Add voice analytics filters and metrics

scripts/demo_smoke.sh

@@ -9,8 +9,9 @@ SMOKE_AUDIO="${SMOKE_AUDIO:-0}"
 SMOKE_VOICE="${SMOKE_VOICE:-0}"
 
 COOKIE_JAR="$(mktemp "${TMPDIR:-/tmp}/dreamweaver-cookie.XXXXXX")"
+VOICE_SMOKE_AUDIO="$(mktemp "${TMPDIR:-/tmp}/dreamweaver-voice-audio.XXXXXX")"
 cleanup() {
-  rm -f "$COOKIE_JAR"
+  rm -f "$COOKIE_JAR" "$VOICE_SMOKE_AUDIO"
 }
 trap cleanup EXIT
 
@@ -174,12 +175,13 @@ if [[ "$SMOKE_VOICE" == "1" ]]; then
   voice_turn_detail_json="$(get_json "$APP_URL/api/voice-sessions/$voice_session_id/turns/$voice_turn_id")"
   assert_jq "$voice_turn_detail_json" '.user_transcript | contains("小熊")' "voice fallback turn should keep user transcript"
   assert_jq "$voice_turn_detail_json" '.assistant_text != null and .assistant_text != ""' "voice fallback turn should return assistant text"
-  assert_jq "$voice_turn_detail_json" '.detected_intent == "start" and .requires_confirmation == false' "first voice turn should start the story without confirmation"
+  assert_jq "$voice_turn_detail_json" '.detected_intent == "start_story" and .requires_confirmation == false' "first voice turn should start the story without confirmation"
   echo "$voice_turn_detail_json" | jq '{id,status,detected_intent,requires_confirmation,assistant_audio_ready,assistant_text}'
 
   say "Submitting voice uploaded turn with demo transcript hint"
+  printf 'dreamweaver-demo-audio' > "$VOICE_SMOKE_AUDIO"
   voice_upload_json="$(post_form "$APP_URL/api/voice-sessions/$voice_session_id/turns" \
-    -F 'audio_file=@/dev/null;filename=turn.webm;type=audio/webm' \
+    -F "audio_file=@${VOICE_SMOKE_AUDIO};filename=turn.webm;type=audio/webm" \
     -F 'duration_ms=900' \
     -F 'transcript_hint=不要让小熊害怕，让月亮姐姐帮它')"
   voice_upload_turn_id="$(jq -r '.turn_id' <<<"$voice_upload_json")"
@@ -187,18 +189,27 @@ if [[ "$SMOKE_VOICE" == "1" ]]; then
 
   voice_upload_detail_json="$(get_json "$APP_URL/api/voice-sessions/$voice_session_id/turns/$voice_upload_turn_id")"
   assert_jq "$voice_upload_detail_json" '.user_transcript | contains("月亮姐姐")' "voice upload turn should expose hinted transcript"
-  assert_jq "$voice_upload_detail_json" '.detected_intent == "correct" and .assistant_text != null' "voice upload correction should continue the narrative"
-  echo "$voice_upload_detail_json" | jq '{id,status,transcription_provider,detected_intent,requires_confirmation,assistant_audio_ready,assistant_text}'
+  assert_jq "$voice_upload_detail_json" '.detected_intent == "correct_story" and .assistant_text != null' "voice upload correction should continue the narrative"
+  assert_jq "$voice_upload_detail_json" '.user_audio_duration_ms == 900' "voice upload turn should expose user audio duration"
+  echo "$voice_upload_detail_json" | jq '{id,status,transcription_provider,user_audio_duration_ms,detected_intent,requires_confirmation,assistant_audio_ready,assistant_text}'
 
   say "Checking voice session detail and analytics"
   voice_detail_json="$(get_json "$APP_URL/api/voice-sessions/$voice_session_id")"
   assert_jq "$voice_detail_json" '.current_turn_index >= 2 and (.recent_turns | length) >= 2 and (.events | length) >= 2 and .can_finalize == true' "voice session should include turns/events and be finalizable"
-  assert_jq "$voice_detail_json" '([.events[].event_type] | index("turn_transcribed")) != null and ([.events[].event_type] | index("turn_narrative_ready")) != null' "voice session should record key turn events"
+  assert_jq "$voice_detail_json" '([.events[].event_type] | index("turn_transcribed")) != null and ([.events[].event_type] | index("assistant_text_ready")) != null' "voice session should record key turn events"
   echo "$voice_detail_json" | jq '{id,status,current_turn_index,can_finalize,latest_detected_intent,events:([.events[].event_type] | unique)}'
 
   voice_analytics_json="$(get_json "$APP_URL/api/voice-sessions/analytics?days=7")"
   assert_jq "$voice_analytics_json" '.window_days == 7 and .total_sessions >= 1 and .total_turns >= 2 and .successful_turns >= 2' "voice analytics should include the smoke session"
-  echo "$voice_analytics_json" | jq '{window_days,total_sessions,total_turns,successful_turns,failed_turns,turn_success_rate,finalize_conversion_rate}'
+  assert_jq "$voice_analytics_json" '.total_user_audio_duration_ms >= 2100 and .avg_user_audio_duration_ms > 0 and .transcription_provider_counts.demo >= 1 and .transcription_provider_counts.fallback >= 1' "voice analytics should expose duration and provider distribution"
+  assert_jq "$voice_analytics_json" '.text_fallback_turns >= 1 and .uploaded_audio_turns >= 1 and .user_audio_turn_rate > 0 and .assistant_audio_ready_rate > 0 and .asr_success_rate > 0 and .tts_success_rate > 0' "voice analytics should expose turn mix and success rates"
+  echo "$voice_analytics_json" | jq '{window_days,total_sessions,total_turns,successful_turns,failed_turns,text_fallback_turns,uploaded_audio_turns,user_audio_turn_rate,assistant_audio_ready_rate,asr_success_rate,tts_success_rate,total_user_audio_duration_ms,avg_user_audio_duration_ms,transcription_provider_counts,confirmation_request_rate,turn_success_rate,finalize_conversion_rate}'
+
+  voice_demo_analytics_json="$(get_json "$APP_URL/api/voice-sessions/analytics?days=7&provider=demo")"
+  assert_jq "$voice_demo_analytics_json" '.provider == "demo" and .uploaded_audio_turns >= 1 and (.transcription_provider_counts | keys == ["demo"])' "voice analytics should filter by ASR provider"
+
+  voice_waiting_analytics_json="$(get_json "$APP_URL/api/voice-sessions/analytics?days=7&session_status=waiting_user")"
+  assert_jq "$voice_waiting_analytics_json" '.session_status == "waiting_user" and .total_sessions >= 1' "voice analytics should filter by session status"
 
   say "Finalizing voice session into story"
   voice_finalize_json="$(post_json "$APP_URL/api/voice-sessions/$voice_session_id/finalize" '{