iris-rag-pipeline

name	iris-rag-pipeline
description	Build production RAG pipelines with IRIS native vector search. Use when asked to implement retrieval-augmented generation, semantic search, or document Q&A on IRIS.
license	MIT
metadata	{"version":"1.0.0","author":"InterSystems Developer Community","compatibility":"iris, python, langchain"}

Purpose

Set up end-to-end RAG pipelines using IRIS as the vector database — no external VectorDB required.

When to Use

• Building a Q&A system over enterprise documents stored in IRIS
• Implementing semantic search over ObjectScript class documentation or IRIS globals
• Adding AI-powered retrieval to an existing Ensemble/HealthConnect workflow

Workflow

Step 1 — Index Documents

from iris_vector_graph import IRISVectorStore

store = IRISVectorStore.connect(
    hostname="localhost", port=1972,
    namespace="USER", username="_SYSTEM", password="SYS"
)
store.add_texts(texts=chunks, metadatas=metadata_list)

Step 2 — Retrieve

results = store.similarity_search(query="your question", k=5)

Step 3 — Generate

Pass retrieved context to the LLM:

context = "\n".join(doc.page_content for doc in results)
prompt = f"Context:\n{context}\n\nQuestion: {query}"
response = llm.invoke(prompt)

IRIS Vector Index Setup

CREATE TABLE VectorStore (
    id INT IDENTITY PRIMARY KEY,
    content VARCHAR(32000),
    embedding VECTOR(FLOAT, 1536),
    metadata VARCHAR(2000)
);
CREATE INDEX hnsw_idx ON VectorStore (embedding) WITH (TYPE='HNSW', M=16, EF_CONSTRUCTION=200);

Common Patterns

• Hybrid search: combine BM25 + vector via iris_vector_graph.HybridSearch
• GraphRAG: use HybridGraphRAG for entity-relationship-aware retrieval
• RAGAS evaluation: use ragas package to score faithfulness, relevancy, context recall