from typing import List

import faiss
from langchain_community.docstore import InMemoryDocstore
from langchain_community.vectorstores import FAISS
from langchain_core.documents import Document
from langchain_huggingface import HuggingFaceEmbeddings



# 准备好向量化的对象
model_name = "BAAI/bge-small-zh-v1.5" # 模型名
model_kwargs = {'device': 'cpu'} # 没有显卡就用cpu，有英伟达显卡写cuda
encode_kwargs = {'normalize_embeddings': True} # set True to compute cosine similarity

#  第一次运行，会自动下载模型（去huggingface上下载），下载到hf默认的缓存目录。
hf_embedding = HuggingFaceEmbeddings(
    model_name=model_name,
    model_kwargs=model_kwargs,
    encode_kwargs=encode_kwargs
)
# 1、

# 把数据库写入磁盘
vector_store = FAISS.load_local('./faiss_db', embeddings=hf_embedding, allow_dangerous_deserialization=True)

vector_store.delete(ids=['id8'])

# results = vector_store.similarity_search('今天的金融投资新闻', k=2)
results = vector_store.similarity_search_with_score('有美食的内容吗', k=4, filter={"source": 'tweet'})  # 带分数
for res, score in results:
    print(type(res))
    print(res.id)
    print(f"* [Score={score:3f}] {res.page_content} [{res.metadata}]")