feat(lifelog-recorder): v0.4 — 段播放 + bottom sheet + 声纹回填

闭环 v0.3 的两个自然延伸： - SegmentPlayer.kt: MediaPlayer 包装，Opus 段原音播放 Android API 21+ MediaPlayer 原生解码 Ogg-Opus，验证 logcat 见「MetaDataBase: OpusHead.. mime: audio/opus brte: 16000 #chn: 1」音频属性设 USAGE_MEDIA + CONTENT_TYPE_SPEECH，单实例复用 - SpeakerMatcher.kt: backfillAll() 遍历所有有 embedding 的段，重算最近邻 profile + cosine 距离；threshold 0.5 复用 RecorderService 里的逻辑；只更新变化的行（避免 Room 全表写） enroll 落库后 MainActivity 自动 launch 调用 - MainActivity: 时间轴 tap row 不再直接弹 enroll dialog，改为 ModalBottomSheet 出「▶ 播放原音 / 记为说话人」二选一 TimelineRow 加 isPlaying 状态指示，正在播放的段前缀 "▶ " DisposableEffect 在 onDispose 调 player.stop() 防泄漏 HANDOFF.md §五更新（v0.4 移到「已完成」），加 SegmentPlayer.kt + SpeakerMatcher.kt 到文件清单。实测（Pixel 8 Pro，电影 audio 输入）： - 段 #11-14 显示 "未匹配 (0.09-0.13)" — 电影角色声纹与 Me anchor cosine 远低于 0.5 → 正确拒绝匹配 - 段 #13 19s opus 文件 MediaPlayer 立刻播放，文件 ~38 KB Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

feat(lifelog-recorder): v0.4 — 段播放 + bottom sheet + 声纹回填
闭环 v0.3 的两个自然延伸： - SegmentPlayer.kt: MediaPlayer 包装，Opus 段原音播放 Android API 21+ MediaPlayer 原生解码 Ogg-Opus，验证 logcat 见「MetaDataBase: OpusHead.. mime: audio/opus brte: 16000 #chn: 1」音频属性设 USAGE_MEDIA + CONTENT_TYPE_SPEECH，单实例复用 - SpeakerMatcher.kt: backfillAll() 遍历所有有 embedding 的段，重算最近邻 profile + cosine 距离；threshold 0.5 复用 RecorderService 里的逻辑；只更新变化的行（避免 Room 全表写） enroll 落库后 MainActivity 自动 launch 调用 - MainActivity: 时间轴 tap row 不再直接弹 enroll dialog，改为 ModalBottomSheet 出「▶ 播放原音 / 记为说话人」二选一 TimelineRow 加 isPlaying 状态指示，正在播放的段前缀 "▶ " DisposableEffect 在 onDispose 调 player.stop() 防泄漏 HANDOFF.md §五更新（v0.4 移到「已完成」），加 SegmentPlayer.kt + SpeakerMatcher.kt 到文件清单。实测（Pixel 8 Pro，电影 audio 输入）： - 段 #11-14 显示 "未匹配 (0.09-0.13)" — 电影角色声纹与 Me anchor cosine 远低于 0.5 → 正确拒绝匹配 - 段 #13 19s opus 文件 MediaPlayer 立刻播放，文件 ~38 KB Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
yibo
1 parent 95af5fc5
Showing 4 changed files with 219 additions and 10 deletions
lifelog-recorder/HANDOFF.md
lifelog-recorder/app/src/main/kotlin/com/lemon/lifelog/MainActivity.kt
lifelog-recorder/app/src/main/kotlin/com/lemon/lifelog/asr/SpeakerMatcher.kt
lifelog-recorder/app/src/main/kotlin/com/lemon/lifelog/audio/SegmentPlayer.kt
@@ -47,10 +47,12 @@ app/src/main/kotlin/com/lemon/lifelog/
 │   └── BootReceiver.kt          — BOOT_COMPLETED 自启 (Android 14 有 caveat 见 §七)
 ├── audio/
 │   ├── AudioCapture.kt          — AudioRecord wrapper，emits FloatArray frames
-│   └── SegmentWriter.kt         — Opus encode via MediaCodec + Ogg muxer，~17× 压缩
+│   ├── SegmentWriter.kt         — Opus encode via MediaCodec + Ogg muxer，~17× 压缩
+│   └── SegmentPlayer.kt         — MediaPlayer 包装，Opus 段原音播放（系统原生支持）
 ├── asr/
 │   ├── ParaformerEngine.kt      — sherpa-onnx paraformer 包装 ←★ 主 ASR
 │   ├── SpeakerEmbedder.kt       — sherpa-onnx 3dspeaker campplus 192-d 嵌入 + cosineSim
+│   ├── SpeakerMatcher.kt        — 回填工具：用当前 profile 集重算所有段的 speakerId
 │   ├── ModelDownloader.kt       — 首次启动下模型（paraformer 232 MB + speaker 28 MB）
 │   └── XfyunIatClient.kt        — 讯飞大模型 WS 客户端，仅 fallback / 云复核用
 └── data/
@@ -155,15 +157,15 @@ VAD 参数（在 `RecorderService.runPipeline` 里调）：
 ✅ **已完成**：
 - v0.2 (2026-06)：Opus 编码 / BootReceiver / 电池白名单
 - v0.3 (2026-06)：声纹识别（3dspeaker campplus）/ FTS4 中文搜索
+- v0.4 (2026-06)：段播放 (MediaPlayer Opus) / bottom sheet 操作菜单 / enroll 后自动回填老段
  
