Update app.py

app.py

@@ -5,9 +5,14 @@ from sentence_transformers import SentenceTransformer
 from transformers import AutoTokenizer, AutoModelForMaskedLM
 from qdrant_client import models
 import logging
+import json
 
 # --- Setup Logging ---
-logging.basicConfig(level=logging.INFO)
+# Configure logging to be more descriptive
+logging.basicConfig(
+    level=logging.INFO,
+    format='%(asctime)s - %(levelname)s - %(message)s',
+)
 logger = logging.getLogger(__name__)
 
 # --- Configuration ---
@@ -53,14 +58,15 @@ async def load_models():
     This ensures models are loaded only once.
     """
     global dense_model, splade_tokenizer, splade_model
-    logger.info(f"Loading models onto device: {DEVICE}")
+    logger.info("Server is starting up... Time to load the ML models.")
+    logger.info(f"I'll be using the '{DEVICE}' for processing.")
     try:
         dense_model = SentenceTransformer(DENSE_MODEL_ID, device=DEVICE)
         splade_tokenizer = AutoTokenizer.from_pretrained(SPLADE_QUERY_MODEL_ID)
         splade_model = AutoModelForMaskedLM.from_pretrained(SPLADE_QUERY_MODEL_ID).to(DEVICE)
-        logger.info("Models initialized successfully.")
+        logger.info("Great news! All models have been loaded successfully.")
     except Exception as e:
-        logger.fatal(f"FATAL: Could not initialize models. Error: {e}")
+        logger.critical(f"Oh no, a critical error occurred while loading models: {e}", exc_info=True)
         # In a real-world scenario, you might want the app to fail startup if models don't load.
         raise e
 
@@ -107,27 +113,36 @@ async def vectorize_query(request: QueryRequest):
     Returns:
         A JSON response containing the dense and sparse vectors.
     """
-    logger.info(f"Received query for vectorization: '{request.query_text}'")
-
+    # --- n8n Logging ---
+    logger.info("=========================================================")
+    logger.info("A new request just arrived! Let's see what we've got.")
+    logger.info(f"The incoming search query from n8n is: '{request.query_text}'")
+    
     # 1. Generate Dense Vector
-    logger.info("Generating dense vector...")
+    logger.info("First, I'm generating the dense vector for semantic meaning...")
     dense_query_vector = dense_model.encode(request.query_text).tolist()
-    logger.info("Dense vector generated.")
+    logger.info("Done with the dense vector. It has %d dimensions.", len(dense_query_vector))
+    logger.info("Here's a small sample of the dense vector: %s...", str(dense_query_vector[:4]))
 
     # 2. Generate Sparse Vector
-    logger.info("Generating sparse vector...")
+    logger.info("Next up, creating the sparse vector for keyword matching...")
     sparse_query_vector = compute_splade_vector(request.query_text)
-    logger.info("Sparse vector generated.")
+    logger.info("Sparse vector is ready. It contains %d important terms.", len(sparse_query_vector.indices))
+    logger.info("Here's a sample of the sparse vector indices: %s...", str(sparse_query_vector.indices[:4]))
 
     # 3. Construct and return the response
-    return VectorResponse(
+    logger.info("Everything looks good. I'm packaging up the vectors to send back.")
+    logger.info("=========================================================")
+    
+    final_response = VectorResponse(
         dense_vector=dense_query_vector,
         sparse_vector=SparseVectorResponse(
             indices=sparse_query_vector.indices,
             values=sparse_query_vector.values
         )
     )
+    return final_response
 
 @app.get("/", include_in_schema=False)
 async def root():
-    return {"message": "Vector Generation API is running. -- VERSION 2 --"}
+    return {"message": "Vector Generation API is running. -- VERSION 2 --"}