import os import json import argparse import hashlib import requests import time from safetensors import safe_open from tqdm import tqdm MAX_RETRIES = 3 BACKOFF_BASE = 2 def retry_request(url, timeout=10, desc=""): attempt = 0 while attempt < MAX_RETRIES: try: response = requests.get(url, timeout=timeout) if response.status_code == 200: return response elif response.status_code == 429: tqdm.write(f"{desc} - Rate limited (HTTP 429). Retrying ({attempt + 1}/{MAX_RETRIES})...") else: tqdm.write(f"{desc} - HTTP {response.status_code}. Retrying ({attempt + 1}/{MAX_RETRIES})...") except requests.exceptions.RequestException as e: tqdm.write(f"{desc} - Error: {e}. Retrying ({attempt + 1}/{MAX_RETRIES})...") attempt += 1 time.sleep(BACKOFF_BASE * (2 ** (attempt - 1))) tqdm.write(f"{desc} - Failed after {MAX_RETRIES} attempts.") return None def compute_sha256(file_path): sha256_hash = hashlib.sha256() with open(file_path, "rb") as f: for byte_block in iter(lambda: f.read(4096), b""): sha256_hash.update(byte_block) return sha256_hash.hexdigest() def extract_local_metadata(safetensors_file): try: with safe_open(safetensors_file, framework="pt") as f: metadata = f.metadata() return metadata except Exception as e: tqdm.write(f"Error reading {safetensors_file}: {e}") return None def fetch_civitai_metadata_by_model_id(model_id): url = f"https://civitai.com/api/v1/models/{model_id}" response = retry_request(url, desc=f"Fetching metadata for model ID {model_id}") if response and response.status_code == 200: return response.json() return None def fetch_civitai_metadata_by_hash(file_hash): url = f"https://civitai.com/api/v1/model-versions/by-hash/{file_hash}" response = retry_request(url, desc=f"Fetching metadata by hash {file_hash[:10]}...") if response and response.status_code == 200: return response.json() return None def download_preview_images(images_list, save_dir, base_filename, delay=0.5, image_pbar=None, use_subdir=True): if use_subdir: subdir = os.path.join(save_dir, f"{base_filename}_previews") os.makedirs(subdir, exist_ok=True) else: subdir = save_dir # save in same folder for idx, img_data in enumerate(images_list): url = img_data.get('url') if not url: continue ext = os.path.splitext(url)[1].split('?')[0] img_name = f"{base_filename}_preview_{idx+1}{ext}" # Paths to check (both flat + subdir) flat_img_path = os.path.join(save_dir, img_name) subdir_img_path = os.path.join(subdir, img_name) if os.path.exists(flat_img_path) or os.path.exists(subdir_img_path): tqdm.write(f"Preview already exists: {img_name} (skipping)") if image_pbar: image_pbar.update(1) continue desc = f"Downloading media {idx + 1}/{len(images_list)}" response = retry_request(url, desc=desc) if response and response.status_code == 200: img_path = subdir_img_path if use_subdir else flat_img_path with open(img_path, 'wb') as img_file: img_file.write(response.content) tqdm.write(f"Saved preview: {img_path}") if image_pbar: image_pbar.update(1) time.sleep(delay) def process_directory(root_dir, force=False, scrape_civitai=False, delay=0.5, previews_subdir=True, max_media=None): safetensors_files = [] for dirpath, dirnames, filenames in os.walk(root_dir): for filename in filenames: if filename.endswith(".safetensors"): safetensors_files.append(os.path.join(dirpath, filename)) print(f"\nFound {len(safetensors_files)} .safetensors files.\n") with tqdm(total=len(safetensors_files), desc="Total Progress", unit="file") as total_pbar: for safetensors_path in safetensors_files: dirpath = os.path.dirname(safetensors_path) filename = os.path.basename(safetensors_path) base_filename = os.path.splitext(filename)[0] json_filename = f"{base_filename}.metadata.json" json_path = os.path.join(dirpath, json_filename) if os.path.exists(json_path) and not force: tqdm.write(f"Skipping (metadata exists): {safetensors_path}") total_pbar.update(1) continue tqdm.write(f"\nProcessing: {safetensors_path}") metadata = extract_local_metadata(safetensors_path) combined_metadata = {'local_metadata': metadata if metadata else {}} civitai_data = None if scrape_civitai: civitai_model_id = None if metadata: if 'ss_civitai_model_id' in metadata: civitai_model_id = metadata['ss_civitai_model_id'] elif 'ss_civitai_url' in metadata: parts = metadata['ss_civitai_url'].split('/') civitai_model_id = next((part for part in parts if part.isdigit()), None) if civitai_model_id: tqdm.write(f"Found model ID in metadata: {civitai_model_id}") civitai_data = fetch_civitai_metadata_by_model_id(civitai_model_id) time.sleep(delay) else: