AI Factory Quickstart v1.3

Hub quick links: Gen AI — Pipelines — Model Serving

What you’ll build

This quickstart gets you from zero to a working Gen AI assistant that answers questions from a Knowledge Base, plus a deployed model endpoint you can call from applications.

Prerequisites

• Prepare your deployment's GPU: Setup GPU
• Familiarity with the Gen AI concepts and Pipelines: Gen AI, Pipelines

Step 1 — Create a Knowledge Base

Create a Knowledge Base and ingest your content so the assistant can retrieve context:

• Create a Knowledge Base
• (Optional) Review Knowledge Base concepts: Knowledge Bases (explained)

Step 2 — Create an Assistant

Build a conversational assistant and attach your Knowledge Base:

• Create an Assistant
• (Optional) Review assistants: Assistants

Step 3 — Deploy a model endpoint

Deploy a model using KServe so your applications can call it directly:

• Deploy NIM containers
• Create an InferenceService
• (Optional) Customize runtime: Configure ServingRuntime

Step 4 — Call the endpoint from an app

Test your deployed model with a simple client and learn how to reach endpoints:

• Quickstart (Python)
• Access KServe endpoints

Step 5 — Monitor and iterate

Keep an eye on model health and performance:

• Monitor InferenceService
• (Optional) Tune Knowledge Base performance: Knowledge Base performance

Next steps

• Learn progressively: Learning paths
• Explore patterns: Use cases and Solutions
• Deep dive into the APIs and reference material via the Gen AI and Model Serving hubs