Update main.py

main.py

@@ -11,21 +11,40 @@ TELEGRAM_API_URL = f"https://api.telegram.org/bot{TELEGRAM_TOKEN}"
 
 app = FastAPI()
 
+# --- MEMORY BANK ---
+# This dictionary remembers which engine you are currently using.
+user_engines = {}
+
 # --- NATIVE AGENT ENGINE ---
-async def run_agent(prompt: str) -> str:
+async def run_agent(prompt: str, model_name: str) -> str:
     try:
         response = await acompletion(
-            model="gemini/gemini-2.5-pro", # Locked onto 2.5 Pro
+            model=model_name, 
             messages=[{"role": "user", "content": prompt}],
             temperature=0.2
         )
         return response.choices[0].message.content or "No response generated."
     except Exception as e:
-        return f"⚠️ Engine Error: {str(e)}"
+        return f"⚠️ Engine Error ({model_name}): {str(e)}"
 
 # --- BACKGROUND TASK: PROCESS AND REPLY ---
 async def process_and_reply(chat_id: int, text: str):
-    response_text = await run_agent(text)
+    user_msg = text.strip().lower()
+    
+    # 1. Check if the user is pulling the Gear Shifter (Slash Commands)
+    if user_msg == "/pro":
+        user_engines[chat_id] = "gemini/gemini-1.5-pro"
+        response_text = "🧠 **Engine Switched: Gemini 1.5 Pro**\n_You are now using the heavy-duty coding engine (50 free requests/day)._"
+    
+    elif user_msg == "/flash":
+        user_engines[chat_id] = "gemini/gemini-1.5-flash"
+        response_text = "⚡ **Engine Switched: Gemini 1.5 Flash**\n_You are now using the ultra-fast casual engine (1,500 free requests/day)._"
+    
+    # 2. If it's a normal message, route it to the currently active engine
+    else:
+        # Default to Flash if we haven't set an engine yet to protect the Pro quota
+        current_model = user_engines.get(chat_id, "gemini/gemini-1.5-flash") 
+        response_text = await run_agent(text, current_model)
     
     # Send the final response back to Telegram
     async with httpx.AsyncClient() as client:
@@ -33,7 +52,8 @@ async def process_and_reply(chat_id: int, text: str):
             f"{TELEGRAM_API_URL}/sendMessage",
             json={
                 "chat_id": chat_id, 
-                "text": response_text
+                "text": response_text,
+                "parse_mode": "Markdown"
             }
         )
 
@@ -42,10 +62,8 @@ async def process_and_reply(chat_id: int, text: str):
 async def health_check():
     return {"status": "alive"}
 
-# Accept EVERY method to prevent "Method Not Allowed" from HF routing
 @app.api_route("/webhook", methods=["GET", "POST", "PUT"])
 async def telegram_webhook(request: Request, background_tasks: BackgroundTasks):
-    # If it's a GET request (like a browser test), just confirm it's alive
     if request.method == "GET":
         return {"status": "Webhook is active. Send a POST request from Telegram!"}
     
@@ -57,10 +75,9 @@ async def telegram_webhook(request: Request, background_tasks: BackgroundTasks):
             chat_id = data["message"]["chat"]["id"]
             user_msg = data["message"]["text"]
             
-            # Run the AI generation in the background so we can instantly reply "OK" to Telegram
+            # Run in the background
             background_tasks.add_task(process_and_reply, chat_id, user_msg)
             
-        # Telegram requires an immediate 200 OK so it doesn't get stuck in a retry loop
         return {"status": "ok"}
     except Exception as e:
         return {"status": f"error reading json: {str(e)}"}