionLocaleID)
+            transcribedAt = try container.decodeIfPresent(Date.self, forKey: .transcribedAt)
+            // Absent on v2 files — decode to empty, back-filled by migration.
+            attempts = try container.decodeIfPresent([TranscriptionAttempt].self, forKey: .attempts) ?? []
+        }
     }
 
     struct Payload: Codable, Sendable, Equatable {
@@ -167,9 +227,16 @@ struct Note: SyncDocument, VersionedDocument, Equatable {
         )
     }
 
-    /// First line of the text, used as the display title.
+    /// What a reader should see as this note's body: the user's own text when
+    /// they wrote any, else the latest transcript. Transcripts live in
+    /// `audio.attempts`, never in `payload.text`.
+    var displayText: String {
+        payload.text.isEmpty ? (payload.audio?.latestAttempt?.text ?? "") : payload.text
+    }
+
+    /// First line of the display text, used as the display title.
     var title: String {
-        payload.text
+        displayText
             .split(separator: "\n", omittingEmptySubsequences: true)
             .first.map(String.init) ?? ""
     }