Kapitel 13/Tutorial.md aktualisiert

Kapitel 13/Tutorial.md

@@ -125,7 +125,7 @@ SFTP_USER=sftp_uploader
 SFTP_KEY=/home/clipper/.ssh/nc_sftp_ed25519
 DROP_BASE="/mnt/hdd/incoming"
 CLIPPER_PEAK_THRESHOLD=0.85
-
+CLIPPER_MATCH_TOLERANCE=4.0
 
 CLIPPER_IN=/srv/clipper/watch
 CLIPPER_OUT=/srv/clipper/out
@@ -826,86 +826,112 @@ Das folgende Schaubild zeigt dir die konkrete Verkabelung
    ```
 
    ```bash
-      #!/usr/bin/env bash
-      set -euo pipefail
+      #!/bin/bash
 
-      . /etc/clipper/clipper.env
+set -euo pipefail
 
-      ID="${1:?need VOD id}"
+# Umgebungsvariablen laden
+if [ -f "/etc/clipper/clipper.env" ]; then
+    set -a
+    . "/etc/clipper/clipper.env"
+    set +a
+else
+    echo "[ERROR] Environment file not found: /etc/clipper/clipper.env" >&2
+    exit 1
+fi
 
-      VOD_IN_MP4="${CLIPPER_OUT}/${ID}/original/${ID}.mp4"
-      OUT_BASE="${CLIPPER_OUT}/${ID}"
-      ANALYSIS="${OUT_BASE}/analysis"
-      LOGDIR="${CLIPPER_LOG}/${ID}"
+# Erwartete Umgebungsvariablen prüfen
+: "${CLIPPER_PEAK_THRESHOLD:?Environment variable CLIPPER_PEAK_THRESHOLD is not set.}"
+: "${CLIPPER_TMP:?Environment variable CLIPPER_TMP is not set.}"
+: "${CLIPPER_LOG:?Environment variable CLIPPER_LOG is not set.}"
 
-      mkdir -p "$ANALYSIS" "$LOGDIR"
-      exec > >(tee -a "${LOGDIR}/analyze.log") 2>&1
-      echo "== Analyze $ID =="
+# ID aus dem Aufruf übernehmen
+VOD_ID="$1"
 
-      echo "[FFMPEG] Szenewechselanalyse läuft..."
-      ffmpeg -hide_banner -loglevel error -i "${VOD_IN_MP4}" \
-      -vf "scale=-2:360,select=gt(scene\,0.30),showinfo" -an -f null - \
-      2> "${LOGDIR}/sceneinfo.log"
+VOD_PATH="/srv/clipper/out/$VOD_ID/original/$VOD_ID.mp4"
+TMP_DIR="$CLIPPER_TMP/$VOD_ID"
+TMP_AUDIO="$TMP_DIR/audio.wav"
+LOG_DIR="$CLIPPER_LOG/$VOD_ID"
+TMP_LOG_AUDIO="$LOG_DIR/audio.log"
+TMP_JSON="$TMP_DIR/candidates.json"
 
-      echo "[FFMPEG] Audiostatistik läuft..."
-      ffmpeg -hide_banner -loglevel error -i "${VOD_IN_MP4}" \
-      -vn -ac 1 -ar 16000 \
-      -af "astats=metadata=1:reset=2,ametadata=print:key=lavfi.astats.Overall.RMS_level" \
-      -f null - \
-      2> "${LOGDIR}/astats.log" || true
+mkdir -p "$TMP_DIR" "$LOG_DIR"
+exec > >(tee -a "${LOG_DIR}/analyze.log") 2>&1
+echo "== Analyze $VOD_ID =="
+echo "[INFO] VOD gefunden: $VOD_PATH"
+echo "[INFO] Extrahiere WAV aus $VOD_PATH → $TMP_AUDIO"
 
-      ANALYSIS="$ANALYSIS" LOGDIR="$LOGDIR" python3 - <<'PY'
-      import os, re, json, sys
-      from datetime import datetime
+# Audio extrahieren mit Logging
+ffmpeg -v warning -i "$VOD_PATH" -ac 1 -ar 16000 -vn "$TMP_AUDIO" 2> "$TMP_LOG_AUDIO"
+echo "[OK] Audio extrahiert: $TMP_AUDIO ($(du -h "$TMP_AUDIO" | cut -f1))"
 
-      def log(msg):
-         timestamp = datetime.now().strftime("%F %T")
-         print(f"[PY] [{timestamp}] {msg}")
+echo "[INFO] Verwende Schwelle: $CLIPPER_PEAK_THRESHOLD"
+echo "[INFO] Starte Python-Analyse..."
 
