debug

app.py

@@ -3,7 +3,6 @@ from fastapi import FastAPI, Depends, HTTPException
 import logging
 from pydantic import BaseModel
 from transformers import AutoTokenizer, AutoModel
-from sentence_transformers import models, SentenceTransformer
 from services.qdrant_searcher import QdrantSearcher
 from services.openai_service import generate_rag_response
 from utils.auth import token_required
@@ -46,7 +45,7 @@ access_token = os.getenv('QDRANT_ACCESS_TOKEN')
 if not qdrant_url or not access_token:
     raise ValueError("Qdrant URL or Access Token is not set. Please set the QDRANT_URL and QDRANT_ACCESS_TOKEN environment variables.")
 
-# Initialize the SentenceTransformer model with trust_remote_code using transformers
+# Load the model and tokenizer with trust_remote_code=True
 try:
     cache_folder = os.path.join(hf_home_dir, "transformers_cache")
     
@@ -54,18 +53,17 @@ try:
     tokenizer = AutoTokenizer.from_pretrained('nomic-ai/nomic-embed-text-v1.5', trust_remote_code=True)
     model = AutoModel.from_pretrained('nomic-ai/nomic-embed-text-v1.5', trust_remote_code=True)
 
-    # Wrap the model into a SentenceTransformer
-    word_embedding_model = models.Transformer(model_name_or_path='nomic-ai/nomic-embed-text-v1.5', model=model, tokenizer=tokenizer)
-    pooling_model = models.Pooling(word_embedding_model.get_word_embedding_dimension())
-    encoder = SentenceTransformer(modules=[word_embedding_model, pooling_model])
-    
-    logging.info("Successfully loaded the SentenceTransformer model.")
+    logging.info("Successfully loaded the model and tokenizer with transformers.")
 except Exception as e:
-    logging.error(f"Failed to load the SentenceTransformer model: {e}")
-    raise HTTPException(status_code=500, detail="Failed to load the SentenceTransformer model.")
+    logging.error(f"Failed to load the model: {e}")
+    raise HTTPException(status_code=500, detail="Failed to load the custom model.")
 
-# Initialize the Qdrant searcher
-searcher = QdrantSearcher(encoder, qdrant_url, access_token)
+# Function to embed text using the model
+def embed_texts(texts):
+    inputs = tokenizer(texts, padding=True, truncation=True, return_tensors="pt")
+    outputs = model(**inputs)
+    embeddings = outputs.last_hidden_state.mean(dim=1)  # Example: mean pooling
+    return embeddings
 
 # Define the request body models
 class SearchDocumentsRequest(BaseModel):
@@ -120,6 +118,10 @@ async def generate_rag_response_api(
             logging.error(f"Search documents error: {error}")
             raise HTTPException(status_code=500, detail=error)
 
+        # Example: Use custom embedding logic
+        # embeddings = embed_texts([hit['text'] for hit in hits])
+        # Use embeddings for further processing...
+
         response, error = generate_rag_response(hits, body.search_query)
         
         if error: