151 lines
4.9 KiB
Python
151 lines
4.9 KiB
Python
import sys
|
|
import argparse
|
|
import json
|
|
import asyncio
|
|
from utils import disclaim_bool_string, prepare_output_folder, create_folder
|
|
from datetime import datetime, timezone
|
|
from dotenv import load_dotenv, find_dotenv
|
|
import os
|
|
|
|
import warnings
|
|
warnings.filterwarnings("ignore")
|
|
|
|
exception_msg = "Exception: %s"
|
|
|
|
from image_extractor import ImageExtractor
|
|
from mllm_management import MLLMManager
|
|
|
|
async def cli(sys_argv):
|
|
parser = argparse.ArgumentParser()
|
|
|
|
parser.add_argument(
|
|
"--page_url",
|
|
type=str,
|
|
help=("Url page to analyze"),
|
|
default="https://www.bbc.com",
|
|
)
|
|
|
|
parser.add_argument(
|
|
"--context_levels",
|
|
type=int,
|
|
default=5,
|
|
help=("HTML context levels around the image"),
|
|
)
|
|
parser.add_argument(
|
|
"--pixel_distance_threshold",
|
|
type=int,
|
|
default=200,
|
|
help=("pixel distance threshold around the image"),
|
|
)
|
|
parser.add_argument(
|
|
"--number_of_images",
|
|
type=int,
|
|
default=10,
|
|
help=("max number of desired images"),
|
|
)
|
|
|
|
parser.add_argument(
|
|
"--save_elaboration",
|
|
action="store_true",
|
|
default=True,
|
|
help=("If True save the elaborated info in a json file"),
|
|
)
|
|
|
|
parser.add_argument(
|
|
"--save_images",
|
|
action="store_true",
|
|
default=True,
|
|
help=("If True save the images"),
|
|
)
|
|
|
|
args = parser.parse_args()
|
|
|
|
print("wcag validator args:",args)
|
|
|
|
if (
|
|
disclaim_bool_string(args.save_elaboration) == True
|
|
or disclaim_bool_string(args.save_images) == True
|
|
): # if something to save
|
|
url_path = args.page_url.replace(":", "").replace("//", "_").replace("/", "_")
|
|
now = datetime.now(timezone.utc)
|
|
now_str = now.strftime("%Y_%m_%d-%H_%M_%S")
|
|
output_dir = prepare_output_folder(url_path, now_str)
|
|
|
|
if disclaim_bool_string(args.save_images) == True:
|
|
images_output_dir = create_folder(
|
|
output_dir, directory_separator="/", next_path="images"
|
|
)
|
|
print("save images path:", images_output_dir)
|
|
|
|
### Task #1: ---------- Image Extractor
|
|
# Create extractor
|
|
image_extractor = ImageExtractor(
|
|
args.page_url,
|
|
context_levels=args.context_levels,
|
|
pixel_distance_threshold=args.pixel_distance_threshold,
|
|
number_of_images=args.number_of_images,
|
|
save_images=args.save_images,
|
|
save_images_path=images_output_dir,
|
|
)
|
|
# Extract images
|
|
print(f"Extracting images from: {args.page_url}")
|
|
images = await image_extractor.extract_images()
|
|
if disclaim_bool_string(args.save_elaboration) == True: # Optionally save to JSON
|
|
|
|
await image_extractor.save_elaboration(images,output_dir=output_dir + "/extracted_images.json")
|
|
#---------------------------------------------
|
|
|
|
### Task #2: ---------- MLLM management
|
|
env_path = find_dotenv(filename=".env")
|
|
_ = load_dotenv(env_path) # read .env file
|
|
|
|
mllm_end_point=os.getenv("mllm_end_point")
|
|
mllm_api_key=os.getenv("mllm_api_key")
|
|
print("mllm_end_point:",mllm_end_point)
|
|
|
|
mllm_manager = MLLMManager(mllm_end_point, mllm_api_key)
|
|
print("mllm_manager.end_point:", mllm_manager.end_point)
|
|
|
|
alt_text_system_prompt = mllm_manager.get_alt_text_system_prompt()
|
|
print("alt_text_system_prompt:", alt_text_system_prompt)
|
|
|
|
|
|
mllm_responses = []
|
|
for img_info in images:
|
|
alt_text="Here is the alt-text of the image: " + img_info["alt_text"]
|
|
image_URL=img_info["url"]
|
|
HTML_context = "Here is the surrounding HTML context of the element: " + img_info["html_context"]
|
|
page_text = "Here is the content of the page: Title of the page: " + str(img_info["page_title"])
|
|
page_text=page_text+", content of the <meta name='description'> tag: " + str(img_info["page_description"])
|
|
page_text=page_text+", content of the <meta name='keywords'> tag: " + str(img_info["page_keywords"])
|
|
# skip headings
|
|
|
|
print("Processing image URL:", image_URL)
|
|
print("Alt-text:", alt_text)
|
|
print("HTML context:", HTML_context)
|
|
print("Page text:", page_text)
|
|
|
|
alt_text_user_prompt=mllm_manager.get_alt_text_user_prompt(
|
|
altTextMessage=alt_text,
|
|
imageURL=image_URL,
|
|
HTMLcontext=HTML_context,
|
|
pageText=page_text,
|
|
)
|
|
print("alt_text_user_prompt:", alt_text_user_prompt)
|
|
|
|
mllm_response = mllm_manager.get_response(
|
|
system_prompt=alt_text_system_prompt,
|
|
user_prompt=alt_text_user_prompt
|
|
|
|
)
|
|
mllm_responses.append(mllm_response['choices'][0]["message"]["content"])
|
|
|
|
if disclaim_bool_string(args.save_elaboration) == True: # Optionally save to JSON
|
|
|
|
with open(output_dir + "/mllm_responses.json", "w", encoding="utf-8") as f:
|
|
json.dump(mllm_responses, f, indent=2, ensure_ascii=False)
|
|
|
|
|
|
if __name__ == '__main__':
|
|
|
|
asyncio.run(cli(sys.argv[1:])) |