Update app.py

app.py

@@ -5,15 +5,19 @@ from prometheus_client import Counter, generate_latest, CONTENT_TYPE_LATEST
 from transformers import pipeline
 import gradio as gr
 
-# ✅ Expose FastAPI app at top level
+# ✅ FastAPI app
 app = FastAPI(title="Incident Triage Bot")
 
 # 📊 Prometheus metrics
 triage_requests = Counter("triage_requests_total", "Total triage requests")
 triage_errors = Counter("triage_errors_total", "Total triage errors")
 
-# 🧠 Load local model
-triage_pipeline = pipeline("text2text-generation", model="google/flan-t5-base")
+# 🧠 Load lightweight model (CPU only)
+triage_pipeline = pipeline(
+    "text2text-generation",
+    model="google/flan-t5-small",
+    device=-1  # Force CPU
+)
 
 # 📥 Request model
 class Incident(BaseModel):
@@ -57,7 +61,7 @@ demo = gr.Interface(
     inputs=gr.Textbox(lines=5, placeholder="Describe the incident..."),
     outputs="text",
     title="Incident Triage Bot",
-    description="Classify incidents and suggest priority using a local Transformers model"
+    description="Classify incidents and suggest priority using a lightweight Transformers model"
 )
 
 @app.get("/gradio")