Update main.py

main.py

@@ -29,16 +29,6 @@ import datetime
 
 os.environ["TOKENIZERS_PARALLELISM"] = os.environ["TOKENIZERS_PARALLELISM"]
 os.environ['ANTHROPIC_API_KEY'] = os.environ['ANTHROPIC_API_KEY']
-index_name = os.environ['PINECONE_INDEX_NAME']
-embeddings = HuggingFaceEmbeddings()
-pinecone.init(
-    api_key=os.environ['PINECONE_API_KEY'],
-    environment=os.environ['PINECONE_ENVIRONMENT']
-)
-vectorstore = Pinecone.from_existing_index(
-    index_name=index_name, embedding=embeddings
-)
-retriever = vectorstore.as_retriever(search_type="similarity_score_threshold", search_kwargs={"score_threshold": .7, "k": 60,"filter": {'categorie': {'$eq': 'OF'}}})
         
 @cl.author_rename
 def rename(orig_author: str):
@@ -89,6 +79,20 @@ async def on_action(action):
     ]
     await cl.Message(author="🌐🌐🌐",content="Fermer le panneau d'information", actions=others).send()
 
+@cl.cache
+def retriever_to_cache():
+    index_name = os.environ['PINECONE_INDEX_NAME']
+    embeddings = HuggingFaceEmbeddings()
+    pinecone.init(
+        api_key=os.environ['PINECONE_API_KEY'],
+        environment=os.environ['PINECONE_ENVIRONMENT']
+    )
+    vectorstore = Pinecone.from_existing_index(
+        index_name=index_name, embedding=embeddings
+    )
+    retriever = vectorstore.as_retriever(search_type="similarity_score_threshold", search_kwargs={"score_threshold": .7, "k": 60,"filter": {'categorie': {'$eq': 'OF'}}})
+    return retriever
+    
 @cl.cache
 def to_cache(file):
     #time.sleep(5)  # Simulate a time-consuming process
@@ -170,7 +174,7 @@ async def start():
         qa = ConversationalRetrievalChain.from_llm(
             streaming_llm,
             chain_type="stuff",
-            retriever=retriever,
+            retriever=retriever_to_cache(),
             #combine_docs_chain=doc_chain,
             #question_generator=question_generator,
             memory=memory,