Rare find

LLM.swift. Run AI models directly on your iPhone, Mac, or Apple Watch without the internet.

github.com/eastriverlee/LLM.swift
by @eastriverlee

Vaya's read on this project

Problem, audience, market, and the verdict — sign in to see it.

Updates

July 2026
  • Release —v3.0.3
  • Merge pull request #85 from eastriverlee/settle-configuration-before-…
  • test: Cover configuration settling
  • fix: Forbid undeclared keys in @Generatable schemas
  • fix: Settle configuration changes before responding
  • Release —v3.0.2
  • Merge pull request #84 from JerryFZhang/fix/per-generation-interruption
  • refactor: Rename interruption epochs to generations
  • fix: scope interruption per generation so stop() can't strand or kill…
  • Release —v3.0.1
  • Merge pull request #82 from JerryFZhang/fix/mid-generation-cancellation
  • fix: Extend interruption to all generation paths
  • Merge branch 'main' into fix/mid-generation-cancellation
  • Release —v3.0.0
  • Merge pull request #83 from eastriverlee/64-support-function-calling
  • chore: Auto-update llama.cpp through a weekly workflow
  • chore: Distribute llama.xcframework as a remote binary target
  • feat: Add function calling via embedded template
  • docs: Document chat templates and function calling
  • feat: Vendor llama.cpp common chat sources
  • fix: make stop() and stream teardown actually cancel an in-flight gen…
  • Merge pull request #79 from DnV1eX/fix/huggingface-model-download
  • fix: update regex pattern for download URL extraction in HuggingFaceM…
December 2025
  • Release —v2.1.0
  • feat: Add thinking output separation for chain-of-thought models
  • add special tokens
  • update llama.xcframework
  • Release —v2.0.2
  • Merge branch 'fusiled/main'
  • fix test cases