Oracle · Software · YoctoIT tech page

Database 23ai

The latest generation of the database: AI Vector Search, JSON Relational Duality and 300+ new features — the foundations for tomorrow's apps.

FOCUS · THE DATABASE THAT EVOLVESVectors next to the tables: AI inside the database, not beside it
Request the Free Assessment Learn more
YoctoIT material for clients and partners · Oracle, Exadata, GoldenGate and the other products mentioned are trademarks of Oracle Corporation e/o sue affiliate.
01 · What it is

Oracle Database 23ai, made clear.

23ai is the long-term release bringing AI into the engine: Vector Search for semantic searches next to relational data, JSON Relational Duality reconciling documents and tables, True Cache, a native SQL firewall. For those on 19c, it's the destination of the next upgrade — to be planned now.

Vector
embeddings and similarity search in the database: RAG without exporting data
Duality
the same data seen as JSON and as tables: the dilemma is over
LTS
the long-support release: the right destination after 19c
Oracle Database 23ai
OFFICIAL ORACLE BRANDING · DATABASE 23AI
MATERIALE UFFICIALE · ORACLE DATABASE 23AI · FONTE: ORACLE BLOG
OFFICIAL MATERIAL · ORACLE DATABASE 23AI · SOURCE: ORACLE BLOG
02 · How to use it well

The things that make the difference.

The new features that matter

Classic applications + AIold and new, same engine
Vector Search
JSON Duality
SQL Firewall
semantics · flexibility · security
AutoUpgrade from 19cthe assisted path
On-prem, ODA, ExaCC, OCIit runs wherever you are
One engine, two eras of applications

A planned upgrade

19c is leaving support: the window to climb to 23ai is designed now, with rehearsals and rollback.

RAG on your own data

Company documents vectorized INSIDE the database: security and permissions already solved.

Simplified development

Duality views and native JSON: modern developers served without a NoSQL on the side.

Security as standard

Native SQL Firewall and data redaction: the database that defends itself.

03 · In depth

AI Vector Search and the new features that matter

Oracle Database 23ai is the long-term release: AI Vector Search brings embeddings and similarity search into the database (VECTOR datatype, HNSW/IVF indexes, distance functions in SQL), JSON Relational Duality makes the same data live as documents and as tables, True Cache adds a consistent read cache in front of the primary. SQL Firewall enters the kernel, schema privileges simplify the grants, and the native boolean closes a twenty-year wait.

  • VECTOR datatype — embeddings next to relational data: RAG without a second database
  • Indici HNSW/IVF — fast similarity search in memory or on storage: chosen by use case
  • JSON Duality — same data, document view and relational view: the dilemma is over
  • True Cache — consistent SQL cache without application logic: reads offloaded
  • SQL Firewall — a whitelist of allowed SQL in the kernel: injection dies right there
  • Long-term — the release to stand on: premier until 2029+, direct upgrade from 19c
04 · Numbers and lifecycle

The numbers that matter.

2029+
premier support: the base for the coming years
19c→23ai
the supported direct upgrade: a single jump
SQL
one language for relational, JSON, graph and vectors
0
apps to rewrite for Duality: you add, you don't migrate
23ai is adopted with method: an assessment of the 19c estate, an upgrade plan and the AI Vector cases identified — before the business asks for them.
05 · Use cases

Where it really pays off.

App modernization

The new features without changing platform.

AI on company data

Semantic search on contracts, tickets, catalogs.

Version consolidation

The estate aligned on the LTS, supported for long.

The database you already have can do new things: the 23ai upgrade is the right project for 2026.