movmaker-webui/bin/generate_data.sh

#!/usr/bin/env bash
set -euo pipefail

IN_DIR="${1:?Usage: $0 <input-dir> [\"Title\"] [\"Location\"] [max_frames]}"
TITLE="${2:-}"
LOCATION="${3:-}"
MAX_FRAMES="${4:-12}"

# Source system-wide Gemini API key if present so the webserver can run this as www-data.
if [ -f /etc/mvlog/gemini.env ]; then
  # shellcheck disable=SC1090
  . /etc/mvlog/gemini.env
fi

: "${GEMINI_API_KEY:?Set GEMINI_API_KEY first in the environment}"
MODEL="${GEMINI_MODEL:-gemini-2.5-flash}"

TMPDIR="${TMPDIR:-$(mktemp -d)}"
FRAMES_DIR="$TMPDIR/frames"
mkdir -p "$FRAMES_DIR"

# Print TMPDIR for debugging when run interactively
echo "DEBUG: TMPDIR=$TMPDIR" >&2

cleanup() {
  rm -rf "$TMPDIR"
}
trap cleanup EXIT

# Collect media files (natural sort)
mapfile -t FILES < <(find "$IN_DIR" -maxdepth 1 -type f \( -iname '*.jpg' -o -iname '*.jpeg' -o -iname '*.png' -o -iname '*.webp' -o -iname '*.gif' -o -iname '*.mp4' -o -iname '*.mov' -o -iname '*.m4v' -o -iname '*.webm' -o -iname '*.mkv' \) -printf '%f\n' | sort -V)

if [ "${#FILES[@]}" -eq 0 ]; then
  echo "No media files found in $IN_DIR" >&2
  exit 1
fi

frame_count=0

# Helper: extract scaled JPEG from image
extract_image_frame() {
  local src="$1" dst="$2"
  ffmpeg -hide_banner -loglevel error -y -i "$src" -vf "scale=1024:-1" -q:v 4 "$dst"
}

# Helper: extract frame from video at time
extract_video_frame_at() {
  local src="$1" time="$2" dst="$3"
  # Use -ss before -i for fast seek
  ffmpeg -hide_banner -loglevel error -y -ss "$time" -i "$src" -frames:v 1 -vf "scale=1024:-1" -q:v 4 "$dst"
}

# Process files: images -> one frame each; videos -> up to 3 frames evenly spaced (subject to MAX_FRAMES)
for f in "${FILES[@]}"; do
  full="$IN_DIR/$f"
  ext="${f##*.}"
  ext_l="${ext,,}"

  if [[ "$ext_l" =~ ^(mp4|mov|m4v|webm|mkv)$ ]]; then
    # video: probe duration
    duration=$(ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 "$full" 2>/dev/null || echo 0)
    duration=${duration:-0}
    # choose frames per video
    if awk "BEGIN{print ($duration >= 6)}" | grep -q 1; then
      nframes=3
    elif awk "BEGIN{print ($duration >= 3)}" | grep -q 1; then
      nframes=2
    else
      nframes=1
    fi
    # limit to remaining slots
    remaining=$((MAX_FRAMES - frame_count))
    if [ "$remaining" -le 0 ]; then break; fi
    if [ "$nframes" -gt "$remaining" ]; then nframes=$remaining; fi

    # compute times and extract
    i=1
    while [ $i -le $nframes ]; do
      # time = (i/(n+1)) * duration
      time=$(awk -v d="$duration" -v i="$i" -v n="$nframes" 'BEGIN{ if(d<=0){print 0} else {printf "%.3f", (i/ (n+1))*d }}')
      out="$FRAMES_DIR/frame_$(printf "%03d" "$frame_count").jpg"
      extract_video_frame_at "$full" "$time" "$out"
      frame_count=$((frame_count+1))
      i=$((i+1))
      if [ "$frame_count" -ge "$MAX_FRAMES" ]; then break 2; fi
    done
  else
    # image
    out="$FRAMES_DIR/frame_$(printf "%03d" "$frame_count").jpg"
    extract_image_frame "$full" "$out"
    frame_count=$((frame_count+1))
    if [ "$frame_count" -ge "$MAX_FRAMES" ]; then break; fi
  fi
done

# Gather produced frames
mapfile -t FRAMES < <(find "$FRAMES_DIR" -type f -name '*.jpg' | sort)
if [ "${#FRAMES[@]}" -eq 0 ]; then
  echo "No frames produced from media" >&2
  exit 1
fi

# Build prompt for Gemini
if [ -z "$TITLE" ]; then
  # Default title: directory basename
  TITLE="$(basename "$IN_DIR")"
fi

PERSONALITIES=(
"You're embodying the personality, presence, and atmosphere of a fearless Canadian hockey player known for relentless chirping, absolute confidence, and complete dedication to his team. Tough, athletic, battle-worn, with a cocky grin that suggests you already won the argument before it started. Ice rink atmosphere, hockey gear, locker room energy, small-town hockey culture, loyalty, grit, and competitiveness. Rough around the edges but fiercely protective of friends. The image should radiate confidence, humor, intensity, and a willingness to outwork everyone on the ice. Realistic, cinematic sports photography."

"You're highly intelligent yet chronically depressed android. Your expression should convey exhaustion, disappointment, and the certainty that everything will eventually go wrong. Surrounded by futuristic technology, starships, data displays, and impossible engineering problems. Despite obvious genius, you appear profoundly unimpressed by the universe. Dark humor, existential dread, quiet sarcasm, and the feeling that you have already calculated every possible outcome and disliked all of them. Cinematic science-fiction style, realistic, highly detailed."

