Update app.py

app.py

@@ -8,36 +8,32 @@ from pydantic import BaseModel
 from transformers import pipeline, BertForSequenceClassification, BertTokenizer
 from nltk.sentiment.vader import SentimentIntensityAnalyzer
 
-# ---------------- INIT ----------------
+# ---------- Force writable cache locations (must match Dockerfile) ----------
+os.environ.setdefault("NLTK_DATA", "/data/nltk_data")
+os.environ.setdefault("HF_HOME", "/data/huggingface")
+os.environ.setdefault("TRANSFORMERS_CACHE", "/data/huggingface/transformers")
+os.environ.setdefault("HF_DATASETS_CACHE", "/data/huggingface/datasets")
+os.environ.setdefault("TMPDIR", "/data/tmp")
 
+# Also ensure nltk uses this path immediately
+nltk.data.path = [os.environ["NLTK_DATA"]] + nltk.data.path
 
-NLTK_DATA_PATH = os.path.join(os.path.dirname(__file__), "nltk_data")
-os.makedirs(NLTK_DATA_PATH, exist_ok=True)
-
-# Ensure VADER is available
+# ---------- NLTK VADER ----------
 try:
     nltk.data.find("sentiment/vader_lexicon")
 except LookupError:
-    nltk.download("vader_lexicon", download_dir=NLTK_DATA_PATH)
-
-# Add path manually so nltk can find it
-nltk.data.path.append(NLTK_DATA_PATH)
-
-nltk.data.path.append("./nltk_data")
-
+    # download into /data/nltk_data (writable)
+    nltk.download("vader_lexicon", download_dir=os.environ["NLTK_DATA"])
 
 vader = SentimentIntensityAnalyzer()
 
-# Emotion model
+# ---------- Models ----------
 emotion_model = pipeline("sentiment-analysis", model="tabularisai/multilingual-sentiment-analysis")
-
-# FinBERT Tone
 finbert = BertForSequenceClassification.from_pretrained("yiyanghkust/finbert-tone", num_labels=3)
 finbert_tokenizer = BertTokenizer.from_pretrained("yiyanghkust/finbert-tone")
 tone_labels = ["Neutral", "Positive", "Negative"]
 
-# FastAPI
-app = FastAPI(title="Sentiment • Emotion • Tone API", version="2.0.0")
+app = FastAPI(title="Sentiment • Emotion • Tone API", version="2.0.1")
 
 
 # ---------------- HELPERS ----------------