Releases: rryam/FoundryKit
Releases · rryam/FoundryKit
FoundryKit v0.0.2
FoundryKit v0.0.2 - Beta 2 Compatibility Update
This release updates FoundryKit to be compatible with the new FoundationModels API changes in macOS 15.2 beta 2.
What's Changed
- Fixed Transcript API: Updated to work with the new collection-based Transcript API where
Transcript
itself is now a collection rather than having anentries
property - Added Rate Limiting Support: Added
.rateLimited
error case to properly handle rate limiting from Foundation Models - Fixed Validation Errors: Resolved naming conflict between validation error types by renaming internal macro validation errors
- Improved Error Handling: Enhanced error mapping between Foundation Models and FoundryKit error types
- Fixed Package URL: Corrected the package URL in README from rudrankriyam to rryam
Breaking Changes
- The Transcript API has changed in macOS 15.2 beta 2. If you were directly accessing
transcript.entries
, you can now usetranscript
directly as it's a collection
Compatibility
- Requires macOS 15.2 beta 2 or later
- Swift 6.0 or later
- Xcode 16.2 beta 2 or later
Installation
dependencies: [
.package(url: "https://github.com/rryam/FoundryKit", from: "0.0.2")
]
Notes
- SimpleJSONValidator has been temporarily disabled for this release while we update it to work with the new APIs
- Full structured generation features are still planned for a future release
Full Changelog: v0.0.1...v0.0.2
0.0.1
FoundryKit v0.0.1 - Initial Release
A unified Swift API for language model inference that seamlessly integrates MLX models and Apple's Foundation Models framework.
Features
- Unified API: Same interface for both MLX and Foundation Models
- Simple Text Generation: Clean, straightforward API for generating text responses
- Foundation Models Style: Uses familiar
Prompt
,Instructions
,Guardrails
- Model Selection: Easy switching between MLX and Foundation Models
- Generation Options: Control temperature, sampling, token limits, and penalties
- Full Swift 6 Support: Complete concurrency safety with Sendable conformance
What's New
- Initial release of FoundryKit
- Support for MLX models (Mistral, Llama, Qwen, etc.)
- Foundation Models framework integration
- Comprehensive test suite
- Example applications demonstrating usage
Installation
dependencies: [
.package(url: "https://github.com/rudrankriyam/FoundryKit", from: "0.0.1")
]
Coming Soon
- Structured generation with
@FoundryGenerable
macro - Response streaming
- Guided JSON generation
- Schema validation