feat: prepare Orion transcription pilot

modules/capabilities/_transcribe-recordings.sh

@@ -92,8 +92,9 @@ process_recording() (
   local transcript_dir=$3
   local archive_dir=$4
   local failed_dir=$5
-  local base stem transcript archive original_path recorded_at transcribed_at
+  local base stem transcript archive original_path recorded_at transcribed_at duration
   local work_dir log_file raw_output rendered_output draft state_file recovery_status
+  local diarization_succeeded=false
 
   base=$(basename "$source")
   stem=${base%.*}
@@ -129,6 +130,20 @@ process_recording() (
   raw_output="$work_dir/transcript"
   rendered_output="$work_dir/transcript.md"
 
+  if ! duration=$(ffprobe \
+    -v error \
+    -show_entries format=duration \
+    -of default=noprint_wrappers=1:nokey=1 \
+    "$source" 2>"$log_file"); then
+    fail_recording "$source" "$failed_dir" "Media validation failed" "$log_file"
+    return 1
+  fi
+  if [[ ! $duration =~ ^[0-9]+([.][0-9]+)?$ || -z ${duration//[.0]/} ]]; then
+    printf 'Invalid media duration: %s\n' "$duration" >"$log_file"
+    fail_recording "$source" "$failed_dir" "Media validation failed" "$log_file"
+    return 1
+  fi
+
   if ! ffmpeg \
     -nostdin \
     -hide_banner \
@@ -148,18 +163,24 @@ process_recording() (
     --file "$work_dir/audio.wav" \
     --language auto \
     --output-json-full \
-    --output-file "$raw_output" >>"$log_file" 2>&1 || [[ ! -s "${raw_output}.json" ]]; then
+    --output-txt \
+    --output-file "$raw_output" >>"$log_file" 2>&1 \
+    || [[ ! -s "${raw_output}.json" || ! -s "${raw_output}.txt" ]]; then
     fail_recording "$source" "$failed_dir" "Whisper transcription failed" "$log_file"
     return 1
   fi
 
-  if ! diarize-transcript \
-    --audio "$work_dir/audio.wav" \
-    --model "$TRANSCRIBE_DIARIZATION_MODEL" \
-    --transcript "${raw_output}.json" \
-    --output "$rendered_output" >>"$log_file" 2>&1 || [[ ! -s "$rendered_output" ]]; then
-    fail_recording "$source" "$failed_dir" "Speaker diarization failed" "$log_file"
-    return 1
+  if [[ -f "$TRANSCRIBE_DIARIZATION_MODEL/.complete" && -f "$TRANSCRIBE_DIARIZATION_MODEL/config.yaml" ]] \
+    && diarize-transcript \
+      --audio "$work_dir/audio.wav" \
+      --model "$TRANSCRIBE_DIARIZATION_MODEL" \
+      --transcript "${raw_output}.json" \
+      --output "$rendered_output" >>"$log_file" 2>&1 \
+    && [[ -s "$rendered_output" ]]; then
+    diarization_succeeded=true
+  else
+    printf 'Speaker diarization unavailable; publishing unlabeled transcript\n' >&2
+    cp -- "${raw_output}.txt" "$rendered_output"
   fi
 
   draft=$(mktemp "$transcript_dir/.transcript.XXXXXX")
@@ -168,12 +189,17 @@ process_recording() (
     printf -- '- Original recording: %s\n' "$original_path"
     printf -- '- Recording timestamp: %s\n' "$recorded_at"
     printf -- '- Transcription timestamp: %s\n' "$transcribed_at"
-    printf -- '- Engine: whisper.cpp %s (Vulkan)\n' "$TRANSCRIBE_ENGINE_VERSION"
+    printf -- '- Duration: %s seconds\n' "$duration"
+    printf -- '- Engine: whisper.cpp %s (%s)\n' "$TRANSCRIBE_ENGINE_VERSION" "$TRANSCRIBE_ENGINE_ACCELERATION"
     printf -- '- Model: %s\n' "$TRANSCRIBE_MODEL_NAME"
     printf -- '- Model SHA-256: %s\n' "$TRANSCRIBE_MODEL_SHA256"
-    printf -- '- Diarization engine: pyannote.audio %s (CPU)\n' "$TRANSCRIBE_DIARIZATION_ENGINE_VERSION"
-    printf -- '- Diarization model: %s\n' "$TRANSCRIBE_DIARIZATION_MODEL_NAME"
-    printf -- '- Diarization model revision: %s\n\n' "$TRANSCRIBE_DIARIZATION_MODEL_REVISION"
+    if [[ $diarization_succeeded == true ]]; then
+      printf -- '- Diarization engine: pyannote.audio %s (CPU)\n' "$TRANSCRIBE_DIARIZATION_ENGINE_VERSION"
+      printf -- '- Diarization model: %s\n' "$TRANSCRIBE_DIARIZATION_MODEL_NAME"
+      printf -- '- Diarization model revision: %s\n\n' "$TRANSCRIBE_DIARIZATION_MODEL_REVISION"
+    else
+      printf -- '- Diarization: unavailable; unlabeled fallback published\n\n'
+    fi
     cat -- "$rendered_output"
   } >"$draft"
   chmod 0644 "$draft"
@@ -203,7 +229,8 @@ process_scope() {
   local processing_dir="$recordings_dir/processing"
   local archive_dir="$recordings_dir/archive"
   local failed_dir="$recordings_dir/failed"
-  local source processing_target recording_status scope_status=0
+  local source processing_target recording_status signature index scope_status=0
+  local -a sources signatures
 
