Upload folder using huggingface_hub

app.py

@@ -4,9 +4,9 @@ from typing import List
 import gradio as gr
 import openai
 import pinecone
-from llama_index import VectorStoreIndex, StorageContext
+from llama_index import VectorStoreIndex, StorageContext, ServiceContext
 from llama_index.chat_engine.types import ChatMode
-from llama_index.llms import ChatMessage, MessageRole
+from llama_index.llms import ChatMessage, MessageRole, OpenAI
 from llama_index.vector_stores import PineconeVectorStore
 
 from environments import OPENAI_API_KEY, PINECONE_API_KEY, PINECONE_INDEX, PASSWORD, LOCAL
@@ -26,10 +26,13 @@ pinecone.init(
 )
 pinecone_index = pinecone.Index(PINECONE_INDEX)
 
+llm = OpenAI(temperature=0.0, model="gpt-3.5-turbo-instruct")
+service_context = ServiceContext.from_defaults(llm=llm)
+
 vector_store = PineconeVectorStore(pinecone_index=pinecone_index)
 storage_context = StorageContext.from_defaults(vector_store=vector_store)
-index = VectorStoreIndex.from_documents([], storage_context=storage_context)
-chat_engine = index.as_chat_engine(chat_mode=ChatMode.CONTEXT, similarity_top_k=2)
+index = VectorStoreIndex.from_documents([], storage_context=storage_context, service_context=service_context)
+chat_engine = index.as_chat_engine(chat_mode=ChatMode.CONTEXT, similarity_top_k=3)
 DENIED_ANSWER_PROMPT = '我是設計用於回答關於信義會地區中心的服務內容'
 
 SYSTEM_PROMPT = f'你是信義會地區中心的智能助理，你必須基於提供的資訊進行總結，用以回答用戶的提問。' \
@@ -54,7 +57,7 @@ CHAT_EXAMPLES = [
 
 
 def convert_to_chat_messages(history: List[List[str]]) -> List[ChatMessage]:
-    chat_messages = []
+    chat_messages = [ChatMessage(role=MessageRole.SYSTEM, content=SYSTEM_PROMPT)]
     for conversation in history[-1:]:
         for index, message in enumerate(conversation):
             if not message:
@@ -119,9 +122,9 @@ def _invoke_chatgpt(history, message, is_include_system_prompt=False):
     history_openai_format.append({"role": "user", "content": message})
 
     response = openai.ChatCompletion.create(
-        model='gpt-3.5-turbo',
+        model='gpt-3.5-turbo-instruct',
         messages=history_openai_format,
-        temperature=1.0,
+        temperature=0.0,
         stream=True
     )
     partial_message = ""
@@ -152,7 +155,6 @@ with gr.Blocks() as demo:
         gr.ChatInterface(predict_without_history, examples=CHAT_EXAMPLES)
     chatbot.like(vote, None, None)
 
-
 if LOCAL:
     demo.queue()
     demo.launch(share=False)