Update app.py

app.py

@@ -455,14 +455,16 @@ If any part of the information cannot be verified from this source, clearly stat
 
         if model == "@cf/meta/llama-3.1-8b-instruct":
             # Use Cloudflare API
+            source_response = ""
             for response in get_response_from_cloudflare(prompt="", context=context, query=query, num_calls=1, temperature=temperature, search_type="web"):
-                accumulated_response += f"Source {i} ({source}):\n\n{response}\n\n"
-                yield accumulated_response, ""
+                source_response += response
+            accumulated_response += f"Source {i} ({source}):\n\n{source_response}\n\n"
+            yield accumulated_response, ""
         else:
             # Use Hugging Face API
             client = InferenceClient(model, token=huggingface_token)
             
-            summary = ""
+            source_response = ""
             for message in client.chat_completion(
                 messages=[{"role": "user", "content": prompt}],
                 max_tokens=2000,
@@ -471,9 +473,9 @@ If any part of the information cannot be verified from this source, clearly stat
             ):
                 if message.choices and message.choices[0].delta and message.choices[0].delta.content:
                     chunk = message.choices[0].delta.content
-                    summary += chunk
-                    accumulated_response = f"{accumulated_response}Source {i} ({source}):\n\n{summary}\n\n"
-                    yield accumulated_response, ""
+                    source_response += chunk
+            accumulated_response += f"Source {i} ({source}):\n\n{source_response}\n\n"
+            yield accumulated_response, ""
     
     # Generate an overall summary after processing all sources
     overall_prompt = f"""Based on the summaries you've generated for each source, provide a concise overall summary that addresses the user's query: '{query}'
@@ -481,9 +483,11 @@ Highlight any conflicting information or gaps in the available data."""
 
     if model == "@cf/meta/llama-3.1-8b-instruct":
         # Use Cloudflare API for overall summary
+        overall_response = ""
         for response in get_response_from_cloudflare(prompt="", context="", query=overall_prompt, num_calls=1, temperature=temperature, search_type="web"):
-            accumulated_response += f"Overall Summary:\n\n{response}\n\n"
-            yield accumulated_response, ""
+            overall_response += response
+        accumulated_response += f"Overall Summary:\n\n{overall_response}\n\n"
+        yield accumulated_response, ""
     else:
         # Use Hugging Face API for overall summary
         overall_summary = ""
@@ -496,8 +500,8 @@ Highlight any conflicting information or gaps in the available data."""
             if message.choices and message.choices[0].delta and message.choices[0].delta.content:
                 chunk = message.choices[0].delta.content
                 overall_summary += chunk
-                accumulated_response += f"Overall Summary:\n\n{overall_summary}\n\n"
-                yield accumulated_response, ""
+        accumulated_response += f"Overall Summary:\n\n{overall_summary}\n\n"
+        yield accumulated_response, ""
 
 def get_response_from_pdf(query, model, selected_docs, num_calls=3, temperature=0.2):
     logging.info(f"Entering get_response_from_pdf with query: {query}, model: {model}, selected_docs: {selected_docs}")