use ip for user sessison tracnking for suggestions

clients/groq_client.py

@@ -12,6 +12,7 @@ import random
 import shutil
 from optimized_quiz import OPTIMIZED_QUESTIONS
 from chat_resources import *
+from datetime import datetime, timedelta
 
 # Load main env
 load_dotenv(dotenv_path=".env")
@@ -36,6 +37,8 @@ LAMBDA_MULT = float(os.getenv("LAMBDA_MULT", 0.5))
 
 class GroqClient:
     def __init__(self):
+        self._sessions = {}  # {ip: {'shown': set(), 'last_activity': datetime}}
+        self.SESSION_TIMEOUT = timedelta(minutes=30)
 
         self.documents = self.load_json_data(DATA_PATH)
         if not self.documents:
@@ -133,19 +136,41 @@ class GroqClient:
 
     # def get_next_questions(self):
     #     return random.sample(OPTIMIZED_QUESTIONS, 3)
+    def cleanup_expired_sessions(self):
+        """Remove expired sessions to avoid memory overload."""
+        current_time = datetime.now()
+        expired_ips = [
+            ip
+            for ip, data in self._sessions.items()
+            if current_time - data["last_activity"] > self.SESSION_TIMEOUT
+        ]
+        for ip in expired_ips:
+            del self._sessions[ip]
 
-    _shown = set()
-
-    def get_next_questions(self):
+    def get_next_questions(self, ip: str):
         """Return 3 non-repeated random questions within the session."""
-        remaining = [q for q in OPTIMIZED_QUESTIONS if q not in self._shown]
+        self.cleanup_expired_sessions()
+
+        current_time = datetime.now()
+        if ip not in self._sessions:
+            self._sessions[ip] = {"shown": set(), "last_activity": current_time}
+        else:
+            if (
+                current_time - self._sessions[ip]["last_activity"]
+                > self.SESSION_TIMEOUT
+            ):
+                self._sessions[ip]["shown"].clear()
+            self._sessions[ip]["last_activity"] = current_time
+
+        shown = self._sessions[ip]["shown"]
+        remaining = [q for q in OPTIMIZED_QUESTIONS if q not in shown]
 
         if len(remaining) < 3:
-            self._shown.clear()
+            shown.clear()
             remaining = OPTIMIZED_QUESTIONS[:]
 
         selected = random.sample(remaining, 3)
-        self._shown.update(selected)
+        shown.update(selected)
         return selected
 
     # ---------------MAIN-----------------
@@ -186,7 +211,6 @@ class GroqClient:
         ]
 
         # Try multiple models with fallback
-
         models_to_try = [
             "openai/gpt-oss-120b",
             "openai/gpt-oss-20b",

main.py

@@ -26,6 +26,7 @@ app.add_middleware(
 
 class QueryRequest(BaseModel):
     query: str
+    ip: str
 
 # Lazy singleton for GroqClient
 @lru_cache(maxsize=1)
@@ -42,7 +43,7 @@ async def chat_endpoint(request: QueryRequest):
         return {
             "status_code": 200,
             "response": response,
-            "next_questions": groq_client.get_next_questions(),
+            "next_questions": groq_client.get_next_questions(request.ip),
         }
     except Exception as e:
         return {"status_code": 500, "response": f"Internal error: {e}", "next_questions": []}