#!/usr/bin/env bash
# Descarcă un video (Facebook, YouTube etc.), extrage audio, transcrie cu Whisper.
# Usage: ./transcribe_video.sh <URL> [language]
#        ./transcribe_video.sh <URL> [language] --save-kb
# Exemple:
#   ./transcribe_video.sh "https://www.facebook.com/share/v/1EdPt3q2sq/"
#   ./transcribe_video.sh "https://www.facebook.com/share/r/1akfPJYvTw/" ro --save-kb
#   ./transcribe_video.sh "https://youtu.be/xyz" ro

set -euo pipefail

URL="${1:-}"
LANG="${2:-en}"
SAVE_KB=0

# Parse flags
for arg in "$@"; do
  if [[ "$arg" == "--save-kb" ]]; then
    SAVE_KB=1
  fi
done

WORKDIR="/tmp/transcribe_$$"
PROJECT_ROOT="$(cd "$(dirname "$0")/.." && pwd)"
KB_DIR="$PROJECT_ROOT/memory/kb"

if [[ -z "$URL" ]]; then
  echo "Usage: $0 <URL> [language (default: en)] [--save-kb]"
  exit 1
fi

export PATH="/home/moltbot/bin:$PATH"

mkdir -p "$WORKDIR"
trap 'rm -rf "$WORKDIR"' EXIT

echo "→ Obțin informații video..."
INFO_JSON=$(yt-dlp "$URL" --dump-json --no-download -q 2>/dev/null || echo "{}")
TITLE=$(echo "$INFO_JSON" | python3 -c "import json,sys; d=json.load(sys.stdin); print(d.get('title','Unknown'))" 2>/dev/null || echo "Unknown")
CREATOR=$(echo "$INFO_JSON" | python3 -c "import json,sys; d=json.load(sys.stdin); print(d.get('uploader') or d.get('channel') or '')" 2>/dev/null || echo "")
DURATION=$(echo "$INFO_JSON" | python3 -c "import json,sys; d=json.load(sys.stdin); s=d.get('duration',0); print(f'{s//60}:{s%60:02d}')" 2>/dev/null || echo "?")

echo "→ Descarc video: $TITLE..."
yt-dlp "$URL" -o "$WORKDIR/video.%(ext)s" --no-playlist -q

VIDEO_FILE=$(ls "$WORKDIR"/video.* 2>/dev/null | head -1)
if [[ -z "$VIDEO_FILE" ]]; then
  echo "Eroare: descărcarea a eșuat."
  exit 1
fi

echo "→ Extrag audio..."
ffmpeg -i "$VIDEO_FILE" -vn -acodec pcm_s16le -ar 16000 -ac 1 "$WORKDIR/audio.wav" -y -loglevel error

echo "→ Transcriu cu Whisper (model: small, limbă: $LANG)..."
TRANSCRIPT=$(python3 -c "
import whisper
model = whisper.load_model('small')
result = model.transcribe('$WORKDIR/audio.wav', language='$LANG')
print(result['text'])
" 2>/dev/null)

echo ""
echo "=== $TITLE ==="
echo "$TRANSCRIPT"
echo ""
echo "✓ Transcriere completă."

if [[ "$SAVE_KB" == "1" ]]; then
  DATE=$(date +%Y-%m-%d)

  # Slug din titlu: lowercase, fără diacritice, doar alfanumerice și cratime
  SLUG=$(echo "$TITLE" | python3 -c "
import sys, re, unicodedata
s = sys.stdin.read().strip()
s = unicodedata.normalize('NFD', s)
s = ''.join(c for c in s if unicodedata.category(c) != 'Mn')
s = s.lower()
s = re.sub(r'[^a-z0-9]+', '-', s)
s = s.strip('-')[:50]
print(s)
")

  # Detectează categoria din URL
  if echo "$URL" | grep -qi "facebook\.com"; then
    CATEGORY="facebook"
    FORMAT="Reel (~${DURATION} min)"
  elif echo "$URL" | grep -qi "youtube\.com\|youtu\.be"; then
    CATEGORY="youtube"
    FORMAT="Video (~${DURATION} min)"
  else
    CATEGORY="media"
    FORMAT="Video (~${DURATION} min)"
  fi

  NOTE_DIR="$KB_DIR/$CATEGORY"
  mkdir -p "$NOTE_DIR"
  NOTE_FILE="$NOTE_DIR/${DATE}_${SLUG}.md"

  cat > "$NOTE_FILE" << NOTEEOF
# $TITLE

**Sursa:** $URL
**Data:** $DATE
**Creator:** $CREATOR
**Format:** $FORMAT
**Tags:** @coaching

---

## TL;DR

<!-- Completează un rezumat de 2-3 rânduri -->

---

## Transcrierea

$TRANSCRIPT
NOTEEOF

  echo ""
  echo "→ Notiță salvată: $NOTE_FILE"

  echo "→ Reindexez KB..."
  python3 "$PROJECT_ROOT/tools/update_notes_index.py"

  echo "✓ KB actualizat. Link: /echo/files.html#memory/kb/$CATEGORY/${DATE}_${SLUG}.md"
fi