Deploy Your Model

Overview

ZETIC.MLange provides a streamlined workflow for deploying your models to production. Choose the method that best fits your workflow:

Web Dashboard

Visual interface for quick deployment

CLI

Automated deployment for CI/CD pipelines

First-time setup required

Start with the Web Dashboard to complete initial setup, then switch to CLI for automation.

Core Concepts

Repository

Your workspace for model version control. Each repository manages multiple versions of a related model family.

username/my-model-repository
├── v1 (uploaded 2024-01-15)
├── v2 (uploaded 2024-02-20) ← default
└── v3 (uploaded 2024-03-10) ← latest

Key features:

Latest model is automatically selected by default
Set a specific version as default model
Version history management for all uploaded models

Model Identifier

Access your models using the format: username/repository_name

Default behavior: Always loads the latest model version

Custom default: Set a specific version as default in repository settings

// Automatically loads the latest or default version
val model = ZeticMLangeModel(
    context, 
    BuildConfig.PERSONAL_KEY, 
    "Steve/MobileCLIP2-image"
)

// Automatically loads the latest or default version
let model = ZeticMLangeModel(
    tokenKey: personalKey,
    name: "Steve/MobileCLIP2-image"
)

// Automatically loads the latest or default version
const model = await ZeticMLange.load(
  personalKey,
  "Steve/MobileCLIP2-image"
);

Deployment Status

Track your model's deployment progress through these stages:

Status	Description	Ready for Use
N/A	Repository created, awaiting first upload	❌
Failed	Error during upload or conversion	❌
Converting	Model uploaded, converting to on-device format	❌
Optimizing	Conversion complete, optimization in progress	✅ Testing only
Ready	Fully optimized and production-ready	✅ Production

Using Optimizing status

Models in Optimizing status are functional but not performance-optimized. Use for testing and development, not production deployments.

Authentication

Personal Key

Your authentication credential for accessing ZETIC.MLange APIs and services.

Generate and Copy Your Personal Key

Copy Personal Key

Navigate to Settings → Personal Access Tokens in the Web Dashboard
Click the Generate New Token button
Click the copy icon next to the generated token to copy it

Save this token securely

You won't be able to see it again!

Store securely

Add your key to local.properties (this file is git-ignored by default):

# local.properties
MLANGE_PERSONAL_KEY=your-personal-key-here

Then load it in your build.gradle.kts:

// app/build.gradle.kts
android {
    defaultConfig {
        val properties = Properties()
        properties.load(project.rootProject.file("local.properties").inputStream())
        
        buildConfigField(
            "String",
            "PERSONAL_KEY",
            "\"${properties.getProperty("MLANGE_PERSONAL_KEY")}\""
        )
    }
}

Create a Config.xcconfig file (add to .gitignore):

# Config.xcconfig
MLANGE_PERSONAL_KEY = your-personal-key-here

Then reference it in your Info.plist:

<!-- Info.plist -->
<key>MLangePersonalKey</key>
<string>$(MLANGE_PERSONAL_KEY)</string>

Use in your application

// Access via BuildConfig
val model = ZeticMLangeModel(
    context,
    BuildConfig.PERSONAL_KEY,
    "Steve/MobileCLIP2-image"
)

// Read from Info.plist
guard let personalKey = Bundle.main.object(
    forInfoDictionaryKey: "MLangePersonalKey"
) as? String else {
    fatalError("Personal key not found")
}

let model = ZeticMLangeModel(
    tokenKey: personalKey,
    name: "Steve/MobileCLIP2-image"
)

Security best practices

Never commit local.properties (Android) or .xcconfig (iOS) files to version control

Add them to .gitignore immediately:

# Android
local.properties

# iOS
Config.xcconfig
*.xcconfig

Rotate keys regularly
Revoke compromised keys immediately in the Web Dashboard

Deploy Your Model

Web Dashboard

CLI

On this page