 剩下：
  
 1. **LLM 纪要 worker**：WorkManager cron。每小时 / 静音 30min 触发，按时间窗 group segments → POST 到 hermes gateway `http://127.0.0.1:8642/v1/chat/completions`（需要先把 hermes api_server platform 配置起来，见 `~/.hermes/config.yaml`）。
 2. **存储 TTL**：删超过 N 天的 .opus，文字永留。Periodic WorkManager job。
 3. **设置页**：存储用量、模型切换、provider 切换、暂停时段配置（v1 定的 MVP 是「一键暂停就够」，但 v2 可能要细化）。
-4. **段播放**：时间轴每条点开能直接播原音（Opus 系统原生支持，MediaPlayer 就能放）。当前点开是「记录说话人」对话框，需要分离 tap vs long-press 或加二级菜单。
-5. **多 anchor 平均 / 声纹回填**：现在 enroll 是单段单 anchor，可以「连选多段都标 我」自动算平均 embedding；并对老段做一次性 backfill 匹配。
-6. **回链 lemonPCRM**：Transcript 提到的人名 / 项目 → GraphStore `MENTIONED_IN` 边。
+4. **多 anchor 平均**：现在 enroll 是单段单 anchor，可以「连选多段都标 我」自动算平均 embedding 提高匹配率。
+5. **回链 lemonPCRM**：Transcript 提到的人名 / 项目 → GraphStore `MENTIONED_IN` 边。
  
 ### 存储预算（v0.2 后）
  
@@ -30,13 +30,16 @@ import androidx.compose.foundation.text.KeyboardOptions
 import androidx.compose.foundation.verticalScroll
 import androidx.compose.material3.AlertDialog
 import androidx.compose.material3.Button
+import androidx.compose.material3.ExperimentalMaterial3Api
 import androidx.compose.material3.HorizontalDivider
 import androidx.compose.material3.MaterialTheme
+import androidx.compose.material3.ModalBottomSheet
 import androidx.compose.material3.OutlinedTextField
 import androidx.compose.material3.Surface
 import androidx.compose.material3.Text
 import androidx.compose.material3.TextButton
 import androidx.compose.material3.TextField
+import androidx.compose.material3.rememberModalBottomSheetState
 import androidx.compose.runtime.Composable
 import androidx.compose.runtime.collectAsState
 import androidx.compose.runtime.DisposableEffect
@@ -54,10 +57,12 @@ import androidx.core.content.ContextCompat
 import com.lemon.lifelog.asr.ModelDownloader
 import com.lemon.lifelog.asr.ParaformerEngine
 import com.lemon.lifelog.asr.SpeakerEmbedder
+import com.lemon.lifelog.asr.SpeakerMatcher
 import com.lemon.lifelog.asr.XfyunIatClient
 import com.lemon.lifelog.asr.toFloatArrayLe
 import com.lemon.lifelog.asr.toLeBytes
 import com.lemon.lifelog.asr.transcribeWithStats
+import com.lemon.lifelog.audio.SegmentPlayer
 import com.lemon.lifelog.data.Repository
 import com.lemon.lifelog.data.Segment
 import com.lemon.lifelog.data.SegmentRow
@@ -112,8 +117,12 @@ private fun MainScreen() {
     val timeline by timelineFlow.collectAsState(initial = emptyList())
     val speakers by repo.speakers.observeAll().collectAsState(initial = emptyList())
  
-    // Enrollment dialog state.
+    // Sheet / dialog state.
+    var actionRow by remember { mutableStateOf<SegmentRow?>(null) }
     var enrollFor by remember { mutableStateOf<SegmentRow?>(null) }
+    val player = remember { SegmentPlayer() }
+    var nowPlayingId by remember { mutableStateOf<Long?>(null) }
+    DisposableEffect(player) { onDispose { player.stop() } }
  
