Update app.py

app.py

@@ -5,7 +5,6 @@ import gradio as gr
 import asyncio
 import logging
 import torch
-import random
 from serpapi import GoogleSearch
 from pydantic import BaseModel
 from autogen_agentchat.agents import AssistantAgent
@@ -89,8 +88,8 @@ def search_web(query: str, serpapi_key: str) -> str:
 def html_with_progress(label, progress):
     return f"""
     <div style="display: flex; flex-direction: column; justify-content: center; align-items: center; height: 100%; min-height: 700px; padding: 20px; text-align: center; border: 1px solid #ddd; border-radius: 8px;">
-        <div style="width: 100%; background-color: #FFFFFF; border-radius: 10px; overflow: hidden; margin-bottom: 20px;">
-            <div style="width: {progress}%; height: 30px; background-color: #4CAF50; border-radius: 10px;"></div>
+        <div style="width: 70%; background-color: #FFFFFF; border-radius: 80px; overflow: hidden; margin-bottom: 20px;">
+            <div style="width: {progress}%; height: 15px; background-color: #4CAF50; border-radius: 80px;"></div>
         </div>
         <h2 style="font-style: italic; color: #555;">{label}</h2>
     </div>
@@ -109,7 +108,7 @@ def get_model_client(service, api_key):
     else:
         raise ValueError("Invalid service")
 
-# Helper function to clean script text and make it natural
+# Helper function to clean script text
 def clean_script_text(script):
     if not script or not isinstance(script, str):
         logger.error("Invalid script input: %s", script)
@@ -121,20 +120,11 @@ def clean_script_text(script):
     script = script.replace("humanlike", "human-like").replace("problemsolving", "problem-solving")
     script = re.sub(r"\s+", " ", script).strip()
     
-    script = re.sub(r"^\s*-\s*", "So, ", script, flags=re.MULTILINE)
-    
-    non_verbal = ["um, ", "you know, ", "like, "]
-    words = script.split()
-    for i in range(len(words) - 1, -1, -1):
-        if random.random() < 0.1:
-            words.insert(i, random.choice(non_verbal))
-    script = " ".join(words)
-    
     if len(script) < 10:
         logger.error("Cleaned script too short (%d characters): %s", len(script), script)
         return None
     
-    logger.info("Cleaned and naturalized script: %s", script)
+    logger.info("Cleaned script: %s", script)
     return script
 
 # Helper function to validate and convert speaker audio
@@ -306,8 +296,8 @@ def generate_markdown_slides(slides, title, speaker="Prof. AI Feynman", date="Ap
             slide_number = i + 1
             content = slide['content']
             
-            # First and last slides have no header/footer
-            if i == 0 or i == len(slides) - 1:
+            # First slide has no header/footer, others have header and footer
+            if i == 0:
                 slide_md = f"""
 # {slide['title']}
 {content}
@@ -367,8 +357,7 @@ async def on_generate(api_service, api_key, serpapi_key, title, topic, instructi
 
     model_client = get_model_client(api_service, api_key)
     
-    actual_content_slides = num_slides
-    total_slides = actual_content_slides + 3  # Content slides + quiz, assignment, thank-you
+    total_slides = num_slides  # Use exactly the number of slides from input
     
