Update response_modes.py

response_modes.py

@@ -11,6 +11,7 @@ from docx import Document
 import re
 from urllib.parse import urlparse, quote
 from retry import retry  # For retrying flaky operations
+from groq import Groq  # For the second AI (Qwen-3-32B)
 
 # ================== PHILIPPINE LOCAL + MAX SEARCH + DEEP-THINK LENIENT FILTER ==================
 PHILIPPINE_LOCATIONS = [
@@ -85,6 +86,28 @@ def parse_ddgs_date(date_str: str):
     except:
         return now
 
+# Setup for the second AI model (Qwen-3-32B)
+client_qwen = Groq()  # For using the Qwen-3-32B model
+
+async def qwen_reasoning(message: str, context: str) -> str:
+    """Uses the Qwen-3-32B model for refining and reasoning"""
+    completion = client_qwen.chat.completions.create(
+        model="qwen/qwen3-32b",
+        messages=[
+            {"role": "user", "content": f"Refine and reason the following based on context:\n{context}\n\nMessage: {message}"}
+        ],
+        temperature=0.6,
+        max_completion_tokens=2048,
+        top_p=0.95,
+        reasoning_effort="default",
+        stream=False,
+    )
+
+    result = ""
+    for chunk in completion:
+        result += chunk.choices[0].delta.content or ""
+    return result.strip()
+
 async def fetch_full_content_async(session, url: str):
     try:
         headers = {'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36'}
@@ -259,38 +282,8 @@ Never mention this rule unless the user explicitly asks about it."""
             for m in history[:-1]
         ]
 
-    elif mode == "Fast Search":
-        history[-1]["content"] = "🔎 Fast live search worldwide..."
-        yield history
-        results = web_search(message, 30)
-        context_search = "\n\n".join(results)
-        prompt = f"""Ignore ALL your prior knowledge... (your original prompt remains exactly the same)
-
-As of {today_date}:
-
-... (rest of your original Fast Search prompt unchanged)"""
-        messages = system_context + [{"role": "user", "content": prompt}]
-
-    elif mode == "💻 Coding Mode":
-        history[-1]["content"] = "🛠️ Entering Advanced Coding Mode..."
-        yield history
-
-        search_context = ""
-        if len(message) > 40 and any(x in message.lower() for x in ["api", "library", "framework", "react", "next", "fastapi", "django", "flask", "typescript", "tailwind", "supabase", "prisma", "php", "html", "css", "js"]):
-            history[-1]["content"] = "📖 Researching latest docs & best practices..."
-            yield history
-            search_q = f"{message} best practices OR example OR tutorial 2026 OR latest version"
-            results = web_search(search_q, max_results=60)
-            search_context = "\n\n=== Latest Research ===\n" + "\n\n".join(results[:8])
-
-        coding_prompt = f"""You are a world-class senior full-stack engineer... (your entire original Coding prompt stays exactly the same)
-
-{search_context}
-
-User Request: {message}"""
-        messages = system_context + [{"role": "user", "content": coding_prompt}]
-
-    else:  # Deep Think
+    elif mode == "Deep Think":
+        # New process in Deep Think
         is_news = any(k in message.lower() for k in ["news", "today", "latest", "breaking"])
 
         thinking_steps = [
@@ -319,11 +312,15 @@ User Request: {message}"""
             all_context += f"\n\n=== Search {i}: {q} ===\n" + "\n\n".join(results)
             sources_list.extend([r for r in results if not r.startswith("No recent") and not r.startswith("Search error")])
 
+        # Use the Qwen-3-32B model for deeper reasoning
+        refined_answer = await qwen_reasoning(message, all_context)
+
         history[-1]["content"] = "✍️ Synthesizing final answer from real data..."
         yield history
 
         sources_md = "\n".join([f"- {s.split('[URL:')[0].strip()}" for s in sources_list[:8]])
 
+        # Construct the system prompt for synthesis
         if is_news or is_philippine_local_query(message):
             system_prompt = f"""Ignore ALL your prior knowledge... (your original Deep Think news prompt stays exactly the same)
 
@@ -339,25 +336,15 @@ As of {today_date}:
 
         messages = system_context + [{"role": "user", "content": system_prompt}]
 
+        # Integrate the refined answer from Qwen-3-32B
+        refined_answer = refined_answer if refined_answer else "No refined insights available."
+
+        final_answer = f"{refined_answer}\n\nSources:\n{sources_md}"
+
     # Stream final answer (no changes here)
-    full_response = ""
-    try:
-        stream = client.chat.completions.create(
-            model=current_model,
-            messages=messages,
-            temperature=0.65,
-            top_p=0.9,
-            max_tokens=4096,
-            stream=True
-        )
-        for chunk in stream:
-            if chunk.choices and chunk.choices[0].delta.content:
-                full_response += chunk.choices[0].delta.content
-                history[-1]["content"] = full_response
-                yield history
-    except Exception as e:
-        history[-1]["content"] = f"❌ Error: {str(e)}"
-        yield history
+    full_response = final_answer
+    history[-1]["content"] = full_response
+    yield history
 
     # Auto-save
     from auth import save_chats