feat : use custom pg vector and gpt4all embeddings

app.py

@@ -1,18 +1,15 @@
-import os
-
 import streamlit as st
 import streamlit.components.v1 as components
 from css import load_css
+from custom_pgvector import CustomPGVector
 from langchain import OpenAI
 from langchain.callbacks import get_openai_callback
 from langchain.chains import ConversationalRetrievalChain
 from langchain.chains.conversation.memory import ConversationBufferMemory
-from langchain.embeddings.openai import OpenAIEmbeddings
-from langchain.vectorstores.pgvector import PGVector
+from langchain.embeddings import GPT4AllEmbeddings
 from message import Message
 
 CONNECTION_STRING = "postgresql+psycopg2://localhost/sorbobot"
-COLLECTION_NAME = ""
 
 
 def initialize_session_state():
@@ -21,15 +18,16 @@ def initialize_session_state():
     if "token_count" not in st.session_state:
         st.session_state.token_count = 0
     if "conversation" not in st.session_state:
-        embeddings = OpenAIEmbeddings()
+        embeddings = GPT4AllEmbeddings()
 
-        store = PGVector(
-            collection_name=COLLECTION_NAME,
-            connection_string=CONNECTION_STRING,
+        db = CustomPGVector(
             embedding_function=embeddings,
+            table_name="article",
+            column_name="abstract_embedding",
+            connection_string=CONNECTION_STRING,
         )
 
-        retriever = store.as_retriever()
+        retriever = db.as_retriever()
 
         llm = OpenAI(
             temperature=0,
@@ -39,7 +37,7 @@ def initialize_session_state():
 
         st.session_state.memory = ConversationBufferMemory()
         st.session_state.conversation = ConversationalRetrievalChain.from_llm(
-            llm=llm, retriever=retriever
+            llm=llm, retriever=retriever, verbose=True
         )