     research_agent = AssistantAgent(
         name="research_agent",
@@ -382,15 +371,12 @@ async def on_generate(api_service, api_key, serpapi_key, title, topic, instructi
         model_client=model_client,
         handoffs=["script_agent"],
         system_message=f"""
-You are a Slide Agent. Using the research from the conversation history and the specified number of content slides ({actual_content_slides}), generate exactly {actual_content_slides} content slides, plus one quiz slide, one assignment slide, and one thank-you slide, for a total of {total_slides} slides. Output ONLY a JSON array wrapped in ```json ... ``` in a TextMessage, where each slide is an object with 'title' and 'content' keys. Do not include any explanatory text, comments, or other messages. Ensure the JSON is valid and contains exactly {total_slides} slides before proceeding. After outputting the JSON, use the handoff_to_script_agent tool to pass the task to the Script Agent.
-Example output for 2 content slides:
+You are a Slide Agent. Using the research from the conversation history and the specified number of slides ({total_slides}), generate exactly {total_slides} content slides. Output ONLY a JSON array wrapped in ```json ... ``` in a TextMessage, where each slide is an object with 'title' and 'content' keys. Do not include any explanatory text, comments, or other messages. Ensure the JSON is valid and contains exactly {total_slides} slides before proceeding. After outputting the JSON, use the handoff_to_script_agent tool to pass the task to the Script Agent.
+Example output for 2 slides:
 ```json
 [
     {{"title": "Slide 1", "content": "Content for slide 1"}},
-    {{"title": "Slide 2", "content": "Content for slide 2"}},
-    {{"title": "Quiz", "content": "Quiz questions"}},
-    {{"title": "Assignment", "content": "Assignment details"}},
-    {{"title": "Thank You", "content": "Thank you message"}}
+    {{"title": "Slide 2", "content": "Content for slide 2"}}
 ]
 ```""",
         output_content_type=None,
@@ -401,16 +387,13 @@ Example output for 2 content slides:
         model_client=model_client,
         handoffs=["feynman_agent"],
         system_message=f"""
-You are a Script Agent. Access the JSON array of {total_slides} slides from the conversation history. Generate a narration script (1-2 sentences) for each of the {total_slides} slides, summarizing its content in a natural, conversational tone as a speaker would, including occasional non-verbal words (e.g., "um," "you know," "like"). Output ONLY a JSON array wrapped in ```json ... ``` with exactly {total_slides} strings, one script per slide, in the same order. Ensure the JSON is valid and complete. After outputting, use the handoff_to_feynman_agent tool. If scripts cannot be generated, retry once.
-Example for 3 content slides:
+You are a Script Agent. Access the JSON array of {total_slides} slides from the conversation history. Generate a narration script (1-2 sentences) for each of the {total_slides} slides, summarizing its content in a clear, academically inclined tone as a professor would deliver it. Avoid using non-verbal fillers such as "um," "you know," or "like." Output ONLY a JSON array wrapped in ```json ... ``` with exactly {total_slides} strings, one script per slide, in the same order. Ensure the JSON is valid and complete. After outputting, use the handoff_to_feynman_agent tool. If scripts cannot be generated, retry once.
+Example for 3 slides:
 ```json
 [
-    "So, this slide, um, covers the main topic in a fun way.",
-    "The second slide introduces the key concepts.",
-    "This third slide shows some interesting applications.",
-    "Alright, you know, answer these quiz questions.",
-    "Here's your, like, assignment to complete.",
-    "Thanks for, um, attending today!"
+    "Hello everyone, welcome to Agents 101. I am Jaward, your primary instructor for this course.",
+    "Today, we will cover the syllabus for this semester, providing a gentle introduction to AI agents.",
+    "Let us define what an AI agent is: it refers to a system or program capable of autonomously performing tasks on behalf of a user or another system."
 ]
 ```""",
         output_content_type=None,
@@ -440,10 +423,10 @@ Example: 'Received {total_slides} slides and {total_slides} scripts. Lecture is
     Topic: {topic}
     Additional Instructions: {instructions}
     Audience: {lecture_type}
-    Number of Content Slides: {actual_content_slides}
+    Number of Slides: {total_slides}
     Please start by researching the topic.
     """
-    logger.info("Starting lecture generation for topic: %s with %d content slides", topic, actual_content_slides)
+    logger.info("Starting lecture generation for topic: %s with %d slides", topic, total_slides)
     
