mylovelycodes

mylovelycodes / LiteRTLM-Swift

Public

Swift package for running LiteRT-LM models on iOS. Wraps Google's C API in a clean, async/await Swift interface.

gemma-4-e2b-it-litert-lm litert litert-lm litert-lm-ios litert-next
10
3
100% credibility
Found Apr 16, 2026 at 10 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Swift
AI Summary

A Swift package providing an easy interface to run on-device AI models for text generation, image analysis, audio transcription, and multimodal tasks on iOS apps.

How It Works

1
📱 Discover smart AI for your iPhone app

You hear about a simple way to add thinking, seeing, and hearing powers to apps you build for iPhones.

2
🔧 Add it to your app project

In your app maker, you paste a web link and click add, so your app gets ready for AI magic.

3
⬇️ Download the AI brain

You grab the big thinking file once, watching a progress bar fill up nicely.

4
🚀 Wake up the AI

Tap to load it, and in seconds your app has a helpful brain ready to go.

5
đź’¬ Chat, describe photos, or listen to sounds

Type questions for answers, show pictures to get descriptions, or play audio for summaries – it all works smoothly.

✨ Your app comes alive with understanding

Now your app thinks, sees, and hears on its own, delighting users with smart responses every time.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 10 to 10 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is LiteRTLM-Swift?

This GitHub Swift repo delivers a Swift Package Manager-ready wrapper for Google's LiteRT-LM C API, letting iOS apps run on-device language models with a clean async/await interface. It handles text generation, vision for image understanding, audio transcription, and multimodal combos like audio-plus-image queries, pulling models like Gemma 4 E2B from HuggingFace via a built-in downloader. Developers get offline inference without wrestling raw C bindings or manual model prep.

Why is it gaining traction?

Unlike raw LiteRT-LM or heavier frameworks, it offers dead-simple SPM integration—add via Xcode or Package.swift—and persistent sessions that slash multi-turn latency via KV cache reuse, dropping time-to-first-token from 20s to 1-2s. Streaming APIs, SwiftUI-bindable progress tracking, and support for WAV/MP3 images in one call make prototyping fast, especially versus cloud-dependent swift github llm options. The async/await api feels native, bridging swift package dependencies to production-ready traits.

Who should use this?

iOS devs building chat apps, photo analyzers, or voice assistants needing offline AI—think AR filters describing scenes or podcasts summarized on-device. Perfect for indie hackers or teams dodging API costs in swift github projects, especially with iPhone 13 Pro+ hardware and increased-memory-limit entitlement. Skip if you're on older iOS or prefer full swift package index listings.

Verdict

Grab it for proofs-of-concept in swift package manager xcode flows—docs shine with copy-paste examples—but at 10 stars and 1.0% credibility score, treat as experimental; rebuild the XCFramework for updates. Solid start for local LLM on iOS, worth watching.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.