#!/bin/bash

# Script to process .webm files and generate metadata using OpenAI API
# Requires: dotenv-cli, jq
# Usage: ./webm_metadata.sh [directory_path]

set -e  # Exit on any error

# Get the directory where the script is located
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"

# Get target directory (default to current directory)
TARGET_DIR="${1:-.}"
TARGET_DIR="$(cd "$TARGET_DIR" && pwd)"  # Convert to absolute path

# Check dependencies
command -v dotenv >/dev/null 2>&1 || { echo "Error: dotenv-cli is required. Install with: npm install -g dotenv-cli" >&2; exit 1; }
command -v jq >/dev/null 2>&1 || { echo "Error: jq is required. Install with: apt-get install jq or brew install jq" >&2; exit 1; }

# Check if .env exists in script directory
ENV_FILE="$SCRIPT_DIR/.env"
if [ ! -f "$ENV_FILE" ]; then
    echo "Error: .env file not found at $ENV_FILE"
    echo "Please create one with OPENAI_API_KEY=your_key"
    exit 1
fi

echo "Script directory: $SCRIPT_DIR"
echo "Target directory: $TARGET_DIR"
echo "Looking for .env at: $ENV_FILE"
echo ""

# Configuration
CSV_FILE="$TARGET_DIR/webm_metadata.csv"
DELAY_SECONDS=1

# Function to check if filename already processed
is_processed() {
    local filename="$1"
    grep -q "^\"$filename\"," "$CSV_FILE" 2>/dev/null
}

# Function to escape CSV field
escape_csv() {
    local field="$1"
    # Escape double quotes and wrap in quotes
    echo "\"$(echo "$field" | sed 's/"/\"\"/g')\""
}

# Function to make API call and extract metadata
process_file() {
    local filename="$1"
    echo "Processing: $filename"

    # Create the system prompt text
    local system_prompt="You are an AI assistant that receives a meme filename. Based on the filename, your task is to generate detailed metadata describing the meme in a structured format.

The metadata should include these fields:
- **type**: either \`video\` or \`image\` depending on the file extension.
- **sub_type**: a classification such as \`background\` or \`overlay\` (choose \`overlay\` if it seems like an edit or reaction video/image).
- **name**: a concise, title-cased name derived from the filename (remove file extension and normalize, without the Meme word if exist).
- **description**: a short paragraph describing the meme content and context inferred from the name, and reaction
- **keywords**: a comma-separated list of relevant keywords extracted from the filename or meme context.
- **media_1_mime_type**: the MIME type derived from the file extension (e.g., \`video/webm\`, \`image/png\`).

Return the output as a JSON object containing these fields. Use null for any unknown or missing values.

---

#### Example input:
\`7th element oiiaa cat.webm\`

#### Example output:
\`\`\`json
{
  \"type\": \"video\",
  \"sub_type\": \"overlay\",
  \"name\": \"7th Element Oiiaa Cat\",
  \"description\": \"a cat edited to mimic or react to Vitas' bizarre vocals from the viral \\\"7th Element\\\" performance.\",
  \"keywords\": \"cat, oiiaa, 7th element, vitas, webm, funny\",
  \"media_1_mime_type\": \"video/webm\"
}
\`\`\`"

    local assistant_example="\`\`\`json
{
  \"type\": \"video\",
  \"sub_type\": \"overlay\",
  \"name\": \"Angry Cat\",
  \"description\": \"A video meme featuring an angry-looking cat, typically used to express frustration, annoyance, or irritation in a humorous way.\",
  \"keywords\": \"angry, cat, reaction, meme, webm, annoyed, frustration\",
  \"media_1_mime_type\": \"video/webm\"
}
\`\`\`"

    # Use jq to properly construct the JSON payload
    local json_payload=$(jq -n \
        --arg system_prompt "$system_prompt" \
        --arg filename "$filename" \
        --arg assistant_example "$assistant_example" \
        '{
          "model": "gpt-4.1-nano",
          "input": [
            {
              "role": "system",
              "content": [
                {
                  "type": "input_text",
                  "text": $system_prompt
                }
              ]
            },
            {
              "role": "user",
              "content": [
                {
                  "type": "input_text",
                  "text": $filename
                }
              ]
            },
            {
              "role": "assistant",
              "content": [
                {
                  "type": "output_text",
                  "text": $assistant_example
                }
              ]
            }
          ],
          "text": {
            "format": {
              "type": "json_object"
            }
          },
          "reasoning": {},
          "tools": [],
          "temperature": 1,
          "max_output_tokens": 2048,
          "top_p": 1,
          "store": true
        }')