-      out = os.environ["ANALYSIS"]
-      logdir = os.environ["LOGDIR"]
+# Python-Analyse
+/srv/clipper/.venv/bin/python3 - <<EOF
+import os
+import json
+import numpy as np
+import soundfile as sf
+import librosa
 
-      scene_ts = []
-      log("Lese sceneinfo.log...")
-      try:
-         with open(os.path.join(logdir, "sceneinfo.log"), errors="ignore") as f:
-            for line in f:
-                  m = re.search(r"pts_time:([0-9]+(?:\.[0-9]+)?)", line)
-                  if m:
-                     scene_ts.append(float(m.group(1)))
-      except Exception as e:
-         log(f"Fehler beim Lesen von sceneinfo.log: {e}")
-         sys.exit(1)
+path = "$TMP_AUDIO"
+outfile = "$TMP_JSON"
+threshold = float(os.getenv("CLIPPER_PEAK_THRESHOLD", "0.85"))
 
-      log(f"{len(scene_ts)} Szenenwechsel gefunden.")
+try:
+    y, sr = librosa.load(path, sr=None, mono=True)
+except Exception as e:
+    print("[ERROR] Laden fehlgeschlagen:", e)
+    exit(1)
 
-      has_audio = False
-      ap = os.path.join(logdir, "astats.log")
-      if os.path.exists(ap):
-         log("Prüfe astats.log auf Audiodaten...")
-         with open(ap, errors="ignore") as f:
-            has_audio = "RMS_level" in f.read()
+if len(y) == 0:
+    print("[WARN] Keine Audio-Daten vorhanden.")
+    with open(outfile, 'w') as f:
+        json.dump([], f)
+    print(f"[DONE] {outfile} geschrieben mit 0 Clip(s)")
+    exit(0)
 
-      log(f"Audioanalyse: {'gefunden' if has_audio else 'nicht vorhanden'}")
+frame_length = sr // 2
+energy = np.array([
+    np.sum(np.abs(y[i:i+frame_length]))
+    for i in range(0, len(y), frame_length)
+])
 
-      cands = [{
-         "start": max(0.0, t - 2.0),
-         "end": t + 6.0,
-         "score": round(0.6 + (0.1 if has_audio else 0), 2),
-         "tags": ["scene-cut"] + (["audio-peak"] if has_audio else [])
-      } for t in scene_ts]
+cutoff = np.quantile(energy, threshold)
+high_energy = np.where(energy > cutoff)[0]
 
-      target = os.path.join(out, "candidates.json")
-      try:
-         with open(target, "w", encoding="utf-8") as f:
-            json.dump(cands, f, ensure_ascii=False, indent=2)
-            log(f"{len(cands)} Kandidaten gespeichert → {target}")
-      except Exception as e:
-         log(f"Fehler beim Schreiben von candidates.json: {e}")
-         sys.exit(2)
-      PY
+if len(high_energy) == 0:
+    print("[INFO] Keine Peaks über Schwelle gefunden.")
+    with open(outfile, 'w') as f:
+        json.dump([], f)
+    print(f"[DONE] {outfile} geschrieben mit 0 Clip(s)")
+    exit(0)
 
-      echo "== Done $ID =="
+# Zeiträume berechnen
+candidates = []
+t_start = None
+for i in range(len(energy)):
+    t = i * frame_length / sr
+    if energy[i] > cutoff:
+        if t_start is None:
+            t_start = t
+    elif t_start is not None:
+        t_end = t
+        if t_end - t_start > 5:
+            candidates.append({"start": round(t_start, 2), "end": round(t_end, 2)})
+        t_start = None
+
+if t_start is not None:
+    t_end = len(y) / sr
+    if t_end - t_start > 5:
+        candidates.append({"start": round(t_start, 2), "end": round(t_end, 2)})
+
+with open(outfile, 'w') as f:
+    json.dump(candidates, f)
+print(f"[DONE] {outfile} geschrieben mit {len(candidates)} Clip(s)")
+EOF
    ```
    1. **SSH Node – Analyze VOD** (Node-Name: Analyze VOD)
       - Node-Typ: SSH