Update app.py

app.py

@@ -4,11 +4,17 @@ from pydantic import BaseModel
 from transformers import AutoTokenizer, AutoModelForSequenceClassification
 import torch
 
-app = FastAPI(title="Routing Service - Space 2")
-
+# -------------------------------
+# Set Hugging Face cache to writable directory
+# -------------------------------
 os.environ["HF_HOME"] = "/data/huggingface-cache"
 os.environ["TRANSFORMERS_CACHE"] = "/data/huggingface-cache"
 
+# -------------------------------
+# FastAPI app
+# -------------------------------
+app = FastAPI(title="Routing Service - Space 2")
+
 # -------------------------------
 # Request Model
 # -------------------------------
@@ -16,16 +22,14 @@ class RoutingRequest(BaseModel):
     text: str
 
 # -------------------------------
-# Load Routing Model (DeBERTa MNLI)
+# Load DeBERTa MNLI Model
 # -------------------------------
 MODEL_NAME = "MoritzLaurer/DeBERTa-v3-base-mnli-fever-anli"
-tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
+tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, use_fast=True)
 model = AutoModelForSequenceClassification.from_pretrained(MODEL_NAME)
 
-# Define your possible departments / labels
-DEPARTMENTS = ["Account", "Software", "Network", "Security", "Hardware",
-       "Infrastructure", "Licensing", "Communication", "RemoteWork",
-       "Training", "Performance"]
+# Departments mapping (example, can adjust for hackathon)
+DEPARTMENTS = ["Networking", "Hardware", "Software", "Security", "General IT"]
 
 # -------------------------------
 # Routing Endpoint
@@ -36,13 +40,11 @@ async def route_ticket(req: RoutingRequest):
     if not text:
         raise HTTPException(status_code=400, detail="Text cannot be empty")
     
-    # Tokenize
     inputs = tokenizer(text, return_tensors="pt", truncation=True)
     outputs = model(**inputs)
     logits = outputs.logits[0]
 
-    # Simple mapping: choose max logit index as department (demo)
-    # For a real hackathon, you may map labels more carefully
+    # Simple mapping: max logit → department
     department_idx = torch.argmax(logits).item() % len(DEPARTMENTS)
     department = DEPARTMENTS[department_idx]