Update app.py

app.py

@@ -56,7 +56,7 @@ async def brave_search(query, count=1):
 
 
 
-pipeline_lock = asyncio.Lock()
+# pipeline_lock = asyncio.Lock()
 
 @traceable
 @log_time
@@ -64,10 +64,10 @@ async def query_teapot(prompt, context, user_input):
     input_text = prompt + "\n" + context + "\n" + user_input
     inputs = tokenizer(input_text, return_tensors="pt")
     
-    async with pipeline_lock:  # Ensure only one call runs at a time
-        output = await asyncio.to_thread(model.generate, **inputs, max_new_tokens=512)
-        output_text = tokenizer.decode(output[0], skip_special_tokens=True)
-        return output_text
+    # async with pipeline_lock:  # Ensure only one call runs at a time
+    output = await asyncio.to_thread(model.generate, **inputs, max_new_tokens=512)
+    output_text = tokenizer.decode(output[0], skip_special_tokens=True)
+    return output_text
 
 
 @log_time