visualize stage logs

app.py

@@ -1,6 +1,6 @@
 # Lectūra Research Demo: A Multi-Agent Tool for Self-taught Mastery.
 # Author: Jaward Sesay
-# © Lectūra Labs. All rights reserved. 
+# © Lectūra Labs 2025. All rights reserved. 
 import os
 import json
 import re
@@ -170,14 +170,48 @@ def create_slides(slides: list[dict], title: str, instructor_name: str, output_d
         logger.error("Failed to create HTML slides: %s", str(e))
         return []
 
-# progress bar
+# Track stage outputs
+stage_outputs = {}
+
+def add_stage_output(stage_name, output):
+    if stage_name not in stage_outputs:
+        stage_outputs[stage_name] = []
+    stage_outputs[stage_name].append(output)
+
+# Dynamic progress bar with accordion
 def html_with_progress(label, progress):
+    accordion_html = ""
+    if stage_outputs:
+        accordion_html = """
+        <div style="width: 70%; margin-top: 20px; border: 1px solid #ddd; border-radius: 25px; overflow: hidden;">
+        """
+        for stage_name, outputs in stage_outputs.items():
+            accordion_html += f"""
+            <div style="border-bottom: 1px solid #ddd;">
+                <div style="padding: 10px; background-color: #f8f9fa; cursor: pointer;" onclick="this.nextElementSibling.style.display = this.nextElementSibling.style.display === 'none' ? 'block' : 'none'">
+                    <h4 style="margin: 0; color: #555;">{stage_name}</h4>
+                </div>
+                <div style="padding: 10px; display: none; background-color: white;">
+            """
+            for output in outputs:
+                accordion_html += f"""
+                    <div style="margin-bottom: 10px; padding: 10px; background-color: #f8f9fa; border-radius: 4px;">
+                        <pre style="margin: 0; white-space: pre-wrap; word-wrap: break-word;">{output}</pre>
+                    </div>
+                """
+            accordion_html += """
+                </div>
+            </div>
+            """
+        accordion_html += "</div>"
+
     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: 70%; background-color: lightgrey; 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 !important;">{label}</h2>
+        {accordion_html}
     </div>
     """
 
@@ -280,7 +314,7 @@ def generate_xtts_audio(tts, text, speaker_wav, output_path):
         return False
     try:
         tts.tts_to_file(text=text, speaker_wav=speaker_wav, language="en", file_path=output_path)
-        logger.info("Generated audio for %s", output_path)
+        logger.info("Generated speech for %s", output_path)
         return True
     except Exception as e:
         logger.error("Failed to generate audio for %s: %s", output_path, str(e))
@@ -418,7 +452,15 @@ def get_gradio_file_url(local_path):
 # Async generate lecture materials and audio
 async def on_generate(api_service, api_key, serpapi_key, title, lecture_content_description, lecture_type, lecture_style, speaker_audio, num_slides):
     print(f"Received serpapi_key: '{serpapi_key}' (type: {type(serpapi_key)}, length: {len(serpapi_key) if serpapi_key else 0})")
-
+    
+    # Reset stage outputs at the start of generation
+    global stage_outputs
+    stage_outputs = {}
+    
+    # Define constants
+    max_audio_retries = 5
+    max_retries = 3
+    
     model_client = get_model_client(api_service, api_key)
     
     # Get the speaker from the speaker_audio path
@@ -576,6 +618,10 @@ Example: 'Received {total_slides} slides, {total_slides} scripts, and HTML files
     
     progress = 0
     label = "Researching lecture topic..."
+    if not serpapi_key:
+        add_stage_output("Research", "No API key for research was provided, proceeding with slide generation.")
+    else:
+        add_stage_output("Research", "Starting research phase...")
     yield (
         html_with_progress(label, progress),
         []
@@ -624,6 +670,8 @@ Example: 'Received {total_slides} slides, {total_slides} scripts, and HTML files
                 if source == "research_agent" and message.target == "slide_agent":
                     progress = 25
                     label = "Slides: generating..."
+                    if hasattr(message, 'content'):
+                        add_stage_output("Research", message.content)
                     yield (
                         html_with_progress(label, progress),
                         []
@@ -636,6 +684,7 @@ Example: 'Received {total_slides} slides, {total_slides} scripts, and HTML files
                         if extracted_json:
                             slides = extracted_json
                             logger.info("Extracted slides JSON from HandoffMessage context: %s", slides)
+                            add_stage_output("Slides", json.dumps(slides, indent=2))
                     if slides is None or len(slides) != total_slides:
                         if slide_retry_count < max_retries:
                             slide_retry_count += 1
@@ -661,6 +710,7 @@ Example: 'Received {total_slides} slides, {total_slides} scripts, and HTML files
                         if extracted_json:
                             scripts = extracted_json
                             logger.info("Extracted scripts JSON from HandoffMessage context: %s", scripts)
+                            add_stage_output("Scripts", json.dumps(scripts, indent=2))
                     progress = 75
                     label = "Review: in progress..."
                     yield (
@@ -669,22 +719,13 @@ Example: 'Received {total_slides} slides, {total_slides} scripts, and HTML files
                     )
                     await asyncio.sleep(0.1)
             
