Part 1 of 3

Introduction to OCI Cloud & MySQL AI Studio

Foundations for the hands-on lab — what OCI provides, how MySQL HeatWave fits in, and what you will use in MySQL AI Studio during the workshop.

What You'll Cover

This opening section equips you with the conceptual background needed for the rest of the workshop. By the end you will understand the OCI services involved, the role of MySQL HeatWave as a unified data platform, and how MySQL Studio provides a single browser-based workspace for SQL, AI chat, and notebooks.

☁️

OCI Cloud Basics

Regions, compartments, VCNs, and the services used throughout this lab — Compute, Networking, Object Storage, and the MySQL Database Service.

MySQL HeatWave

A single managed service combining OLTP, real-time analytics, Lakehouse, AutoML, and GenAI — no ETL, no separate engines.

🧪

MySQL Studio

The new browser-based workspace in OCI: SQL editor, AI-assisted chat, Jupyter-compatible notebooks, and project-based organization — all in one place.

OCI & MySQL HeatWave at a Glance

MySQL HeatWave is a fully managed cloud database service available on OCI, AWS, and Azure. It extends the MySQL DB System with a massively parallel, in-memory HeatWave Cluster that accelerates analytics, machine learning, and generative AI workloads — all without moving data out of the database.

Fig 1 — MySQL HeatWave High-Level Architecture
OCI Region / VCN Client Apps ODBC / JDBC / SDKs MySQL Studio SQL · Chat · Notebooks MySQL Shell CLI / VS Code Plugin REST API HeatWave REST MySQL DB System Enterprise Edition · InnoDB Storage Engine OLTP Transactions · Data Masking · Encryption Group Replication · High Availability HeatWave Cluster In-Memory Hybrid Columnar Engine Scale-out up to 512 nodes Auto Query Offloading Analytics OLAP · Real-time AutoML Train · Predict · Explain GenAI LLMs · RAG · Vector Store Lakehouse Object Storage Queries OCI Object Storage OCI GenAI Service HeatWave Autopilot

Key architectural highlights: the HeatWave Cluster stores data in a hybrid columnar format in memory, enabling efficient analytical processing alongside the InnoDB row-store used for OLTP. Query offloading is automatic — the optimizer decides at runtime whether to execute on InnoDB or HeatWave. The Autopilot system provides ML-driven advisors for provisioning, data placement, query optimization, and error recovery.

MySQL HeatWave: One Service, Five Workloads

Unlike traditional cloud database stacks that require separate services for each workload type, MySQL HeatWave consolidates transactions, analytics, machine learning, generative AI, and lakehouse queries into a single managed service. This eliminates ETL pipelines, reduces operational complexity, and lowers cost.

CapabilityWhat It DoesKey Feature
OLTP Transactional workloads on InnoDB with Enterprise Edition security Group Replication, High Availability across 3 ADs
Analytics Real-time OLAP on transactional data — no ETL duplication Hybrid columnar engine, auto query offloading
AutoML Build, train, and explain ML models in-database ML_TRAIN, ML_PREDICT_TABLE, NL2ML
GenAI In-database LLMs, vector store, RAG, and NL2SQL ML_RAG, ML_GENERATE, HeatWave Chat
Lakehouse Query structured, semi-structured, and unstructured data in Object Storage Auto Parallel Load, Delta Lake support

MySQL HeatWave GenAI

HeatWave GenAI is the core technology we'll leverage throughout this workshop. It provides integrated, automated, and secure generative AI capabilities directly within the database — no separate vector database, no external LLM orchestration, and no data movement required.

Fig 2 — HeatWave GenAI: RAG Pipeline (Simplified)
Documents PDF · DOCX · HTML in Object Storage Auto Parallel Load Chunk + Embed multilingual-e5-small HeatWave Vector Store HNSW Index Hybrid Search User Query Natural Language Encode Query Same embedding model as vector store Similarity Search Top-K Retrieval LLM In-DB or OCI GenAI Llama 3.2 · Cohere Query + Context → Augmented Prompt Answer with citations

The beauty of HeatWave GenAI is that the entire RAG pipeline — document ingestion, chunking, embedding generation, vector storage, similarity search, prompt augmentation, and LLM inference — is orchestrated through standard SQL calls. A single CALL sys.ML_RAG() invocation handles retrieval-augmented generation end to end.