tqdm.write("No CivitAI model ID found in metadata. Trying hash lookup...") file_hash = compute_sha256(safetensors_path) civitai_data = fetch_civitai_metadata_by_hash(file_hash) time.sleep(delay) if civitai_data: civitai_meta = { 'civitai_model_id': civitai_data.get('modelId') or civitai_data.get('id'), 'civitai_model_version_id': civitai_data.get('id'), 'civitai_name': civitai_data.get('name'), 'description': civitai_data.get('description'), 'tags': civitai_data.get('tags'), 'trainedWords': civitai_data.get('trainedWords'), 'images': civitai_data.get('images') } combined_metadata['civitai_metadata'] = civitai_meta images_list = civitai_meta.get('images', []) if images_list: # Apply max_media logic if max_media == 0: tqdm.write("Skipping download of preview images/videos (user selected 0).") else: if max_media is not None: images_list = images_list[:max_media] with tqdm(total=len(images_list), desc="Image/Video Progress", leave=False) as image_pbar: download_preview_images( images_list, dirpath, base_filename, delay=delay, image_pbar=image_pbar, use_subdir=previews_subdir ) else: tqdm.write("No preview images/videos found.") else: tqdm.write("No CivitAI data found (model ID or hash lookup failed).") with open(json_path, "w", encoding="utf-8") as f: json.dump(combined_metadata, f, indent=4, ensure_ascii=False) tqdm.write(f"Saved metadata to: {json_path}") total_pbar.update(1) def interactive_menu(): print("\n=== LoRA Metadata Scraper Config ===\n") scrape_civitai = input("A) Scrape CivitAI? (Y/N) [Default: N]: ").strip().lower() == 'y' delay_choice = input("B) Use default delay (0.5s), no delay (0), or custom? (D/N/C) [Default: D]: ").strip().lower() if delay_choice == 'n': delay = 0.0 elif delay_choice == 'c': delay = float(input("Enter delay in seconds (e.g., 0.5): ").strip()) else: delay = 0.5 force = input("C) Force re-scrape if metadata exists? (Y/N) [Default: N]: ").strip().lower() == 'y' loras_only = input("D) Scan only the LoRAs folder? (Y/N) [Default: Y]: ").strip().lower() != 'n' previews_subdir = input("E) Save preview images in a subdirectory? (Y/N) [Default: Y]: ").strip().lower() != 'n' media_choice = input("F) How many preview images/videos to download? (A=All [default], N=None, X=Number): ").strip().lower() if media_choice == 'n': max_media = 0 elif media_choice == 'a' or media_choice == '': max_media = None else: try: max_media = int(media_choice) except ValueError: max_media = None # fallback to all print("\n=== Starting with your selected options ===\n") return force, scrape_civitai, delay, loras_only, previews_subdir, max_media if __name__ == "__main__": print(">>> Script started") parser = argparse.ArgumentParser(description="Scrape and save metadata for .safetensors files.") parser.add_argument("--force", action="store_true", help="Force re-scrape even if metadata file exists.") parser.add_argument("--scrape-civitai", action="store_true", help="Enable scraping CivitAI metadata + images.") parser.add_argument("--delay", type=float, default=0.5, help="Delay time (seconds) between API/image steps (default: 0.5s).") parser.add_argument("--interactive", action="store_true", help="Run in interactive mode.") parser.add_argument("--loras-only", action="store_true", help="Scan only the LoRAs folder (models/loras).") parser.add_argument("--previews-subdir", dest="previews_subdir", action="store_true", help="Save preview images in a subdirectory.") parser.add_argument("--no-previews-subdir", dest="previews_subdir", action="store_false", help="Save preview images in the same folder.") parser.add_argument("--max-media", type=int, default=None, help="Max number of preview images/videos to download (0 = none).") parser.set_defaults(previews_subdir=True) args = parser.parse_args() if args.interactive: force, scrape_civitai, delay, loras_only, previews_subdir, max_media = interactive_menu() else: force, scrape_civitai, delay, loras_only, previews_subdir, max_media = ( args.force, args.scrape_civitai, args.delay, args.loras_only, args.previews_subdir, args.max_media ) script_dir = os.path.dirname(os.path.abspath(__file__)) if loras_only: comfyui_dir = os.path.abspath(os.path.join(script_dir, "..", "..", "models", "loras")) else: comfyui_dir = os.path.abspath(os.path.join(script_dir, "..", "..")) tqdm.write(f"Scanning directory: {comfyui_dir}") process_directory( comfyui_dir, force=force, scrape_civitai=scrape_civitai, delay=delay, previews_subdir=previews_subdir, max_media=max_media )