The advent of the era of large models has made AI application development easier and time-saving, especially under the paradigm of CVP Stack, developers can even make a complete application in a weekend.
This article will apply theory to practice, and show you how to use Milvus, Xinference, Llama 2-70B open source model and LangChain to build a full-featured question answering system. Xinference makes it simple and efficient to deploy the Llama 2 model locally, while Milvus provides efficient vector storage and retrieval capabilities.
Get rid of the dependence on OpenAI, build a full-process AI application with the help of the open source ecosystem, start now!
01.
Project Introduction
-
The kite
Milvus (https://milvus.io/docs/overview.md) is a vector database whose main function is to store, index and manage large-scale embedding vectors generated by deep neural networks and other machine learning models. Unlike traditional relational databases, Milvus specializes in querying input vectors and is capable of indexing trillion-scale vectors.
Milvus is designed from the ground up with special consideration for handling embedding vectors from unstructured data. With the development of the Internet, unstructured data such as emails, papers, sensor data, photos, etc. have become more and more common. In order for computers to understand and process these