utils path

utils/coherence_bbscore.py

@@ -3,8 +3,6 @@ import math, re, unicodedata
 from typing import List, Dict, Any, Optional, Tuple
 import numpy as np
 import os, re, unicodedata, numpy as np
-# get the reranked results with no scores
-from retrieve_n_rerank import retrieve_and_rerank
 
 try:
     from sentence_transformers import SentenceTransformer
@@ -244,12 +242,14 @@ def coherence_assessment_std(
     }
 
 # Get the coherence report
-def coherence_report(embedder="MoritzLaurer/deberta-v3-base-zeroshot-v2.0",
+def coherence_report(embedder="BAAI/bge-m3",
                         input_text=None,
                         reranked_results=None,
                         run_zero_shot=True):
     embedder = Embedder(embedder) if isinstance(embedder, str) else embedder
     if reranked_results is None:
+        # Import here to avoid circular imports
+        from utils.retrieve_n_rerank import retrieve_and_rerank
         reranked_results = retrieve_and_rerank(input_text)
     if not reranked_results:
         return []

utils/generation_streaming.py

@@ -6,28 +6,30 @@ import time
 import json
 
 # encode the text
-from encoding_input import encode_text
+from utils.encoding_input import encode_text
 
 # rertrieve and rerank the documents
-from retrieve_n_rerank import retrieve_and_rerank
+from utils.retrieve_n_rerank import retrieve_and_rerank
 
 # sentiment analysis on reranked documents
-from sentiment_analysis import get_sentiment
+from utils.sentiment_analysis import get_sentiment
 
 # coherence assessment reports
-from coherence_bbscore import coherence_report
+from utils.coherence_bbscore import coherence_report
 
 # Get the vectorstore
-from loading_embeddings import get_vectorstore
-vectorstore = get_vectorstore()
+from utils.loading_embeddings import get_vectorstore
 
 # build message from model generation
-from model_generation import build_messages
+from utils.model_generation import build_messages
+import os
 
-API_KEY = "sk-do-8Hjf0liuGQCoPwglilL49xiqrthMECwjGP_kAjPM53OTOFQczPyfPK8xJc"
+API_KEY = os.getenv("API_KEY", "sk-do-8Hjf0liuGQCoPwglilL49xiqrthMECwjGP_kAjPM53OTOFQczPyfPK8xJc")
 MODEL = "llama3.3-70b-instruct"
 
 def generate_response_stream(query: str, enable_sentiment: bool, enable_coherence: bool):
+    # Initialize vectorstore when needed
+    vectorstore = get_vectorstore()
 
     
     

utils/retrieve_n_rerank.py

@@ -1,6 +1,6 @@
 # load the encoded text and vectorstore
-from encoding_input import encode_text
-from loading_embeddings import get_vectorstore
+from utils.encoding_input import encode_text
+from utils.loading_embeddings import get_vectorstore
 from sentence_transformers import CrossEncoder
 import numpy as np
 import faiss