Update router.py

router.py

@@ -1,3 +1,4 @@
+
 import asyncio
 import time
 import requests
@@ -19,9 +20,7 @@ from rag_engine import rag_response
 # =====================================
 
 IMAGE_SPACE_URL = "https://your-image-space.hf.space/generate"
-
 CACHE_TTL_SECONDS = 300  # 5 minutes
-
 response_cache = {}
 
 
@@ -54,7 +53,6 @@ def set_cache(cache_key, response):
 # =====================================
 
 def build_messages(system_prompt, memory, user_prompt):
-
     messages = []
 
     if system_prompt:
@@ -67,19 +65,19 @@ def build_messages(system_prompt, memory, user_prompt):
 
 
 # =====================================
-# IMAGE SERVICE
+# IMAGE SERVICE (Async Safe)
 # =====================================
 
-def call_image_microservice(prompt):
-
+async def call_image_microservice(prompt):
     try:
-        response = requests.post(
-            IMAGE_SPACE_URL,
-            json={"prompt": prompt},
-            timeout=60
+        return await asyncio.to_thread(
+            lambda: requests.post(
+                IMAGE_SPACE_URL,
+                json={"prompt": prompt},
+                timeout=60
+            ).json()
         )
-        return response.json()
-    except:
+    except Exception:
         return {"error": "Image service unavailable"}
 
 
@@ -99,7 +97,7 @@ async def async_gemini(messages):
 # MAIN ROUTER
 # =====================================
 
-def route_request(prompt, user_id):
+async def route_request(prompt, user_id):
 
     cache_key = f"{user_id}:{prompt}"
 
@@ -115,7 +113,7 @@ def route_request(prompt, user_id):
     # ==========================
     if prompt.startswith("/image"):
         clean_prompt = prompt.replace("/image", "").strip()
-        return call_image_microservice(clean_prompt)
+        return await call_image_microservice(clean_prompt)
 
     # ==========================
     # RAG QUICK RESPONSE
@@ -134,7 +132,6 @@ def route_request(prompt, user_id):
     # CLASSIFY
     # ==========================
     classification = classify_prompt(prompt)
-
     intent = classification.get("intent", "chat")
     needs_search = classification.get("needs_search", False)
 
@@ -158,7 +155,7 @@ def route_request(prompt, user_id):
     if intent == "reasoning":
 
         messages = build_messages(system_prompt, memory, prompt)
-        response = call_gemini(messages)
+        response = await async_gemini(messages)
 
         save_message(user_id, "user", prompt)
         save_message(user_id, "assistant", response)
@@ -185,16 +182,11 @@ Use web data if helpful.
 
         messages = build_messages(system_prompt, memory, enriched_prompt)
 
-        async def run_parallel():
-            llama_task = asyncio.create_task(async_llama(messages))
-            gemini_task = asyncio.create_task(async_gemini(messages))
-
-            llama_answer = await llama_task
-            gemini_answer = await gemini_task
-
-            return llama_answer, gemini_answer
+        llama_task = asyncio.create_task(async_llama(messages))
+        gemini_task = asyncio.create_task(async_gemini(messages))
 
-        llama_answer, gemini_answer = asyncio.run(run_parallel())
+        llama_answer = await llama_task
+        gemini_answer = await gemini_task
 
         winner = judge_answers(llama_answer, gemini_answer)
         final_answer = gemini_answer if winner == 2 else llama_answer
@@ -211,7 +203,7 @@ Use web data if helpful.
     # ==========================
     messages = build_messages(system_prompt, memory, prompt)
 
-    response = call_llama(messages)
+    response = await async_llama(messages)
 
     save_message(user_id, "user", prompt)
     save_message(user_id, "assistant", response)