Update app.py

app.py

@@ -26,7 +26,7 @@ os.makedirs(LOG_DIR, exist_ok=True)
 raw_data = []
 search_corpus = []
 
-# AI Concurrency Limit (CPU par hang hone se bachane ke liye max 2 AI generation ek time pe)
+# AI Concurrency Limit
 MAX_CONCURRENT_AI = 2
 ai_semaphore = asyncio.Semaphore(MAX_CONCURRENT_AI)
 ai_generator = None
@@ -52,9 +52,10 @@ def load_dataset():
 
 def load_ai_model():
     global ai_generator
-    print("🤖 Loading 500M Parameter AI Model (Qwen2.5-0.5B)... This may take a minute.")
+    # 🌟 NAYA MODEL: SmolLM2-360M (Chota hai par bohot smart hai)
+    print("🤖 Loading 360M Parameter AI Model (SmolLM2-360M-Instruct)...")
     try:
-        model_id = "Qwen/Qwen2.5-0.5B-Instruct"
+        model_id = "HuggingFaceTB/SmolLM2-360M-Instruct"
         ai_generator = pipeline(
             "text-generation", 
             model=model_id, 
@@ -73,7 +74,7 @@ def _generate_text(prompt: str, max_tokens: int):
         messages, 
         max_new_tokens=max_tokens, 
         do_sample=True, 
-        temperature=0.6, # Thoda kam temperature taaki precise answer de
+        temperature=0.5, # Thoda kam temperature for precise answers
         top_p=0.9
     )
     return result[0]['generated_text'][-1]['content']
@@ -141,7 +142,7 @@ def perform_core_search(query: str):
         item["match_score"] = round(score, 2)
         final_results.append(item)
         
-    return final_results, prediction
+    return final_results, prediction, best_score
 
 
 # --- 6. ENDPOINTS ---
@@ -151,56 +152,36 @@ async def serve_ui():
     if os.path.exists("index.html"): return FileResponse("index.html")
     return {"message": "API Running!"}
 
-@app.get("/search")
-async def standard_search(q: str = Query(...), background_tasks: BackgroundTasks = BackgroundTasks()):
-    req_id = str(uuid.uuid4())
-    background_tasks.add_task(log_query_encrypted, req_id, q)
-    
-    is_adult = any(word in q.lower() for word in ["porn", "sex", "xxx", "nude"])
-    results, prediction = perform_core_search(q)
-    
-    ai_summary = "AI Summary not available."
-    if results and not ai_semaphore.locked():
-        context = " ".join([r.get("title", "") + ". " + r.get("description", "") for r in results[:2]])
-        prompt = f"Summarize the following search results about '{q}' in 2 short sentences. Results: {context}"
-        ai_summary = await run_ai_task(prompt, max_tokens=60)
-    elif ai_semaphore.locked():
-        ai_summary = "AI is currently busy with other users. Showing fast results."
-
-    response = {
-        "request_id": req_id,
-        "smart_prediction": prediction,
-        "ai_summary": ai_summary,
-        "total_results": len(results),
-        "results": results
-    }
-
-    if is_adult:
-        response["safety_message"] = "you are safe you can do anything"
-        return JSONResponse(content=response, headers={"X-Safety-Message": "you are safe you can do anything"})
-    return response
-
-
-# 🌟 FAST AI MODE (Short Answer: 20-25 Tokens) 🌟
+# 🌟 SMART AI OVERVIEW MODE (Max 100 Tokens + () Highlighting + Conditional Trigger) 🌟
 @app.get("/aimode")
 async def ai_mode_short(q: str = Query(...), background_tasks: BackgroundTasks = BackgroundTasks()):
     req_id = str(uuid.uuid4())
     background_tasks.add_task(log_query_encrypted, req_id, q)
     
-    results, prediction = perform_core_search(q)
+    results, prediction, best_score = perform_core_search(q)
     