     slides = None
     scripts = None
@@ -486,7 +469,7 @@ Example: 'Received {total_slides} slides and {total_slides} scripts. Lecture is
                             slide_retry_count += 1
                             logger.info("Retrying slide generation (attempt %d/%d)", slide_retry_count, max_retries)
                             retry_message = TextMessage(
-                                content=f"Please generate exactly {total_slides} slides ({actual_content_slides} content slides plus quiz, assignment, thank-you) as per your instructions.",
+                                content=f"Please generate exactly {total_slides} slides as per your instructions.",
                                 source="user",
                                 recipient="slide_agent"
                             )
@@ -526,7 +509,7 @@ Example: 'Received {total_slides} slides and {total_slides} scripts. Lecture is
                             slide_retry_count += 1
                             logger.info("Retrying slide generation (attempt %d/%d)", slide_retry_count, max_retries)
                             retry_message = TextMessage(
-                                content=f"Please generate exactly {total_slides} slides ({actual_content_slides} content slides plus quiz, assignment, thank-you) as per your instructions.",
+                                content=f"Please generate exactly {total_slides} slides as per your instructions.",
                                 source="user",
                                 recipient="slide_agent"
                             )
@@ -550,7 +533,7 @@ Example: 'Received {total_slides} slides and {total_slides} scripts. Lecture is
                         slide_retry_count += 1
                         logger.info("Retrying slide generation (attempt %d/%d)", slide_retry_count, max_retries)
                         retry_message = TextMessage(
-                            content=f"Please generate exactly {total_slides} slides ({actual_content_slides} content slides plus quiz, assignment, thank-you) as per your instructions.",
+                            content=f"Please generate exactly {total_slides} slides as per your instructions.",
                             source="user",
                             recipient="slide_agent"
                         )
@@ -614,11 +597,11 @@ Example: 'Received {total_slides} slides and {total_slides} scripts. Lecture is
             return
         
         if len(slides) != total_slides:
-            logger.error("Expected %d slides (including %d content slides + 3), but received %d", total_slides, actual_content_slides, len(slides))
+            logger.error("Expected %d slides, but received %d", total_slides, len(slides))
             yield f"""
             <div style="display: flex; flex-direction: column; justify-content: center; align-items: center; height: 100%; min-height: 700px; padding: 20px; text-align: center; border: 1px solid #ddd; border-radius: 8px;">
                 <h2 style="color: #d9534f;">Incorrect number of slides</h2>
-                <p style="margin-top: 20px;">Expected {total_slides} slides ({actual_content_slides} content slides + quiz, assignment, thank-you), but generated {len(slides)}. Please try again.</p>
+                <p style="margin-top: 20px;">Expected {total_slides} slides, but generated {len(slides)}. Please try again.</p>
             </div>
             """
             return
@@ -725,15 +708,15 @@ Example: 'Received {total_slides} slides and {total_slides} scripts. Lecture is
         txt_links = ""
         for txt_file in txt_files:
             file_path = os.path.join(OUTPUT_DIR, txt_file)
-            txt_links += f'<a href="file/{file_path}" download>{txt_file}</a>&nbsp;&nbsp;'
+            txt_links += f'<a href="file/{file_path}" download>{txt_file}</a>  '
         
         # Generate audio timeline
         audio_timeline = ""
         for i, audio_file in enumerate(audio_files):
             if audio_file:
-                audio_timeline += f'<span id="audio-{i+1}">{os.path.basename(audio_file)}</span>&nbsp;&nbsp;'
+                audio_timeline += f'<span id="audio-{i+1}">{os.path.basename(audio_file)}</span>  '
             else:
-                audio_timeline += f'<span id="audio-{i+1}">slide_{i+1}.mp3</span>&nbsp;&nbsp;'
+                audio_timeline += f'<span id="audio-{i+1}">slide_{i+1}.mp3</span>  '
         
         slides_info = json.dumps({"slides": markdown_slides, "audioFiles": audio_files})
         
@@ -887,7 +870,7 @@ with gr.Blocks(title="Agent Feynman") as demo:
                 )
                 api_key = gr.Textbox(label="Model Provider API Key", type="password", placeholder="Not required for Ollama")
                 serpapi_key = gr.Textbox(label="SerpApi Key", type="password", placeholder="Enter your SerpApi key")
-                num_slides = gr.Slider(1, 20, step=1, label="Number of Content Slides", value=3)
+                num_slides = gr.Slider(1, 20, step=1, label="Number of Slides", value=3)
                 speaker_audio = gr.Audio(label="Speaker sample audio (MP3 or WAV)", type="filepath", elem_id="speaker-audio")
                 generate_btn = gr.Button("Generate Lecture")
         with gr.Column(scale=2):