[fix]: Fixed code issues

src/gpp.py

@@ -19,7 +19,8 @@ from typing import List, Dict, Any, Optional
 import re
 
 from src import EmbeddingConfig, GPPConfig
-from src.utils import OpenAIEmbedder, LLMClient, logger
+from src.utils import OpenAIEmbedder, LLMClient
+from src import logger
 
 def parse_markdown_table(md: str) -> Optional[Dict[str, Any]]:
     """

src/qa.py

@@ -10,6 +10,7 @@ Each component is modular and can be swapped or extended (e.g., add HyDE retriev
 """
 import os
 from typing import List, Dict, Any, Tuple
+import streamlit as st
 
 from src import RerankerConfig, logger
 from src.utils import LLMClient
@@ -19,13 +20,18 @@ class Reranker:
     """
     Cross-encoder re-ranker using a transformer-based sequence classification model.
     """
+    @staticmethod
+    @st.cache_resource(show_spinner="Loading reranker model...")
+    def load_model_and_tokenizer(model_name, device):
+        from transformers import AutoTokenizer, AutoModelForSequenceClassification
+        tokenizer = AutoTokenizer.from_pretrained(model_name)
+        model = AutoModelForSequenceClassification.from_pretrained(model_name)
+        model.to(device)
+        return tokenizer, model
+
     def __init__(self, config: RerankerConfig):
         try:
-            from transformers import AutoTokenizer, AutoModelForSequenceClassification
-            import torch
-            self.tokenizer = AutoTokenizer.from_pretrained(config.MODEL_NAME)
-            self.model = AutoModelForSequenceClassification.from_pretrained(config.MODEL_NAME)
-            self.model.to(config.DEVICE)
+            self.tokenizer, self.model = self.load_model_and_tokenizer(config.MODEL_NAME, config.DEVICE)
         except Exception as e:
             logger.error(f'Failed to load reranker model: {e}')
             raise

src/retriever.py

@@ -1,18 +1,23 @@
 import os
 from typing import List, Dict, Any
+import streamlit as st
 
-from src.config import RetrieverConfig
-from src import logger  # Use logger from src/__init__.py
+from src import RetrieverConfig, logger
 
 class Retriever:
     """
     Hybrid retriever combining BM25 sparse and dense retrieval (no Redis).
     """
+    @staticmethod
+    @st.cache_resource(show_spinner="Loading embedding model...")
+    def load_embedder(model_name):
+        from sentence_transformers import SentenceTransformer
+        return SentenceTransformer(model_name)
+
     def __init__(self, chunks: List[Dict[str, Any]], config: RetrieverConfig):
         # Lazy import heavy libraries
         import numpy as np
         import hnswlib
-        from sentence_transformers import SentenceTransformer
         from rank_bm25 import BM25Okapi
         self.chunks = chunks
         try:
@@ -21,7 +26,7 @@ class Retriever:
                 raise ValueError("Chunks must be a list of dicts.")
             corpus = [c.get('narration', '').split() for c in chunks]
             self.bm25 = BM25Okapi(corpus)
-            self.embedder = SentenceTransformer(config.DENSE_MODEL)
+            self.embedder = self.load_embedder(config.DENSE_MODEL)
             dim = len(self.embedder.encode(["test"])[0])
             self.ann = hnswlib.Index(space='cosine', dim=dim)
             self.ann.init_index(max_elements=len(chunks))

src/utils.py

@@ -6,7 +6,7 @@ import openai
 from typing import List
 from openai import AzureOpenAI
 from langchain_openai import AzureOpenAIEmbeddings
-from src import logger  # Import logger from src/__init__.py
+from src import logger
 
 
 class LLMClient: