app.py

# import os
# import subprocess

# # Install BrowserGym dependencies before running the main application
# def install_browsergym():
#     try:
#         print("Installing BrowserGym dependencies...")
#         subprocess.run("cd BrowserGym && make install", shell=True, check=True)
#         print("BrowserGym installation completed successfully")
#     except subprocess.CalledProcessError as e:
#         print(f"Error installing BrowserGym: {e}")
#         raise

# install_browsergym()

import os
import logging
import gradio as gr
import openai
import multiprocessing

from process_run import process_run

# Configure logging
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(levelname)s - %(message)s',
    handlers=[
        logging.StreamHandler(),
    ]
)
logger = logging.getLogger(__name__)
logger.setLevel('INFO')

# Set your OpenAI API key
openai.api_key = os.getenv("OPENAI_API_KEY")


# Example instructions to display
EXAMPLES = [
    "When did the solar system form? Find on wikipedia.",
    "Find the rating of Monopoly (1935) on boardgamegeek.com",
]

URL_EXAMPLES = [
    "about:blank",
    "https://www.wikipedia.org",
    "https://www.boardgamegeek.com"
]

def main():
    logger.info("Starting BrowserGym web agent")
    
    with gr.Blocks(title="WebShephered Demo") as demo:
        # Add CSS for outlined groups
        gr.Markdown("# WebShephered Demo")
        with gr.Row():
            with gr.Column(scale=2):
                with gr.Column():
                    instruction = gr.Textbox(
                        label="Instruction",
                        placeholder="Enter your instruction here",
                        lines=2,
                    )
                    gr.Examples(
                        examples=[[e] for e in EXAMPLES],
                        inputs=instruction,
                        cache_examples=False,
                    )
                
                gr.Markdown("\n\n")
                
                with gr.Column():
                    start_url = gr.Textbox(
                        label="Starting URL",
                        placeholder="URL to start the browser at",
                        value="about:blank"
                    )
                    gr.Examples(
                        examples=URL_EXAMPLES,
                        inputs=start_url,
                        cache_examples=False,
                    )

                gr.Markdown("\n\n")

                model_name = gr.Dropdown(
                    label="Agent Model",
                    choices=["gpt-4o"],
                    value="gpt-4o"
                )
                run_btn = gr.Button("Run Demo")

                gr.Markdown("---")
                
                with gr.Column():
                    gr.Markdown("## Current State")
                    state_view = gr.Markdown()
                    browser_view = gr.Image(label="Browser View")

                    gr.Markdown("### Task Checklist from WebShephered")
                    checklist_view = gr.Markdown()
                    
                    gr.Markdown("### Action Selection in current step")
                    with gr.Row() as rm_row:
                        rm_cards_container = gr.HTML()
            with gr.Column(scale=2):
                gr.Markdown("## Trajectory")
                trajectory_container = gr.HTML() # Placeholder for our custom trajectory component

        
            
        run_btn.click(
            fn=process_run,
            inputs=[instruction, model_name, start_url],
            outputs=[state_view, browser_view, checklist_view, rm_cards_container, trajectory_container],
            api_name="run_agent",
            concurrency_limit=32,
            show_progress=True
        )

    logger.info("Launching Gradio interface")
    # Set max_threads to allow multiple concurrent requests
    demo.launch(share=True, max_threads=32)

if __name__ == "__main__":
    # Add support for multiprocessing on Windows
    multiprocessing.freeze_support()
    main()