Connecting Oracle APEX to Google Gemini AI: A Step-by-Step Guide

Introduction

Oracle APEX has quietly made adding generative AI to low-code apps as simple as filling out a form. No middleware, no custom REST wrappers, no separate AI service to host. In this post, I'll walk through exactly how I connected Google Gemini to my APEX workspace using APEX's native Generative AI Services, and what it unlocks once it's set up.

Why this matters

Most teams assume "adding AI" means weeks of integration work and a dedicated ML engineer. Oracle APEX flips that: plugging in a foundation model is now a configuration task, not an engineering project. Once connected, AI capabilities like Builder AI, natural language search, and AI-assisted page generation work automatically — across every app in your workspace.

Prerequisites
An Oracle APEX workspace ( tested on 26.1)
A Google Gemini API key (Google AI Studio)
Workspace admin access to configure Generative AI Services
Step-by-step setup

1. Navigate to Generative AI Services
   From Workspace Utilities → Generative AI Services → Create/Edit.

2. Set the AI Provider
   Choose Google Gemini from the AI Provider dropdown, and give it a name (e.g. "Gemini API Key").

3. Configure the Base URL and Endpoint URL
   Both point to Gemini's API:

https://generativelanguage.googleapis.com/v1beta

1. Add credentials
   Toggle "Create Credentials" on, and paste in your Gemini API key. APEX stores this securely — you won't need to manage it in code anywhere.

2. Enable it for App Builder
   Toggle Used by App Builder and, optionally, Default for New Apps, so every new app in the workspace automatically uses Gemini.

3. Test the connection
   Click Test Connection to confirm APEX can reach Gemini before saving.

What this unlocks

Once configured, several AI-powered features become available immediately:

Builder AI — generate pages, forms, and reports from natural language descriptions
AI-assisted natural language search in Interactive Reports
AI-generated data models and content during app creation

**Model version compatibility

Not every Gemini model works out of the box with APEX's Generative AI Services. I found that Gemini Flash 3 and onward is what you need — earlier model versions either failed to respond correctly or weren't supported by APEX's integration layer. If you're setting this up and things aren't working, check your Model ID first before troubleshooting anything else.

API key scope
Make sure the API key you generate in Google AI Studio has access to the Generative Language API specifically — a key created for a different Google Cloud service won't authenticate correctly here.

Test Connection is your friend

Don't skip the Test Connection button before saving. It catches URL/endpoint mismatches and auth issues immediately, rather than surfacing as a vague error later inside Builder AI or natural language search.

Takeaway

Low-code and generative AI are converging faster than most people realize. If you're building on Oracle APEX and haven't tried wiring in a foundation model yet, it's a 10-minute setup that changes what your app builder can do.

Have you connected an LLM into APEX or another low-code platform? I'd love to hear what you built.