Update app.py

app.py

@@ -20,13 +20,8 @@ MAX_HISTORY_TURNS = 10  # Maximum conversation turns to keep in context
 client = genai.Client(api_key=GEMINI_KEY)
 user_memory = {}  # { user_id: { "history": [], "last_sync": timestamp } }
 
-# --- Helper: sync single user's history (added) ---
+# --- Helper: sync single user's history ---
 def sync_user(uid):
-    """
-    Attempt to sync a single user's most recent history to LAMBDA_URL.
-    Updates user_memory[uid]['last_sync'] on success and logs results.
-    This runs in its own thread when called from update_user_history and in the flush loop.
-    """
     try:
         data = user_memory.get(uid)
         if not data:
@@ -40,18 +35,15 @@ def sync_user(uid):
     except Exception as e:
         app.logger.warning(f"Failed sync for {uid}: {e}")
 
-# --- Background thread for periodic flush (uses helper) ---
+# --- Background thread for periodic flush ---
 def flush_loop():
     while True:
         now = time.time()
         for uid, data in list(user_memory.items()):
             if now - data.get("last_sync", 0) >= FLUSH_INTERVAL and data.get("history"):
-                # Offload the actual network call to a short-lived thread so the loop isn't blocked.
                 threading.Thread(target=sync_user, args=(uid,), daemon=True).start()
         time.sleep(5)
 
-# NOTE: don't start the thread at module import time to avoid issues with Flask reloader.
-# We'll start it once when the app starts serving requests.
 def start_flush_thread_once():
     t = threading.Thread(target=flush_loop, daemon=True)
     t.start()
@@ -110,29 +102,21 @@ HTML = """
 # --- Gemini Generation with History ---
 def generate_from_gemini(prompt, image_bytes=None, history=None):
     start_time = time.time()
-    
-    # Build contents list with history
     contents = []
-    
-    # Add historical messages (limit to recent turns to avoid token limits)
+
     if history:
         recent_history = history[-MAX_HISTORY_TURNS:]
         for entry in recent_history:
-            # Add user message
             user_parts = [types.Part.from_text(text=entry["prompt"])]
             contents.append(types.Content(role="user", parts=user_parts))
-            
-            # Add model response
             model_parts = [types.Part.from_text(text=entry["response"])]
             contents.append(types.Content(role="model", parts=model_parts))
     
-    # Add current user message
     current_parts = []
     if prompt: 
         current_parts.append(types.Part.from_text(text=prompt))
     if image_bytes: 
         current_parts.append(types.Part.from_bytes(data=image_bytes, mime_type="image/jpeg"))
-    
     contents.append(types.Content(role="user", parts=current_parts))
     
     cfg = types.GenerateContentConfig(response_mime_type="text/plain")
@@ -160,7 +144,6 @@ def get_user_history(uid):
             resp = requests.get(f"{LAMBDA_URL}?user_id={uid}", timeout=5)
             resp.raise_for_status()
             loaded_history = resp.json().get("history", [])
-            # Only keep the most recent MAX_HISTORY_TURNS when loading
             user_memory[uid] = {
                 "history": loaded_history[-MAX_HISTORY_TURNS:], 
                 "last_sync": 0
@@ -174,13 +157,9 @@ def get_user_history(uid):
 def update_user_history(uid, prompt, response):
     entry = {"prompt": prompt, "response": response, "timestamp": time.time()}
     user_memory.setdefault(uid, {"history": [], "last_sync": 0})["history"].append(entry)
-    
-    # Trim history to MAX_HISTORY_TURNS to prevent unbounded growth
     if len(user_memory[uid]["history"]) > MAX_HISTORY_TURNS:
         user_memory[uid]["history"] = user_memory[uid]["history"][-MAX_HISTORY_TURNS:]
         app.logger.debug(f"Trimmed history for {uid} to {MAX_HISTORY_TURNS} turns")
-    
-    # Attempt immediate async sync so you can see logs quickly (added)
     threading.Thread(target=sync_user, args=(uid,), daemon=True).start()
 
 # --- Routes ---
@@ -201,22 +180,16 @@ def gen():
         return jsonify({"error": "No prompt or image provided"}), 400
 
     try:
-        # Load user's conversation history
         history = get_user_history(uid)
-        
-        # Generate response with history context
         result = generate_from_gemini(prompt, img_bytes, history=history)
-        
-        # Update history with new exchange
         update_user_history(uid, prompt, result["text"])
-        
         return jsonify({"result": result["text"], "timing": result["timing"]})
     except Exception as e:
         app.logger.exception("Generation failed")
         return jsonify({"error": str(e)}), 500
 
-# Ensure background thread starts when app begins serving requests
-@app.before_first_request
+# Flask 3.x compatible event
+@app.before_serving
 def _start_background_tasks():
     start_flush_thread_once()