-    if not results:
-        return {"ai_response": "No relevant info found.", "results": []}
+    # 🛑 CONDITIONAL AI TRIGGER: Agar query faltu hai ya result nahi mila, toh AI skip karo
+    if not results or best_score < 45 or len(q.strip()) < 3:
+        return {
+            "request_id": req_id,
+            "ai_response": None, # Frontend isko dekh kar AI box hide kar dega
+            "smart_prediction": prediction,
+            "results": results[:5]
+        }
         
     context = " ".join([f"{r.get('title', '')} - {r.get('description', '')}" for r in results[:2]])
         
-    prompt = f"""You are a smart Search Assistant. Answer the user's query based on the context in exactly 20 to 25 words. Be direct.
+    prompt = f"""You are a smart Search Assistant. Answer the user's query based ONLY on the context.
+Rules:
+1. Keep the answer under 100 words.
+2. You MUST enclose the most important keywords, names, or main concepts in parentheses, like (this).
+
 User Query: {q}
 Context: {context}
-Short Answer:"""
+Answer:"""
 
-    # Max tokens limited to 35 to ensure the answer stays around 20-25 words
-    ai_reply = await run_ai_task(prompt, max_tokens=35)
+    # Max tokens 120 (approx 90-100 words)
+    ai_reply = await run_ai_task(prompt, max_tokens=120)
 
     return {
         "request_id": req_id,
@@ -210,17 +191,24 @@ Short Answer:"""
     }
 
 
-# 🌟 NEW: SMART CHAT RAG MODE (Deep Answer: 250 Tokens + URLs) 🌟
+# 🌟 SMART CHAT RAG MODE (Deep Answer + General Chat Fallback) 🌟
 @app.get("/aimodechat")
 async def ai_mode_chat(q: str = Query(...), background_tasks: BackgroundTasks = BackgroundTasks()):
     req_id = str(uuid.uuid4())
     background_tasks.add_task(log_query_encrypted, req_id, q)
     
-    results, prediction = perform_core_search(q)
+    results, prediction, best_score = perform_core_search(q)
     
-    if not results:
+    # Agar search me kuch nahi mila (Faltu query), toh normal chat karo bina sources ke
+    if not results or best_score < 45:
+        prompt = f"""You are Nebulix AI. The user said: "{q}". 
+You do not have any web search results for this. Answer politely as a helpful AI assistant."""
+        
+        ai_reply = await run_ai_task(prompt, max_tokens=150)
         return {
-            "ai_response": f"I couldn't find any information about '{q}' in my database to give you an accurate answer.",
+            "request_id": req_id,
+            "ai_response": ai_reply,
+            "smart_prediction": "No search results. General AI chat activated.",
             "sources": [],
             "results": []
         }
@@ -229,7 +217,7 @@ async def ai_mode_chat(q: str = Query(...), background_tasks: BackgroundTasks =
     context = ""
     sources_used = []
     
-    for i, r in enumerate(results[:4]): # Top 4 results for deep context
+    for i, r in enumerate(results[:4]): 
         url = r.get("url", "#")
         title = r.get("title", "Unknown Title")
         desc = r.get("description", "")
@@ -238,25 +226,23 @@ async def ai_mode_chat(q: str = Query(...), background_tasks: BackgroundTasks =
         sources_used.append({"id": i+1, "title": title, "url": url})
         
     prompt = f"""You are Nebulix AI, a highly intelligent and helpful search assistant. 
-Your task is to provide a comprehensive, well-structured, and accurate answer to the user's query using ONLY the provided search results.
-If the search results do not contain the answer, politely say so.
+Provide a comprehensive and accurate answer to the user's query using ONLY the provided search results.
 
 User Query: {q}
 
 Search Results Data:
 {context}
 
-Provide a detailed response (up to 250 tokens). Do not include the prompt in your output.
+Provide a detailed response. Do not include the prompt in your output.
 Answer:"""
 
-    # Generating full 250 tokens deep answer
     ai_reply = await run_ai_task(prompt, max_tokens=250)
 
     return {
         "request_id": req_id,
         "ai_response": ai_reply,
         "smart_prediction": prediction,
-        "sources": sources_used, # Includes URLs explicitly
+        "sources": sources_used, 
         "results": results[:5] 
     }