Introduction

The User Guide describes the tools you work with once an administrator has deployed the SAS Agentic AI Accelerator - see the Administration Guide for that part.

• Model Definition Builder - add an LLM or embedding model to the environment with one definition.yaml and the mdb command line, check it locally, and register and publish it to SAS Model Manager and the SAS Container Runtime.
• Prompt Builder - test a prompt across several LLMs at once, let a judge model rank the responses, keep a versioned experiment history in SAS Model Manager, optimize the prompt with DSPy, and turn the best prompt into a scoreable model for SAS Intelligent Decisioning. The same tool designs decision templates for decision models and, as its third kind, RAG setups - the RAG Builder is part of it.
• RAG - turn a folder of documents into a governed vector-store collection with the RAG custom steps or with a RAG setup in the Prompt Builder, ask it questions, and publish the retrieval as a model a decision can call.

The SAS code, custom steps and Python samples behind these tools are listed in the repository's SAS Viya Integrations and Non-SAS Viya Integrations folders.