ReadyFlow overview: RAG Query Milvus [Technical Preview]

You can use the RAG Query Milvus [Technical Preview] ReadyFlow to query Milvus VectorDB with an embedded prompt.

This ReadyFlow generates and vectorizes a query from a user-specified question. It then performs a similarity search using one of five search metrics (L2, IP, COSINE, JACCARD, HAMMING) on the specified Milvus collection and returns the result with the shortest distance, extracting the specified fields. The flow then performs a lexical search based on the specified fields and returns all rows with the returned 'section', storing them as a text attribute. A prompt is created from the original user query and generated text document. The prompt is then passed to ChatGPT for an answer to the user query. A Milvus access token and an OpenAI API key are required to run this flow. Define a KPI on the failure_AskChatGPT connection to monitor failed responses.

RAG Query Milvus [Technical Preview] ReadyFlow details
Source	Custom prompt-query
Source Format	Natural language
Destination	Milvus
Destination Format	Vector DB