Update stay alive

app.py

@@ -3,7 +3,7 @@ import json
 import time
 import numpy as np
 from sentence_transformers import SentenceTransformer
-from fastapi import FastAPI, HTTPException
+from fastapi import FastAPI, HTTPException, BackgroundTasks
 from fastapi.responses import StreamingResponse
 from pydantic import BaseModel
 from llama_cpp import Llama
@@ -239,4 +239,27 @@ async def warm_up_model():
     logger.info("Model warm-up completed.")
     # Log initial RAM usage
     ram_stats = get_ram_usage()
-    logger.info(f"Initial RAM usage after startup: {ram_stats}")
+    logger.info(f"Initial RAM usage after startup: {ram_stats}")
+
+# Add a background task to keep the model warm
+@app.on_event("startup")
+async def setup_periodic_tasks():
+    asyncio.create_task(keep_model_warm())
+    logger.info("Periodic model warm-up task scheduled")
+
+async def keep_model_warm():
+    """Background task that keeps the model warm by sending periodic requests"""
+    while True:
+        try:
+            logger.info("Performing periodic model warm-up")
+            dummy_query = "Say only the word 'ok.'"
+            dummy_history = []
+            # Process a dummy query through the generator to keep it warm
+            async for _ in stream_response(dummy_query, dummy_history):
+                pass
+            logger.info("Periodic warm-up completed")
+        except Exception as e:
+            logger.error(f"Error in periodic warm-up: {str(e)}")
+        
+        # Wait for 13 minutes before the next warm-up
+        await asyncio.sleep(13 * 60)