Key SQL routines you'll use: sys.ML_RAG() for RAG queries, sys.ML_GENERATE() for content generation, sys.ML_EMBED_ROW() for single-row embeddings, sys.heatwave_load() for ingesting documents into the vector store, and sys.NL_SQL() for natural-language-to-SQL conversion.

MySQL Studio: Your Workshop Workspace

MySQL Studio is the browser-based environment where you'll spend most of this workshop. It was designed to lower the barrier to entry for working with MySQL HeatWave's advanced capabilities by unifying SQL editing, AI-assisted chat, and Jupyter-compatible notebooks in a single project-based workspace.

Fig 3 — MySQL Studio Components
MySQL Studio Browser-Based Workspace SQL Editor Schema browser · Explain Plan Charts · NL2SQL AI Chat RAG-enabled conversations Ask Oracle · Vector Store select Notebooks Jupyter-compatible Python · SQL · Ibis DataFrames Project workspace — SQL scripts, chat sessions, notebooks, and data files auto-saved together

Studio connects directly to your MySQL DB System in OCI. All artifacts are organized into projects: SQL scripts, chat sessions, notebooks, and supporting files are stored together and automatically saved. This means you can switch between writing SQL, chatting with your data, and running Python analysis without leaving the browser.

📝

SQL Editor

Write and run SQL with schema browsing, syntax highlighting, Explain Plan visualization, and results rendered as tables or charts. Natural Language to SQL converts plain English into executable queries.

💬

AI Chat

RAG-enabled conversations where you can select vector stores and ask questions grounded in your proprietary documents. "Ask Oracle" provides contextual help for SQL and database topics.

📓

Notebooks

Jupyter-compatible notebooks with Python, SQL, and Ibis DataFrame support. Ready-to-run guides cover AutoML, Lakehouse patterns, RAG workflows, and interactive visualizations.

Supported LLMs & Embedding Models

HeatWave GenAI supports both in-database LLMs (running directly on HeatWave nodes) and external models accessed via OCI Generative AI Service or Amazon Bedrock. Here are the key models available as of late 2025:

ModelTypeContextNotes
llama3.2-3b-instruct-v1 In-DB LLM Default Default LLM for content generation
llama3.1-8b-instruct-v1 In-DB LLM 8B params Higher quality, more compute
mistral-7b-instruct-v3 In-DB LLM 7B params Alternative in-database option
Cohere / Meta via OCI GenAI External Varies Pre-trained foundational models, VLMs
multilingual-e5-small Embedding Default Default embedding model for vector store

Workshop tip: GenAI features are available at no additional cost for MySQL HeatWave users, including on the HeatWave.Free shape. For this workshop we'll primarily use the default in-database LLM and embedding model.

Workshop Flow Preview

Here's what you'll build across the three sections:

Fig 4 — Workshop Journey
Part 1 Introduction OCI · HeatWave · Studio ← You are here Part 2 MySQL Studio Tour SQL · Vectors · GenAI queries Hands-on exploration Part 3 Build RAG Agent Embeddings · Vector Index Custom Tools · Chatbot

What You Should Be Ready For

After completing this introduction you should understand:

  • The OCI services involved in this workshop (MySQL Database Service, Object Storage, Networking)
  • How MySQL HeatWave unifies OLTP, analytics, ML, GenAI, and Lakehouse in a single service
  • The role of HeatWave GenAI — in-database LLMs, vector store, RAG, and NL2SQL
  • What MySQL Studio is and how its three modalities (SQL, Chat, Notebooks) work together
  • Where to click and what resources are available for the hands-on labs ahead