llama_index/example/build_index.py

import os
import logging
import sys
from llama_index import SimpleDirectoryReader, GPTSimpleVectorIndex

logging.basicConfig(stream=sys.stdout, level=logging.INFO)
logging.getLogger().addHandler(logging.StreamHandler(stream=sys.stdout))

os.environ['OPENAI_API_KEY'] = 'sk-xx'

def build_index():
    documents = SimpleDirectoryReader('data').load_data()
    # 按最大token数500来把原文档切分为多个小的chunk，每个chunk转为向量，并构建索引
    index = GPTSimpleVectorIndex(documents, chunk_size_limit=500)
    # 保存索引
    index.save_to_disk('index.json')

def query():
    # 加载索引
    new_index = GPTSimpleVectorIndex.load_from_disk('index.json')
    # 查询索引
    response = new_index.query("What did the author do in 9th grade?")
    # 打印答案
    print(response)

if __name__=='__main__':
    build_index()