     // Battery optimisation exemption — recompute whenever activity resumes
     // (user may have just returned from the system settings dialog).
@@ -210,7 +219,11 @@ private fun MainScreen() {
             verticalArrangement = Arrangement.spacedBy(4.dp),
         ) {
             items(timeline, key = { it.id }) { row ->
-                TimelineRow(row, onClick = { enrollFor = row })
+                TimelineRow(
+                    row = row,
+                    isPlaying = nowPlayingId == row.id,
+                    onClick = { actionRow = row },
+                )
             }
         }
  
@@ -247,6 +260,29 @@ private fun MainScreen() {
         Text(asrLog, style = MaterialTheme.typography.bodySmall)
     }
  
+    // Bottom sheet with row actions.
+    actionRow?.let { row ->
+        SegmentActionSheet(
+            row = row,
+            isPlaying = nowPlayingId == row.id,
+            onDismiss = { actionRow = null },
+            onPlay = {
+                if (nowPlayingId == row.id) {
+                    player.stop()
+                    nowPlayingId = null
+                } else {
+                    nowPlayingId = row.id
+                    player.play(row.audioPath) { nowPlayingId = null }
+                }
+                actionRow = null
+            },
+            onEnroll = {
+                enrollFor = row
+                actionRow = null
+            },
+        )
+    }
+
     // Speaker enrollment dialog.
     enrollFor?.let { row ->
         EnrollSpeakerDialog(
@@ -258,7 +294,9 @@ private fun MainScreen() {
                 enrollFor = null
                 if (target != null && name.isNotBlank()) {
                     scope.launch {
-                        enrollSpeakerForSegment(activity, repo, target.id, name)
+                        enrollSpeakerForSegment(repo, target.id, name)
+                        // Backfill all older segments with embeddings against the new profile.
+                        SpeakerMatcher.backfillAll(repo)
                     }
                 }
             },
@@ -266,6 +304,54 @@ private fun MainScreen() {
     }
 }
  
+@OptIn(ExperimentalMaterial3Api::class)
+@Composable
+private fun SegmentActionSheet(
+    row: SegmentRow,
+    isPlaying: Boolean,
+    onDismiss: () -> Unit,
+    onPlay: () -> Unit,
+    onEnroll: () -> Unit,
+) {
+    val sheetState = rememberModalBottomSheetState()
+    ModalBottomSheet(
+        onDismissRequest = onDismiss,
+        sheetState = sheetState,
+    ) {
+        Column(
+            modifier = Modifier.fillMaxWidth().padding(16.dp),
+            verticalArrangement = Arrangement.spacedBy(8.dp),
+        ) {
+            Text(
+                "段 #${row.id}  ·  ${"%.1fs".format(row.durationMs / 1000.0)}",
+                style = MaterialTheme.typography.titleMedium,
+            )
+            row.text?.let {
+                Text(it, style = MaterialTheme.typography.bodyMedium)
+            }
+            row.speakerName?.let { name ->
+                Text(
+                    "已识别：$name  (sim=${"%.2f".format(row.speakerSim ?: 0f)})",
+                    style = MaterialTheme.typography.labelMedium,
+                )
+            }
+            Spacer(Modifier.height(8.dp))
+            Button(
+                onClick = onPlay,
+                modifier = Modifier.fillMaxWidth(),
+            ) {
+                Text(if (isPlaying) "■ 停止" else "▶ 播放原音")
+            }
+            Button(
+                onClick = onEnroll,
+                modifier = Modifier.fillMaxWidth(),
+            ) {
+                Text("记为说话人")
+            }
+        }
+    }
+}
+
 @Composable
 private fun EnrollSpeakerDialog(
     row: SegmentRow,
@@ -317,7 +403,6 @@ private fun EnrollSpeakerDialog(
  * Future segments will be auto-matched against it (see RecorderService.refreshProfiles).
  */
 private suspend fun enrollSpeakerForSegment(
-    ctx: Context,
     repo: Repository,
     segmentId: Long,
     name: String,
@@ -373,7 +458,7 @@ private fun requestBatteryOptExemption(ctx: Context) {
 private val TIMELINE_FMT = SimpleDateFormat("HH:mm:ss", Locale.US)
  
