Update app.py

app.py

@@ -100,15 +100,25 @@ def get_model(temperature, top_p, repetition_penalty):
         huggingfacehub_api_token=huggingface_token
     )
 
-def generate_chunked_response(model, prompt, max_tokens=1000, max_chunks=5):
+def generate_chunked_response(model, prompt, max_tokens=200):
     full_response = ""
-    for i in range(max_chunks):
-        chunk = model(prompt + full_response, max_new_tokens=max_tokens)
-        chunk = chunk.strip()
-        if chunk.endswith((".", "!", "?")):
+    total_length = len(prompt.split())  # Approximate token count of prompt
+    
+    while total_length < 7800:  # Leave some margin
+        try:
+            chunk = model(prompt + full_response, max_new_tokens=min(200, 7800 - total_length))
+            chunk = chunk.strip()
+            if not chunk:
+                break
             full_response += chunk
+            total_length += len(chunk.split())  # Approximate token count
+            
+            if chunk.endswith((".", "!", "?")):
+                break
+        except Exception as e:
+            print(f"Error generating response: {str(e)}")
             break
-        full_response += chunk
+    
     return full_response.strip()
 
 def manage_conversation_history(question, answer, history, max_history=5):
@@ -223,8 +233,14 @@ def google_search(term, num_results=20, lang="en", timeout=5, safe="active", ssl
     return all_results
 
 def summarize_content(content, model):
+    # Approximate the token limit using character count
+    # Assuming an average of 4 characters per token
+    max_chars = 7000 * 4  # Leave some room for the prompt
+    if len(content) > max_chars:
+        content = content[:max_chars] + "..."
+    
     summary_prompt = f"""
-    Summarize the following content in a concise manner:
+    Summarize the following content concisely:
     {content}
     Summary:
     """
@@ -262,12 +278,28 @@ def ask_question(question, temperature, top_p, repetition_penalty, web_search):
         database = FAISS.load_local("faiss_database", embed, allow_dangerous_deserialization=True)
     else:
         database = None
-    
+
     if web_search:
         search_results = google_search(question)
         model = get_model(temperature, top_p, repetition_penalty)
         
-        summaries = [summarize_content(result["text"], model) for result in search_results]
+        summaries = []
+        for result in search_results:
+            try:
+                summary = summarize_content(result["text"], model)
+                summaries.append(summary)
+            except Exception as e:
+                print(f"Error summarizing content: {str(e)}")
+                summaries.append("Error: Unable to summarize this content.")
+        
+        # Combine summaries, ensuring we don't exceed the token limit
+        combined_summaries = ""
+        for summary in summaries:
+            if len((combined_summaries + summary).split()) > 7000:
+                break
+            combined_summaries += summary + "\n\n"
+        
+        context_str = combined_summaries
         titles = [result["title"] for result in search_results]
         ranks = rank_search_results(titles, summaries, model)