Objective

The Agentix Playground is a sample iOS app that demonstrates how to integrate mimik AI into your iOS projects. It allows you to download and run language models directly on your device—enabling powerful, private, offline-capable AI through a hybrid edge-cloud architecture.

The app also supports activating online services (e.g., Google Gemini) and using them for prompt generation or response validation. You can mix and match on-device and cloud services as needed.

When running on Apple Silicon Macs, the app unlocks additional capabilities such as support for Vision-Enabled Language Models. You can upload images and receive intelligent visual feedback—within the same Swift-based codebase.

The latest versions now include:

Built-in authentication with your mimik Developer Console account.
Automatic token retrieval after login (no manual copy/paste).
Automatic project creation (“Agentix Playground”) in your Developer Console (with approval).
For developers compiling from source: runtime license setup and customization remain available.

This example is built using ClientLibrary and ClientRuntime, part of the mimik Client Library v5.10.0. This unified abstraction layer simplifies working with AI models across device, edge, and cloud environments, supporting real-time streaming and discovery.

Why this platform matters

For Developers (Powerful)	For Managers (Magical)
Single code path – Interact with both on-device and cloud-hosted AI models using the same API, reducing branching logic and simplifying maintenance.	One simple workflow – Build once, and your app taps into both on-device and cloud AI without extra effort.
Composable model execution – Chain on-device and online models together for validation, fallback, or multi-step inference without writing custom glue code.	Smarter together – Seamlessly combine the strengths of local and online AI for faster, more reliable results.
Offline-first on-device models – Local models run without any network dependency, no background telemetry, and no data leaving the user’s phone.	Always available offline – On-device AI keeps working with no internet required and no hidden data sharing.
Privacy by default – On-device inference is fully local; no hidden telemetry or background data transfer.	Built-in trust – User data stays private by default, making compliance and customer confidence effortless.
Vision VLM support – Natively run and integrate vision-language models for multimodal apps (image + text understanding) without custom hacks.	Next-gen AI experiences – Power products with image + text intelligence, from smart assistants to rich content understanding.
Capabilities beyond Apple’s APIs – Provides features not supported by Core ML or Apple’s frameworks (e.g., direct chaining of heterogeneous models, hybrid execution strategies).	Unleashes new possibilities – Unlocks capabilities that Apple’s tools don’t allow, giving your product a competitive edge.

Getting Started

🆕 Authentication with your mimik Developer Console account is now required when the app launches.
The built-in UI flow will guide you through signing up or logging in.

⚠️ Vision-enabled features are available only on Apple Silicon Macs.

🚀 Option 1: Try It with TestFlight

Prefer not to build the app yourself?

Open this TestFlight link on your iOS device.

Accept the invitation and install the app.

🛠️ Option 2: Compile It Yourself

Prerequisites

You must use a physical iOS device connected to your Mac.

The iOS Simulator is not supported due to native dependencies.

Alternatively, you can run the app on an Apple Silicon Mac.

Get the Code

Clone the project and open it in Xcode:

git clone https://github.com/mimikgit/mimik-ai-agentix-playground-example-iOS.git

Install Dependencies (CocoaPods)

The project uses two custom CocoaPods:

These are already declared in the Podfile.

Step 1: Navigate to the Source Directory

cd mimik-ai-agentix-playground-example-iOS/Source/

Step 2: Install Pods

pod install --repo-update

Configure Runtime Credentials

🆕 Developer ID Tokens are retrieved automatically after login.
Manual copy/paste of config-developer-id-token is no longer required.

Runtime License

For TestFlight builds, the runtime license is built in automatically.
For source builds, you must still copy/paste your runtime license manually from developer console:

open config-developer-runtime-license

This step will be automated in a future release with a new API call.
Developer console tutorial

API Key

open config-developer-api-key

Paste your API key to authenticate specific app calls if needed.

Model Configurations

open Model-Configs/config-model-04-gemma2b.json; open Model-Configs/config-model-05-gemma11.json

These JSON files define model download URLs and settings. You can leave them as-is unless you're customizing models.

Launch in Xcode

Open the workspace:

open mimik-agentix.xcworkspace

And run on either a physical iOS device or Apple Silicon Mac.

Running the Example on Devices

iOS Devices

Connect a physical iOS device and select it in Xcode as the run target. For optimal performance, use a device capable of running AI models. Simulator is not supported.

Run the app and follow the on-device prompts to continue.

⚠️ iOS Simulators are not supported, use a physical device.

Apple Silicon Macs

Apple Silicon Macs can run iOS apps natively on macOS, including the iOS mim OE binary.

Select My Mac (Designed for iPad) in Xcode as the run target.

Run the app and follow the on-device prompts to continue.

📚 Learn more: Adapting iOS code to macOS (Apple Silicon)

Getting Started in the App

After launching the app, you’ll first be prompted to sign in with your mimik Developer Console account.

Options include Sign in, Sign up, or Reset Password.
A new project (Agentix Playground) can be automatically created in your Developer Console (after user approval).
Authentication is required on every app relaunch.

Developers compiling from source may extend the code to persist login if desired.