Using IRIS as a vector database

InterSystems IRIS embedded vector search capabilities lets us search unstructured and semi-structured data. Data is converted to vectors (also called ‘embeddings’) and then stored and indexed in InterSystems IRIS for semantic search, retrieval-augmented generation (RAG), text analysis, recommendation engines, and other use cases.

This is a simple demo of IRIS being used as a vector database and similarity search on IRIS.

Prerequisites:

1. Python
2. InterSystems IRIS for Health - as it will be used as the vector database

Repository: https://github.com/piyushisc/vectorsearchusingiris

Steps to follow:

1. Clone the repo.
2. Open VS Code, connect to desired instance and namespace of IRIS and compile the classes.
3. Open IRIS Terminal and invoke the command do ##class(vectors.vectorstore).InsertEmbeddings(), which reads the text from the file text.txt and generate embeddings and store them in IRIS.
4. Invoke the command do ##class(vectors.vectorstore).VectorSearch("search_terms") with desired words to perform similarity search. IRIS will return top three closest match: