Create main.py

main.py

@@ -0,0 +1,48 @@
+import os
+from fastapi import FastAPI, HTTPException
+from pydantic import BaseModel
+from sentence_transformers import SentenceTransformer
+
+# Optimize for the 2 vCPU limit on the free tier
+os.environ["OMP_NUM_THREADS"] = "2"
+os.environ["MKL_NUM_THREADS"] = "2"
+
+app = FastAPI(title="Embedding API", description="Production-grade text embedding endpoint.")
+
+# Initialize the model at startup
+MODEL_ID = "Snowflake/snowflake-arctic-embed-l"
+
+try:
+    print(f"Loading {MODEL_ID} into memory...")
+    model = SentenceTransformer(MODEL_ID, device="cpu")
+    print("Model loaded successfully.")
+except Exception as e:
+    print(f"Failed to load model: {e}")
+    model = None
+
+class EmbeddingRequest(BaseModel):
+    texts: list[str]
+    is_query: bool = False
+
+@app.get("/health")
+def health_check():
+    if model is None:
+        raise HTTPException(status_code=503, detail="Model not loaded")
+    return {"status": "healthy", "model": MODEL_ID}
+
+@app.post("/embed")
+def generate_embeddings(request: EmbeddingRequest):
+    if model is None:
+        raise HTTPException(status_code=503, detail="Model initialization failed.")
+    
+    try:
+        # Snowflake architecture requires a specific prefix for search queries
+        if request.is_query:
+            inputs = [f"Represent this sentence for searching relevant passages: {text}" for text in request.texts]
+        else:
+            inputs = request.texts
+            
+        embeddings = model.encode(inputs, normalize_embeddings=True)
+        return {"embeddings": embeddings.tolist()}
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))