Update app.py

app.py

@@ -1,10 +1,9 @@
 import os
 from fastapi import FastAPI
-from pydantic import BaseModel
+import uvicorn
 from sentence_transformers import SentenceTransformer
 from threading import Thread
 from typing import Optional
-from fastapi.responses import JSONResponse
 
 # Initialize FastAPI
 app = FastAPI(title="Sentence Transformer API")
@@ -13,40 +12,46 @@ app = FastAPI(title="Sentence Transformer API")
 model: Optional[SentenceTransformer] = None
 
 def load_model():
-    """Load the embedding model in background"""
+    """Background thread for model loading"""
     global model
     try:
         model = SentenceTransformer('bge-small-zh')
-        print("✅ Model loaded successfully")
+        print("Model loaded successfully")
     except Exception as e:
-        print(f"❌ Model loading failed: {str(e)}")
+        print(f"Model loading failed: {str(e)}")
 
-# Start loading model in background
+# Start model loading immediately
 Thread(target=load_model, daemon=True).start()
 
-class TextInput(BaseModel):
-    text: str
-
 @app.get("/")
 def health_check():
-    """Health check endpoint"""
+    """Required health check endpoint"""
     return {
         "status": "ready",
         "model_loaded": model is not None
     }
 
 @app.post("/embed")
-async def embed_text(input: TextInput):
-    """Embed text and return vector"""
+async def embed_text(text: str):
     if model is None:
-        return JSONResponse(status_code=503, content={"error": "Model is still loading"})
+        return {"error": "Model still loading"}, 503
     
-    try:
-        embedding = model.encode([input.text])[0]
-        return {
-            "text": input.text,
-            "embedding": embedding.tolist(),
-            "dimension": len(embedding)
-        }
-    except Exception as e:
-        return JSONResponse(status_code=500, content={"error": str(e)})
+    # Convert text to list if single string
+    input_text = [text] if isinstance(text, str) else text
+    
+    embeddings = model.encode(input_text)
+    
+    return {
+        "text": text,
+        "embedding": embeddings.tolist(),
+        "dimension": len(embeddings[0]) if isinstance(embeddings, list) else len(embeddings)
+    }
+
+if __name__ == "__main__":
+    port = int(os.environ.get("PORT", 7860)) 
+    uvicorn.run(
+        "app:app",
+        host="0.0.0.0",
+        port=port,
+        reload=False  # Disable auto-reload in production
+    )