Update app.py

app.py

@@ -1,57 +1,43 @@
-# app.py
+
 
 # AI Agent Framework Imports
 from smolagents import CodeAgent, HfApiModel, tool, load_tool
 
-# Web Scraping and Parsing
-import requests
-from bs4 import BeautifulSoup
-
 # Selenium & webdriver-manager for Browser Automation
 from selenium import webdriver
 from selenium.webdriver.chrome.options import Options
 from selenium.webdriver.chrome.service import Service
 from selenium.webdriver.common.by import By
 from webdriver_manager.chrome import ChromeDriverManager
+import re
+import time
+
+# Standard Library Imports (allowed)
 import time
-import os
-
-# NLP and Text Processing (Optional for Future Enhancements)
-import nltk
-from nltk.tokenize import sent_tokenize
-import spacy
-
-# Code Parsing and Explanation
-import ast
-from pygments import highlight
-from pygments.lexers import PythonLexer
-from pygments.formatters import HtmlFormatter
-
-# UI Overlay & Interaction
-import gradio as gr
-import cv2  # Optional: for visual highlights
 import yaml
 
 # Final Answer and UI Handling
 from tools.final_answer import FinalAnswerTool
 from Gradio_UI import GradioUI
 
-# Helper function to automatically obtain and configure the Selenium driver
+# --------------------------------------------
+# Helper function to obtain and configure the Selenium driver automatically.
 def get_driver():
     options = Options()
-    options.add_argument("--headless")  # Run browser in headless mode
+    options.add_argument("--headless")            # Run in headless mode
     options.add_argument("--disable-gpu")
     options.add_argument("--no-sandbox")
     options.add_argument("--disable-dev-shm-usage")
-    # Automatically install and use the appropriate ChromeDriver
+    # Use webdriver-manager to install the proper ChromeDriver
     service = Service(ChromeDriverManager().install())
     driver = webdriver.Chrome(service=service, options=options)
     return driver
 
-# Tool to analyze a webpage or document and simplify its content
+# --------------------------------------------
+# Tool to analyze a webpage or document and provide a simplified summary.
 @tool
 def analyze_and_simplify(url: str) -> str:
-    """Fetches webpage content, extracts text, and provides a simplified summary.
+    """Fetches webpage content using Selenium, strips HTML tags, and provides a simplified summary.
     
     Args:
         url: The webpage URL to analyze.
@@ -60,17 +46,23 @@ def analyze_and_simplify(url: str) -> str:
         driver = get_driver()
         driver.get(url)
         time.sleep(5)  # Wait for content to load
-        soup = BeautifulSoup(driver.page_source, "html.parser")
+        
+        # Fetch page source and use regex to strip HTML tags.
+        page_source = driver.page_source
         driver.quit()
-        text = ' '.join([p.get_text() for p in soup.find_all("p")])
+        # Remove HTML tags (very simple approach)
+        text = re.sub(r'<[^>]+>', '', page_source)
+        text = text.strip()
         if not text:
             return "No readable text found on the page."
-        summary = f"Simplified Summary: {text[:500]}... (summary continues)"
+        # Return a truncated summary of the text.
+        summary = f"Simplified Summary: {text[:500]}..."
         return summary
     except Exception as e:
         return f"Error analyzing webpage: {str(e)}"
 
-# Tool to detect ambiguous directions and suggest clarification
+# --------------------------------------------
+# Tool to detect ambiguous instructions.
 @tool
 def detect_ambiguity(content: str) -> str:
     """Checks for vague instructions and suggests clarifications.
@@ -78,32 +70,34 @@ def detect_ambiguity(content: str) -> str:
     Args:
         content: Text to analyze.
     """
-    # Placeholder logic for ambiguity detection
+    # Placeholder logic for ambiguity detection.
     return "Ambiguity detected. Click 'Is this ambiguous?' for help."
 
-# Tool to explain technical terms or assumed knowledge in simple language
+# --------------------------------------------
+# Tool to explain technical terms in simple language.
 @tool
 def explain_assumed_knowledge(term: str) -> str:
-    """Defines difficult or technical terms in a simple way.
+    """Defines technical terms in a simple way.
     
     Args:
         term: The term to explain.
     """
     return f"Definition of '{term}': [Detailed beginner-friendly explanation here]"
 
-# Tool to highlight a UI element and optionally automate its interaction
+# --------------------------------------------
+# Tool to highlight a UI element and optionally perform a click.
 @tool
 def highlight_elements(step: str, element: str, auto_execute: bool = False) -> str:
     """Highlights a UI element and optionally performs an action.
     
     Args:
         step: The current step in the guide.
-        element: The UI element to highlight (expected as an XPath).
-        auto_execute: If True, the agent will click the element automatically.
+        element: The UI element to highlight (as an XPath).
+        auto_execute: If True, the agent will automatically click the element.
     """
     try:
         driver = get_driver()
-        # Navigate to a placeholder page; replace "about:blank" with the actual URL if needed
+        # Here we simply navigate to a blank page (or a given URL if needed).
         driver.get("about:blank")
         if auto_execute:
             elem = driver.find_element(By.XPATH, element)
@@ -116,17 +110,19 @@ def highlight_elements(step: str, element: str, auto_execute: bool = False) -> s
     except Exception as e:
         return f"Error in highlight_elements: {str(e)}"
 
-# Tool to explain a single line of code in simple terms
+# --------------------------------------------
+# Tool to explain a single line of code.
 @tool
 def explain_code_line(line: str) -> str:
     """Explains what a line of code does in simple terms.
     
     Args:
-        line: The code to explain.
+        line: The code line to explain.
     """
     return f"Explanation for: {line} [Insert explanation here]"
 
-# Tool to provide an interactive teacher box for user queries
+# --------------------------------------------
+# Tool to provide an interactive teacher box for user queries.
 @tool
 def teacher_box_query(question: str) -> str:
     """Allows users to ask the AI questions while browsing.
@@ -136,7 +132,8 @@ def teacher_box_query(question: str) -> str:
     """
     return f"AI Answer: [Response for '{question}']"
 
-# Tool to toggle between manual and automatic navigation modes
+# --------------------------------------------
+# Tool to toggle between manual and automatic navigation modes.
 @tool
 def toggle_auto_execution(enable: bool) -> str:
     """Lets the user turn automatic navigation on or off.
@@ -146,19 +143,20 @@ def toggle_auto_execution(enable: bool) -> str:
     """
     return "Auto-execution enabled." if enable else "Manual step-by-step mode enabled."
 
-# Configure the AI Model
+# --------------------------------------------
+# Configure the AI Model.
 model = HfApiModel(
-    max_tokens=2096,       # Maximum response length
-    temperature=0.5,       # Controls response randomness
-    model_id='Qwen/Qwen2.5-Coder-32B-Instruct',  # Selected model
+    max_tokens=2096,                       # Maximum response length.
+    temperature=0.5,                       # Controls response randomness.
+    model_id='Qwen/Qwen2.5-Coder-32B-Instruct',  # Selected model.
     custom_role_conversions=None
 )
 
-# Load prompt templates from a YAML configuration file
+# Load prompt templates from a YAML configuration file.
 with open("prompts.yaml", 'r') as stream:
     prompt_templates = yaml.safe_load(stream)
 
-# Create the AI agent with the defined tools
+# Create the AI agent with the defined tools.
 agent = CodeAgent(
     model=model,
     tools=[
@@ -171,10 +169,11 @@ agent = CodeAgent(
         teacher_box_query,
         toggle_auto_execution
     ],
-    max_steps=6,             # Maximum number of reasoning steps
-    verbosity_level=1,       # Level of detail in agent responses
+    max_steps=6,             # Maximum number of reasoning steps.
+    verbosity_level=1,       # Level of detail in agent responses.
     prompt_templates=prompt_templates
 )
 
-# Launch the interactive UI using Gradio
+# Launch the interactive UI using Gradio.
 GradioUI(agent).launch()
+