mlx_localllm 0.2.12 copy "mlx_localllm: ^0.2.12" to clipboard
mlx_localllm: ^0.2.12 copied to clipboard

Published 12 hours ago
SDKFlutter
PlatformiOSmacOS
→

Metadata

A high-performance local LLM plugin for macOS using Apple's MLX framework.

More...

0.2.12 #

  • Comprehensive Documentation Update: Greatly expanded README.md with detailed API references, parameter tables, and a thinking mode guide. Added extensive DartDoc comments to lib/mlx_localllm.dart for better IDE support.

0.2.11 #

  • Improved Inference Logging: Implemented line-buffered logging for real-time inference monitoring with better readability.
  • Jinja Template Support: Added support for passing chat_template_kwargs to the model tokenizer, enabling dynamic control over model behavior.
  • Thinking Mode Fix: Injected add_generation_prompt by default and optimized output cleaning to ensure <think> tags and reasoning processes are correctly rendered.
  • Type-Safe Parameter Passing: Implemented robust type conversion for additionalContext in Swift to handle various input formats from Dart.

0.2.10 #

  • Fixed iOS SPM Support: Added ios/mlx_localllm/Package.swift to satisfy native Swift Package Manager requirements on iOS.
  • Unified Native Structure: Ensured both ios and macos directories have identical Swift Package structures for better compatibility.

0.2.9 #

  • iOS Support: Added full support for iOS (iPhone/iPad) using Apple's MLX framework.
  • x86 Architecture Support (Mocking): Added conditional compilation to support x86 (Intel Mac) builds. Inference is mocked to return "Unsupported architecture" while allowing the project to compile and link.
  • Refactored Native Code: Unified Swift code for both iOS and macOS through shared source files.
  • Improved Pub.dev Score: Fixed lint errors in tests and updated example project structure.

0.2.8 #

  • Fixed Native Errors: Fixed NativeMLXService missing members (unloadModel, checkModelExists, deleteModel).
  • Synced API Names: Unified MethodChannel and method names between Dart and Swift.
  • Stable Example: Verified that the example build succeeds on macOS ARM with SPM.

0.2.7 #

  • macOS ARM Focus: Optimized and stabilized specifically for Apple Silicon macOS. Dropped experimental iOS support to ensure maximum performance and reliability on desktop.
  • Full API Implementation: Finalized loadModel, unloadModel, generate, downloadModel (with progress stream), checkModelExists, and deleteModel.
  • State Management: Improved model lifecycle management with automatic caching and reference tracking.
  • EventChannel Support: Added robust event stream for real-time progress and status reporting during model downloads and loading operations.

0.2.6 #

  • Fixed SPM Repository URL: Corrected the Swift Package Manager repository URL for mlx-swift-lm.
  • Synced Versions: Synchronized version strings across pubspec.yaml and native .podspec files.

0.2.5 #

  • Fixed Dependency: Corrected MLX dependency repository from mlx-swift-llm to mlx-swift-lm. This ensures the latest MLX core features are correctly integrated.

0.2.4 #

  • Fixed SPM Detection: Restructured native directories to ios/mlx_localllm/ and macos/mlx_localllm/ as required by pub.dev score analyzer (pana). This ensures full native Swift Package Manager support is correctly recognized.
  • Code Formatting: Fixed Dart code formatting issues to satisfy pub.dev quality checks.

0.2.3 #

  • Fixed Documentation: Fixed an issue where the mockup image link was using an insecure local path. Now using a relative path compatible with pub.dev.

0.2.2 #

  • Full Swift Package Manager Support: Added Package.swift to both ios/ and macos/ directories. This fully satisfies native SPM requirements and improves the platform score on pub.dev.

0.2.1 #

  • Native SPM support: Added native Swift Package Manager support in pubspec.yaml. No manual Xcode configuration required for MLX dependencies.
  • Improved pub.dev platform score.

0.2.0 #

  • Added iOS support (iPhone, iPad).
  • Unified Swift codebase for iOS and macOS.
  • Set iOS deployment target to 17.0.
  • Updated example project with iOS project structure.

0.1.0 #

  • Initial release.
  • Support for local LLM inference on macOS (Apple Silicon).
  • Native MLX framework integration.
  • Robust model downloader with mirror support.
  • Progress reporting and download cancellation.
  • Model management (isDownloaded, deleteModel with error feedback).

← Metadata

1
likes
160
points
0
downloads

Publisher

unverified uploader

Weekly Downloads

Metadata

A high-performance local LLM plugin for macOS using Apple's MLX framework.

Repository (GitHub)
View/report issues

Documentation

API reference

License

MIT (license)

Dependencies

flutter, http, path, path_provider, plugin_platform_interface

More

Packages that depend on mlx_localllm

Packages that implement mlx_localllm

Back