Optimize timeouts: 20s orchestrator, 15s LLM client, language-aware fallback

app/core/llm_client.py

@@ -49,7 +49,8 @@ from app.core.personas import PERSONAS
 from app.core.time_utils import TimeAwareBehavior
 
 # Shared HTTP Client for performance (Connection Pooling)
-_shared_client = httpx.AsyncClient(timeout=30.0)
+# Timeout reduced to 15s to fit within GUVI's 30s limit (with buffer for multiple calls)
+_shared_client = httpx.AsyncClient(timeout=15.0)
 
 
 from enum import Enum

app/utils/guvi_handler.py

@@ -237,8 +237,8 @@ class GUVIHandler:
             
             logger.debug(f"🔥 Orchestrator reached with language: {request_language}") # [DEBUG] Verify flow
             try:
-                # [LATENCY] Strict 25s Timeout to satisfy GUVI's 30s limit
-                # If LLM hangs, we abort and return fallback immediately
+                # [LATENCY] Strict 20s Timeout to satisfy GUVI's 30s limit
+                # Buffer: 20s processing + 10s network/serialization headroom
                 result = await asyncio.wait_for(
                     orchestrator.process_message(
                         message=scammer_text,
@@ -250,16 +250,18 @@ class GUVIHandler:
                         should_finalize=is_finalizing_turn, # [RESTORED] Critical for callback trigger
                         request_language=request_language  # [FIX] Pass GUVI language preference
                     ),
-                    timeout=25.0
+                    timeout=20.0
                 )
             except asyncio.TimeoutError:
-                logger.error(f"DATA TIMEOUT ({session_id}): Orchestrator took >25s. Forcing fallback.")
+                logger.error(f"DATA TIMEOUT ({session_id}): Orchestrator took >20s. Forcing fallback.")
                 # Construct a minimal valid 'result' to allow fall-through to standard response builder
+                # Use language-aware fallback message
+                fallback_msg = "Hello? My network is slow, please wait..." if request_language == "english" else "Hello? Thoda network slow hai mera..."
                 result = {
                     "status": "partial_success",
                     "is_scam": False, # Fail open (continue)
                     "threat_level": "MEDIUM", 
-                    "honeypot_response": {"message": "Hello? Thoda network slow hai mera.", "persona": "fallback"},
+                    "honeypot_response": {"message": fallback_msg, "persona": "fallback"},
                     "conversation": {"message_count": db_history_len + 1},
                     "aggregated_intelligence": conv.get("aggregated_intelligence", {}),
                     "confidence": 0.0,

scripts/quick_test.py

@@ -42,6 +42,12 @@ def test_honeypot():
             "sender": "scammer",
             "text": test_message,
             "timestamp": int(time.time() * 1000)
+        },
+        "conversationHistory": [],
+        "metadata": {
+            "channel": "SMS",
+            "language": "English",  # Request English responses
+            "locale": "IN"
         }
     }