   if ! mkdir -p "$inbox_dir" "$processing_dir" "$archive_dir" "$failed_dir" "$transcript_dir"; then
     return 1
@@ -219,23 +246,42 @@ process_scope() {
     fi
   done < <(find "$processing_dir" -maxdepth 1 -type f -print0 | sort -z)
 
-  while IFS= read -r -d '' source; do
-    if ! is_recording "$source"; then
-      continue
+  while true; do
+    sources=()
+    signatures=()
+
+    while IFS= read -r -d '' source; do
+      if is_recording "$source" && signature=$(stat -c '%s:%y:%i' -- "$source"); then
+        sources+=("$source")
+        signatures+=("$signature")
+      fi
+    done < <(find "$inbox_dir" -maxdepth 1 -type f -print0 | sort -z)
+
+    if [[ ${#sources[@]} -eq 0 ]]; then
+      break
     fi
 
-    processing_target="$processing_dir/$(basename "$source")"
-    if ! move_no_clobber "$source" "$processing_target"; then
-      fail_recording "$source" "$failed_dir" "Processing filename already exists: $processing_target"
-      scope_status=1
-      continue
-    fi
-    process_recording "$processing_target" "$inbox_dir" "$transcript_dir" "$archive_dir" "$failed_dir"
-    recording_status=$?
-    if [[ $recording_status -ne 0 ]]; then
-      scope_status=1
-    fi
-  done < <(find "$inbox_dir" -maxdepth 1 -type f -print0 | sort -z)
+    sleep "${TRANSCRIBE_SETTLE_SECONDS:-300}"
+
+    for index in "${!sources[@]}"; do
+      source=${sources[$index]}
+      if [[ ! -f "$source" ]] || [[ $(stat -c '%s:%y:%i' -- "$source") != "${signatures[$index]}" ]]; then
+        continue
+      fi
+
+      processing_target="$processing_dir/$(basename "$source")"
+      if ! move_no_clobber "$source" "$processing_target"; then
+        fail_recording "$source" "$failed_dir" "Processing filename already exists: $processing_target"
+        scope_status=1
+        continue
+      fi
+      process_recording "$processing_target" "$inbox_dir" "$transcript_dir" "$archive_dir" "$failed_dir"
+      recording_status=$?
+      if [[ $recording_status -ne 0 ]]; then
+        scope_status=1
+      fi
+    done
+  done
 
   return "$scope_status"
 }
@@ -243,14 +289,6 @@ process_scope() {
 runtime_dir="${XDG_RUNTIME_DIR:?}/transcribe-recordings"
 mkdir -p "$runtime_dir"
 
-if [[ ! -f "$TRANSCRIBE_DIARIZATION_MODEL/.complete" || ! -f "$TRANSCRIBE_DIARIZATION_MODEL/config.yaml" ]]; then
-  printf 'Speaker diarization model is not installed; run prepare-transcription-diarization-model\n' >&2
-  exit 1
-fi
-
 set +o errexit
 process_scope "$TRANSCRIBE_SYNC_ROOT/recordings" "$TRANSCRIBE_SYNC_ROOT/notes/transcripts"
-personal_status=$?
-process_scope "$TRANSCRIBE_SYNC_ROOT/work/recordings" "$TRANSCRIBE_SYNC_ROOT/work/notes/transcripts"
-work_status=$?
-exit "$((personal_status != 0 || work_status != 0))"
+exit $?