-            elif source == "research_agent" and isinstance(message, TextMessage) and "handoff_to_slide_agent" in message.content:
-                logger.info("Research Agent completed research")
-                progress = 25
-                label = "Research complete. Generating slides..."
-                yield (
-                    html_with_progress(label, progress),
-                    []
-                )
-                await asyncio.sleep(0.1)
-            
             elif source == "slide_agent" and isinstance(message, (TextMessage, StructuredMessage)):
                 logger.debug("Slide Agent message received")
                 extracted_json = extract_json_from_message(message)
                 if extracted_json:
                     slides = extracted_json
                     logger.info("Slide Agent generated %d slides: %s", len(slides), slides)
+                    add_stage_output("Slides", json.dumps(slides, indent=2))
                     if len(slides) != total_slides:
                         if slide_retry_count < max_retries:
                             slide_retry_count += 1
@@ -700,6 +741,9 @@ Example: 'Received {total_slides} slides, {total_slides} scripts, and HTML files
                     html_files = create_slides(slides, title, instructor_name)
                     if not html_files:
                         logger.error("Failed to generate HTML slides")
+                        add_stage_output("Slides", "Failed to generate HTML slides")
+                    else:
+                        add_stage_output("Slides", f"Successfully generated {len(html_files)} HTML slides")
                     progress = 50
                     label = "Scripts: generating..."
                     yield (
@@ -726,14 +770,17 @@ Example: 'Received {total_slides} slides, {total_slides} scripts, and HTML files
                 if extracted_json:
                     scripts = extracted_json
                     logger.info("Script Agent generated scripts for %d slides: %s", len(scripts), scripts)
+                    add_stage_output("Scripts", json.dumps(scripts, indent=2))
                     for i, script in enumerate(scripts):
                         script_file = os.path.join(OUTPUT_DIR, f"slide_{i+1}_script.txt")
                         try:
                             with open(script_file, "w", encoding="utf-8") as f:
                                 f.write(script)
                             logger.info("Saved script to %s", script_file)
+                            add_stage_output("Scripts", f"Saved script for slide {i+1}")
                         except Exception as e:
                             logger.error("Error saving script to %s: %s", script_file, str(e))
+                            add_stage_output("Scripts", f"Failed to save script for slide {i+1}: {str(e)}")
                     progress = 75
                     label = "Script complete. Reviewing lecture materials..."
                     yield (
@@ -758,6 +805,7 @@ Example: 'Received {total_slides} slides, {total_slides} scripts, and HTML files
                 logger.info("Instructor Agent completed lecture review: %s", message.content)
                 progress = 90
                 label = "Lecture materials ready. Generating lecture speech..."
+                add_stage_output("Review", message.content)
                 file_paths = [f for f in os.listdir(OUTPUT_DIR) if f.endswith(('.md', '.txt'))]
                 file_paths.sort()
                 file_paths = [os.path.join(OUTPUT_DIR, f) for f in file_paths]
@@ -852,32 +900,8 @@ Example: 'Received {total_slides} slides, {total_slides} scripts, and HTML files
         
         for i, script in enumerate(scripts):
             cleaned_script = clean_script_text(script)
-            audio_file = os.path.join(OUTPUT_DIR, f"slide_{i+1}.mp3")
-            script_file = os.path.join(OUTPUT_DIR, f"slide_{i+1}_script.txt")
-            
-            try:
-                with open(script_file, "w", encoding="utf-8") as f:
-                    f.write(cleaned_script or "")
-                logger.info("Saved script to %s: %s", script_file, cleaned_script)
-            except Exception as e:
-                logger.error("Error saving script to %s: %s", 
-                script_file, str(e))
+            audio_file = os.path.join(OUTPUT_DIR, f"slide_{i+1}_audio.mp3")
             
-            if not cleaned_script:
-                logger.error("Skipping audio for slide %d due to empty or invalid script", i + 1)
-                audio_files.append(None)
-                audio_urls[i] = None
-                progress = 90 + ((i + 1) / len(scripts)) * 10
-                label = f"Generating lecture speech for slide {i + 1}/{len(scripts)}..."
-                yield (
-                    html_with_progress(label, progress),
-                    file_paths,
-                    None
-                )
-                await asyncio.sleep(0.1)
-                continue
-            
-            max_audio_retries = 5
             for attempt in range(max_audio_retries + 1):
                 try:
                     current_text = cleaned_script
@@ -891,11 +915,12 @@ Example: 'Received {total_slides} slides, {total_slides} scripts, and HTML files
                     if not success:
                         raise RuntimeError("TTS generation failed")
                     
-                    logger.info("Generated audio for slide %d: %s", i + 1, audio_file)
+                    logger.info("Generated speech for slide %d: %s", i + 1, audio_file)
                     audio_files.append(audio_file)
-                    audio_urls[i] = f"/gradio_api/file={audio_file}"
+                    audio_urls[i] = get_gradio_file_url(audio_file)
                     progress = 90 + ((i + 1) / len(scripts)) * 10
                     label = f"Generating lecture speech for slide {i + 1}/{len(scripts)}..."
+                    add_stage_output("Speech", f"Generated speech for slide {i + 1}")
                     file_paths.append(audio_file)  
                     yield (
                         html_with_progress(label, progress),
@@ -912,6 +937,7 @@ Example: 'Received {total_slides} slides, {total_slides} scripts, and HTML files
                         audio_urls[i] = None
                         progress = 90 + ((i + 1) / len(scripts)) * 10
                         label = f"Generating lecture speech for slide {i + 1}/{len(scripts)}..."
+                        add_stage_output("Speech", f"Failed to generate audio for slide {i + 1}: {str(e)}")
                         yield (
                             html_with_progress(label, progress),
                             file_paths,
@@ -2241,7 +2267,6 @@ with gr.Blocks(
         return gr.update(visible=False), gr.update(visible=True), gr.update(visible=False), gr.update(value=quiz, visible=True)
 
     # Helper to get fallback lecture context from form fields
-
     def get_fallback_lecture_context(lecture_context, title_val, desc_val, style_val, audience_val):
         # If slides/scripts missing, use form fields
         if lecture_context and (lecture_context.get("slides") or lecture_context.get("scripts")):