Spaces:
Sleeping
Sleeping
File size: 4,006 Bytes
a56a0d0 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 |
import streamlit as st
import os
import pickle
import faiss
import logging
from multiprocessing import Lock
from multiprocessing.managers import BaseManager
from llama_index.callbacks import CallbackManager, LlamaDebugHandler
from llama_index import VectorStoreIndex, Document,Prompt, SimpleDirectoryReader, ServiceContext, StorageContext, load_index_from_storage
from llama_index.chat_engine import CondenseQuestionChatEngine;
from llama_index.node_parser import SimpleNodeParser
from llama_index.langchain_helpers.text_splitter import TokenTextSplitter
from llama_index.constants import DEFAULT_CHUNK_OVERLAP
from llama_index.response_synthesizers import get_response_synthesizer
from llama_index import ListIndex, SimpleWebPageReader,TrafilaturaWebReader
import tiktoken
from logging import getLogger, StreamHandler, Formatter
custom_prompt = Prompt("""\
以下はこれまでの会話履歴と、ドキュメントを検索して回答する必要がある、ユーザーからの会話文です。
会話と新しい会話文に基づいて、検索クエリを作成します。回答は日本語で行います。
新しい会話文が挨拶の場合、挨拶を返してください。
新しい会話文が質問の場合、検索した結果の回答を返してください。
答えがわからない場合は正直にわからないと回答してください。
会話履歴:
{chat_history}
新しい会話文:
{question}
Search query:
""")
chat_history = []
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger("__name__")
logger.debug("調査用ログ")
st.title("💬 ChatbotWebRead")
if "webmessages" not in st.session_state:
st.session_state["webmessages"] = [{"role": "assistant", "content": "お困りごとはございますか？"}]
for msg in st.session_state.webmessages:
st.chat_message(msg["role"]).write(msg["content"])
if prompt := st.chat_input():
st.session_state.webmessages.append({"role": "user", "content": prompt})
st.chat_message("user").write(prompt)
response = st.session_state.web_chat_engine.chat(prompt)
msg = str(response)
st.session_state.webmessages.append({"role": "assistant", "content": msg})
st.chat_message("assistant").write(msg)
if st.button("リセット",use_container_width=True):
st.session_state.web_chat_engine.reset()
st.session_state.webmessages = [{"role": "assistant", "content": "お困りごとはございますか？"}]
logger.info("reset")
def initialize_webindex():
logger.info("initialize_web_index start")
text_splitter = TokenTextSplitter(separator="。", chunk_size=1500
, chunk_overlap=DEFAULT_CHUNK_OVERLAP
, tokenizer=tiktoken.encoding_for_model("gpt-3.5-turbo").encode)
node_parser = SimpleNodeParser(text_splitter=text_splitter)
service_context = ServiceContext.from_defaults(node_parser=node_parser)
d = 1536
k=2
faiss_index = faiss.IndexFlatL2(d)
# デバッグ用
llama_debug_handler = LlamaDebugHandler()
callback_manager = CallbackManager([llama_debug_handler])
service_context = ServiceContext.from_defaults(callback_manager=callback_manager)
webDocuments = SimpleWebPageReader(html_to_text=True).load_data(
["https://www.stylez.co.jp/"]
)
logger.info(webDocuments)
webIndex = ListIndex.from_documents(webDocuments,service_context=service_context)
response_synthesizer = get_response_synthesizer(response_mode='compact')
webQuery_engine = webIndex.as_query_engine(response_synthesizer=response_synthesizer)
st.session_state.web_chat_engine = CondenseQuestionChatEngine.from_defaults(
query_engine=webQuery_engine,
condense_question_prompt=custom_prompt,
chat_history=chat_history,
verbose=True
)
if __name__ == "__main__":
# init the global index
logger.info("main start")
if "web_chat_engine" not in st.session_state:
initialize_webindex()
logger.info("initializing index...") |