Update app.py

app.py

@@ -9,21 +9,47 @@ from fastapi import FastAPI, HTTPException
 from pydantic import BaseModel
 from sentence_transformers import SentenceTransformer
 from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
-os.environ["HF_HOME"] = "/tmp/huggingface"
-os.environ["TRANSFORMERS_CACHE"] = "/tmp/huggingface"
-os.environ["SENTENCE_TRANSFORMERS_HOME"] = "/tmp/huggingface"
+
+# ✅ Set a writable cache directory inside the container
+os.environ["HF_HOME"] = "/app/cache"
+os.environ["TRANSFORMERS_CACHE"] = "/app/cache"
+os.environ["SENTENCE_TRANSFORMERS_HOME"] = "/app/cache"
+
 # ✅ Initialize FastAPI
 app = FastAPI()
 
-# ✅ Set OpenRouter API Key
+# ✅ Securely Fetch API Key
 OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY")
 if not OPENROUTER_API_KEY:
     raise ValueError("❌ OPENROUTER_API_KEY is missing. Set it as an environment variable.")
 
 OPENROUTER_BASE_URL = "https://openrouter.ai/api/v1/chat/completions"
-embedding_model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2", cache_folder="/app/cache")
-summarization_model = AutoModelForSeq2SeqLM.from_pretrained("google/long-t5-tglobal-base", cache_dir="/app/cache")
-summarization_tokenizer = AutoTokenizer.from_pretrained("google/long-t5-tglobal-base", cache_dir="/app/cache")
+
+# ✅ Load AI Models with explicit caching & remote code trust
+try:
+    embedding_model = SentenceTransformer(
+        "sentence-transformers/all-MiniLM-L6-v2",
+        cache_folder="/app/cache",
+        trust_remote_code=True  # ✅ Fix potential caching issues
+    )
+    summarization_model = AutoModelForSeq2SeqLM.from_pretrained(
+        "google/long-t5-tglobal-base",
+        cache_dir="/app/cache",
+        trust_remote_code=True  # ✅ Trust remote code
+    )
+    summarization_tokenizer = AutoTokenizer.from_pretrained(
+        "google/long-t5-tglobal-base",
+        cache_dir="/app/cache",
+        trust_remote_code=True
+    )
+    print("✅ Models Loaded Successfully!")
+except Exception as e:
+    print(f"❌ Model loading error: {e}")
+
+# ✅ API Health Check
+@app.get("/")
+def health_check():
+    return {"status": "FastAPI is running!"}
 
 # ✅ Load Datasets
 try: