Update app.py

app.py

@@ -28,11 +28,15 @@ def root():
 
 @app.post("/chat")
 def chat(req: ChatRequest):
-    output = llm(
-        f"<|user|>{req.message}<|assistant|>",
-        max_tokens=300,
-        stop=["<|end|>"]
-    )
+   output = llm(
+    f"<|user|>{req.message}<|assistant|>",
+    max_tokens=512,
+    temperature=0.7,
+    top_p=0.9,
+    repeat_penalty=1.1,
+    stop=["<|end|>"]
+)
+
     return {"reply": output["choices"][0]["text"]}
 
 # ⭐ THIS PART WAS MISSING