openai embedding base search example

openai-embedding-base-search-example

Article search examples based on the embedding interface and GPT3.5 interface of OpenAI (of course, most of the code is given through ChatGPT, I only made some changes to encapsulation)

PS. Code is just a simple example. In fact, the embedding data that should be stored in the article using a vector database.

process

• Generate embedding data for all articles (only do it once)
• Generate embedding data for user input
• Through the similarity algorithm, we can find the first 3 articles with high similarity to user input
• Split all paragraphs of these 3 articles
• GPT model respectively. If the model gives a positive answer, the paragraph will be returned to the output.

run

• Need to install nodejs v18+
• Create a .env file in the root directory of this project and set the OpenAPI API KEY environment variable

OPENAI_API_KEY=sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
#如果需要使用代理访问，请提供下面的环境变量
# https_proxy=http://127.0.0.1:1080

• Install dependency npm install
• Run npm run dev

More references

• Domain knowledge questions and answers based on embedding: https://github.com/openai/openai-cookbook/blob/main/examples/Question_answering_using_embeddings.ipynb
• Generate OpenAI API KEY: https://platform.openai.com/account/api-keys