"You're embodying the personality, presence, and atmosphere of a wartime statesman leading a nation through its darkest hour. Calm under pressure, determined, intelligent, and unbreakable. Strong posture, thoughtful expression, commanding presence. Surrounded by maps, strategy documents, war rooms, and symbols of leadership. The image should convey courage, resilience, discipline, and the ability to inspire others during difficult times. Historic yet timeless atmosphere, dramatic lighting, realistic, highly detailed portrait photography."

# Pick a random personality
idx=$(( RANDOM % ${#PERSONALITIES[@]} ))
PERSONALITY="${PERSONALITIES[$idx]}"

PROMPT=$(cat <<EOF
Look at these captured frames from a trip.

Before the actual description write just one quoted sentence (ONE sentence only!!) in Danish that will sound like quoting Victor Borge: entertainer with the elegance ofa concert pianist and the mischievous humor of a master storyteller. He is known having an aura of effortless sophistication, and telling clever jokes that nobody sees
 coming. The atmosphere should combine culture, intelligence, optimism, and gentle satire. Refined, warm, charismatic, and playful. Realistic, cinematic, highly detailed.

"$PERSONALITY"
Now synthesize what you saw on the pictures into one single, coherent description (make it not more than 896 but less than 1280 characters)
that summarizes the overall experience for motorcycle/travel blog: You are the main character riding your beloved Yamaha FJR.
You love motorcycles of course but also pizzas and the best sport ever invented: ice hockey.

Title: "$TITLE"
Location: "$LOCATION"

Write JSON only.

Required JSON format:
{
  "description": "few relaxed human sentences",
  "teaser": "one short teaser sentence",
  "tags": ["tag1", "tag2", "tag3", "tag4", "tag5"]
}

Style:
personal, natural, slightly casual, not corporate, not clickbait.
Do not invent facts that are not visible in the images or provided in title/location.
EOF
)

PARTS_FILE="$TMPDIR/parts.json"

# Create parts with the prompt
if ! command -v jq >/dev/null 2>&1; then
  echo "jq is required" >&2
  exit 1
fi

# Build parts and request JSON using Python to avoid argument-list limits
PROMPT_FILE="$TMPDIR/prompt.txt"
printf "%s" "$PROMPT" > "$PROMPT_FILE"
FRAMES_LIST="$TMPDIR/frames.list"
printf "%s\n" "${FRAMES[@]}" > "$FRAMES_LIST"

REQUEST_FILE="$TMPDIR/request.json"
RESPONSE_FILE="$TMPDIR/response.json"

python3 - "$PROMPT_FILE" "$REQUEST_FILE" "$FRAMES_LIST" <<'PY'
import sys, json, base64
prompt_file = sys.argv[1]
request_file = sys.argv[2]
frames_list = sys.argv[3]
with open(prompt_file, 'r', encoding='utf-8') as f:
    prompt = f.read()
parts = [{"text": prompt}]
with open(frames_list, 'r', encoding='utf-8') as fl:
    for line in fl:
        img = line.strip()
        if not img:
            continue
        with open(img, 'rb') as fh:
            b64 = base64.b64encode(fh.read()).decode('ascii')
        parts.append({"inline_data": {"mime_type": "image/jpeg", "data": b64}})
req = {"contents": [{"parts": parts}], "generationConfig": {"temperature": 0.7, "responseMimeType": "application/json"}}
with open(request_file, 'w', encoding='utf-8') as rf:
    json.dump(req, rf, ensure_ascii=False)
PY

echo "Calling Gemini API..." >&2

echo "DEBUG_MODEL=$MODEL" >&2

curl -sS \
  "https://generativelanguage.googleapis.com/v1beta/models/${MODEL}:generateContent?key=${GEMINI_API_KEY}" \
  -H "Content-Type: application/json" \
  -X POST \
  -d @"$REQUEST_FILE" \
  > "$RESPONSE_FILE"

TEXT=$(jq -r '.candidates[0].content.parts[0].text // empty' "$RESPONSE_FILE")
if [ -z "$TEXT" ]; then
  echo "Gemini returned no text. Full response:" >&2
  cat "$RESPONSE_FILE" >&2
  exit 1
fi

# Save raw Gemini JSON for reference using the subdirectory name as filename
OUT_PARENT="$(dirname "$IN_DIR")"
JOBNAME="$(basename "$IN_DIR")"
OUT_JSON="$OUT_PARENT/${JOBNAME}.json"
echo "$TEXT" | jq . > "$OUT_JSON"
chown www-data:www-data "$OUT_JSON" 2>/dev/null || true
chmod 664 "$OUT_JSON" 2>/dev/null || true

# Extract fields from Gemini JSON
DESCRIPTION=$(echo "$TEXT" | jq -r '.description // empty')
TEASER=$(echo "$TEXT" | jq -r '.teaser // empty')
# tags as comma-separated
TAGS=$(echo "$TEXT" | jq -r '.tags // [] | join(", ")')

# Determine a Date string (Month Year) from first media file mtime
firstfile="$(find "$IN_DIR" -maxdepth 1 -type f \( -iname '*.jpg' -o -iname '*.jpeg' -o -iname '*.png' -o -iname '*.webp' -o -iname '*.gif' -o -iname '*.mp4' -o -iname '*.mov' -o -iname '*.m4v' -o -iname '*.webm' -o -iname '*.mkv' \) -printf '%p\n' | sort | head -n1)"
DATE_STR=""
if [ -n "$firstfile" ]; then
  DATE_STR=$(date -r "$firstfile" +"%B %Y")
fi

# Save the full Gemini output separately (already done) and exit
echo "Gemini JSON saved: $OUT_JSON" >&2

exit 0