    # Make the API call
    local response=$(curl -s "https://api.openai.com/v1/responses" \
        -H "Content-Type: application/json" \
        -H "Authorization: Bearer $OPENAI_API_KEY" \
        -d "$json_payload")

    # Check for API errors
    if echo "$response" | jq -e '.error' >/dev/null 2>&1; then
        echo "API Error for $filename:"
        echo "$response" | jq '.error'
        return 1
    fi

    # Extract the response text from the correct path
    local response_text=$(echo "$response" | jq -r '.output[0].content[0].text // empty')

    if [ -z "$response_text" ]; then
        echo "Error: Empty response for $filename"
        return 1
    fi

    # With JSON mode enabled, response should always be direct JSON
    local json_content="$response_text"

    # Validate that it's valid JSON before parsing
    if ! echo "$json_content" | jq . >/dev/null 2>&1; then
        echo "Error: Invalid JSON response for $filename"
        echo "Content: $json_content"
        return 1
    fi

    # Parse individual fields
    local type=$(echo "$json_content" | jq -r '.type // "unknown"')
    local sub_type=$(echo "$json_content" | jq -r '.sub_type // "unknown"')
    local name=$(echo "$json_content" | jq -r '.name // "unknown"')
    local description=$(echo "$json_content" | jq -r '.description // "unknown"')
    local keywords=$(echo "$json_content" | jq -r '.keywords // "unknown"')
    local media_1_mime_type=$(echo "$json_content" | jq -r '.media_1_mime_type // "unknown"')

    # Write to CSV
    local csv_line="$(escape_csv "$filename"),$(escape_csv "$type"),$(escape_csv "$sub_type"),$(escape_csv "$name"),$(escape_csv "$description"),$(escape_csv "$keywords"),$(escape_csv "$media_1_mime_type")"
    echo "$csv_line" >> "$CSV_FILE"

    echo "✓ Successfully processed: $filename"
}

# Export functions so they're available in the dotenv subshell
export -f is_processed
export -f escape_csv
export -f process_file
export CSV_FILE
export DELAY_SECONDS
export TARGET_DIR

# Change to script directory so dotenv can find the .env file
cd "$SCRIPT_DIR"

# Use dotenv to run the main processing logic
dotenv -- bash -c '
# Check if OPENAI_API_KEY is loaded
if [ -z "$OPENAI_API_KEY" ]; then
    echo "Error: OPENAI_API_KEY not found in environment variables"
    exit 1
fi

echo "✓ OpenAI API key loaded successfully"
echo "CSV file will be: $CSV_FILE"
echo "Delay between requests: ${DELAY_SECONDS}s"
echo ""

# Create CSV with headers if it doesn'\''t exist
if [ ! -f "$CSV_FILE" ]; then
    echo "filename,type,sub_type,name,description,keywords,media_1_mime_type" > "$CSV_FILE"
    echo "Created $CSV_FILE with headers"
fi

echo "Starting processing of .webm files..."

# Count total files and processed files
total_files=$(ls -1 "$TARGET_DIR"/*.webm 2>/dev/null | wc -l)
processed_count=0
skipped_count=0
error_count=0

if [ "$total_files" -eq 0 ]; then
    echo "No .webm files found in $TARGET_DIR"
    exit 0
fi

echo "Found $total_files .webm files"

# Process each .webm file
for webm_file in "$TARGET_DIR"/*.webm; do
    if [ ! -f "$webm_file" ]; then
        continue
    fi

    # Get just the filename (not full path) for processing
    filename=$(basename "$webm_file")

    # Check if already processed
    if is_processed "$filename"; then
        echo "⏭️  Skipping (already processed): $filename"
        ((skipped_count++))
        continue
    fi

    # Process the file
    if process_file "$filename"; then
        ((processed_count++))
    else
        echo "❌ Failed to process: $filename"
        ((error_count++))
    fi

    # Progress update
    total_done=$((processed_count + skipped_count + error_count))
    echo "Progress: $total_done/$total_files (Processed: $processed_count, Skipped: $skipped_count, Errors: $error_count)"
    echo ""

    # Rate limiting delay (skip on last file)
    if [ "$total_done" -lt "$total_files" ]; then
        echo "Waiting ${DELAY_SECONDS}s before next request..."
        sleep "$DELAY_SECONDS"
    fi
done

echo "===================="
echo "Processing complete!"
echo "Total files: $total_files"
echo "Newly processed: $processed_count"
echo "Already processed (skipped): $skipped_count"
echo "Errors: $error_count"
echo "CSV file: $CSV_FILE"
echo "===================="
'