Update app.py

app.py

@@ -12,7 +12,7 @@ import gradio as gr
 
 tokenizer = AutoTokenizer.from_pretrained("answerdotai/ModernBERT-base")
 sentence_model = SentenceTransformer('all-MiniLM-L6-v2')
-max_tokens = 4000
+max_tokens = 3000
 
 def clean_text(text):
     text = re.sub(r'\[speaker_\d+\]', '', text)
@@ -164,9 +164,9 @@ def analyze_segment_with_gemini(cluster_text, is_full_text=False):
         prompt = f"""
             Analyze the following text (likely a transcript or document) and:
             
-            1. First, identify distinct segments or topics within the text
+            1. First, do text segmentation and identify DISTINCT key topics within the text
             2. For each segment/topic you identify:
-               - Provide a concise topic name (3-5 words)
+               - Provide a SPECIFIC and UNIQUE topic name (3-5 words) that clearly differentiates it from other segments
                - List 3-5 key concepts discussed in that segment
                - Write a brief summary of that segment (3-5 sentences)
                - Create 5 quiz questions based DIRECTLY on the content in that segment
@@ -215,7 +215,7 @@ def analyze_segment_with_gemini(cluster_text, is_full_text=False):
     else:
         prompt = f"""
             Analyze the following text segment and provide:
-            1. A concise topic name (3-5 words)
+            1. A SPECIFIC and DESCRIPTIVE topic name (3-5 words) that precisely captures the main focus
             2. 3-5 key concepts discussed
             3. A brief summary (6-7 sentences)
             4. Create 5 quiz questions based DIRECTLY on the text content (not from your summary)
@@ -298,7 +298,7 @@ def process_document_with_quiz(text):
     token_count = len(tokenizer.encode(text))
     print(f"Text contains {token_count} tokens")
     
-    if token_count < 12000:
+    if token_count < 7000:
         print("Text is short enough to analyze directly without text segmentation")
         full_analysis = analyze_segment_with_gemini(text, is_full_text=True)