 @Composable
-private fun TimelineRow(row: SegmentRow, onClick: () -> Unit) {
+private fun TimelineRow(row: SegmentRow, isPlaying: Boolean, onClick: () -> Unit) {
     val time = TIMELINE_FMT.format(Date(row.startTs))
     val durSec = "%.1fs".format(row.durationMs / 1000.0)
     val speakerLabel = when {
@@ -382,6 +467,7 @@ private fun TimelineRow(row: SegmentRow, onClick: () -&gt; Unit) {
         row.speakerSim != null -> "未匹配 (${"%.2f".format(row.speakerSim)})"
         else -> "—"
     }
+    val playMarker = if (isPlaying) "▶ " else ""
     Column(
         modifier = Modifier
             .fillMaxWidth()
@@ -389,7 +475,7 @@ private fun TimelineRow(row: SegmentRow, onClick: () -&gt; Unit) {
             .padding(vertical = 4.dp),
     ) {
         Text(
-            "$time  ·  $durSec  ·  #${row.id}  ·  $speakerLabel",
+            "$playMarker$time  ·  $durSec  ·  #${row.id}  ·  $speakerLabel",
             style = MaterialTheme.typography.labelSmall,
         )
         Text(
+package com.lemon.lifelog.asr
+
+import android.util.Log
+import com.lemon.lifelog.data.Repository
+import kotlinx.coroutines.Dispatchers
+import kotlinx.coroutines.withContext
+
+/**
+ * Backfill speakerId / speakerSim for all segments that already have a stored embedding,
+ * using the current SpeakerProfile set.  Cheap enough to run synchronously after
+ * any enroll / delete profile event (192-d cosine × O(segments × profiles)).
+ */
+object SpeakerMatcher {
+    private const val TAG = "SpeakerMatcher"
+    const val MATCH_THRESHOLD = 0.5f
+
+    suspend fun backfillAll(repo: Repository): BackfillStats = withContext(Dispatchers.IO) {
+        val profiles = repo.speakers.all().map { p ->
+            Cached(p.id, p.name, p.embedding.toFloatArrayLe())
+        }
+        val segments = repo.segments.allWithEmbedding()
+        var assigned = 0
+        var unassigned = 0
+        for (seg in segments) {
+            val emb = seg.embedding?.toFloatArrayLe() ?: continue
+            val (newId, newSim) = if (profiles.isEmpty()) {
+                null to null
+            } else {
+                var bestId: Long? = null
+                var bestSim = -1f
+                for (p in profiles) {
+                    val sim = cosineSim(emb, p.embedding)
+                    if (sim > bestSim) {
+                        bestSim = sim
+                        bestId = p.id
+                    }
+                }
+                if (bestSim >= MATCH_THRESHOLD) bestId to bestSim else null to bestSim
+            }
+            if (newId != seg.speakerId || newSim != seg.speakerSim) {
+                repo.segments.updateSpeaker(seg.id, newId, newSim)
+            }
+            if (newId != null) assigned++ else unassigned++
+        }
+        Log.i(TAG, "backfill: ${segments.size} segments, $assigned matched, $unassigned unmatched")
+        BackfillStats(total = segments.size, matched = assigned, unmatched = unassigned)
+    }
+
+    private data class Cached(val id: Long, val name: String, val embedding: FloatArray)
+
+    data class BackfillStats(val total: Int, val matched: Int, val unmatched: Int)
+}
+package com.lemon.lifelog.audio
+
+import android.media.AudioAttributes
+import android.media.MediaPlayer
+import android.util.Log
+
+/**
+ * One-at-a-time MediaPlayer for Opus segments.
+ * Android MediaPlayer natively decodes Ogg-Opus (API 21+).
+ *
+ * Lifecycle: call [play] to start a new clip (any previous one is released);
+ * [stop] when navigating away.
+ */
+class SegmentPlayer {
+    companion object {
+        private const val TAG = "SegmentPlayer"
+    }
+
+    private var player: MediaPlayer? = null
+    private var currentPath: String? = null
+
+    val nowPlaying: String? get() = currentPath
+
+    fun play(path: String, onComplete: () -> Unit) {
+        stop()
+        currentPath = path
+        player = MediaPlayer().apply {
+            setAudioAttributes(
+                AudioAttributes.Builder()
+                    .setUsage(AudioAttributes.USAGE_MEDIA)
+                    .setContentType(AudioAttributes.CONTENT_TYPE_SPEECH)
+                    .build(),
+            )
+            try {
+                setDataSource(path)
+                setOnCompletionListener {
+                    currentPath = null
+                    onComplete()
+                }
+                setOnErrorListener { _, what, extra ->
+                    Log.w(TAG, "MediaPlayer error what=$what extra=$extra path=$path")
+                    currentPath = null
+                    onComplete()
+                    true
+                }
+                prepare()
+                start()
+                Log.i(TAG, "playing $path duration=${duration}ms")
+            } catch (t: Throwable) {
+                Log.e(TAG, "failed to play $path", t)
+                release()
+                currentPath = null
+                onComplete()
+            }
+        }
+    }
+
+    fun stop() {
+        player?.let { mp ->
+            try {
+                if (mp.isPlaying) mp.stop()
+            } catch (_: IllegalStateException) {
+            }
+            mp.release()
+        }
+        player = null
+        currentPath = null
+    }
+}