What's Changed

• Expand test coverage by @mattt in #133
• Document Swift Package Manager trait resolution workaround in README by @mattt in #136
• Add UnavailableReason for MLXLanguageModel by @mattt in #138
• Serialize Linux URLSession request paths to mitigate _MultiHandle race by @mattt in #134
• Implement safe MLX KV reuse with scoped GPU cache eviction by @mattt in #147
• Update CI workflow to test with --traits AsyncHTTPClient by @mattt in #148
• Preserve named string enum dependencies in schema conversion by @mattt in #149
• Add additionalContext support to MLXLanguageModel by @noorbhatia in #145

Full Changelog: 0.7.1...0.8.0

Contributors

• @mattt
• @noorbhatia

What's Changed

• Make tool schema inclusion behavior consistent with Foundation Models by @mattt in #129
• Fix race condition in LanguageModelSession when streaming responses by @atdrendel in #126
• Make follow-up changes to Locked by @mattt in #131
• Fix incorrect formatting in OpenAI responses API for tool by @noorbhatia in #125

Full Changelog: 0.7.0...0.7.1

Contributors

• @mattt
• @atdrendel
• @noorbhatia

What's Changed

• Document guided generation support for MLX and Llama by @mattt in #106
• Add tool execution delegate by @mattt in #109
• Implement guided generation for AnthropicLanguageModel by @mattt in #110
• Implement guided generation for OpenAI models by @mattt in #111
• Implement guided generation for GeminiLanguageModel by @mattt in #115
• Add support for Open Responses by @mattt in #119
• Implement guided generation for CoreML by @mattt in #116
• Implement guided generation for OllamaLanguageModel by @mattt in #121
• Restructure usage section of README by @mattt in #120

Full Changelog: 0.6.0...0.7.0

Contributors

• @mattt

What's Changed

• Implement guided generation for SystemLanguageModel by @mattt in #59
• Fixed build errors under macCatalyst. by @chenyunguiMilook in #96
• Add custom JSON mode generation option for Gemini language model by @mattt in #84
• Fix running tests on macOS 15 by @atdrendel in #100
• Implement prewarm for MLXLanguageModel by @noorbhatia in #97
• Implement structured output generation for both LlamaLanguageModel / MLXLanguageModel by @eastriverlee in #75
• Fix simulator tests on Xcode 26 by @atdrendel in #101
• Fix tool message content format for DeepSeek API compatibility by @ajason in #99
• Add public initializers in LanguageModelSession by @mattt in #102

New Contributors

• @chenyunguiMilook made their first contribution in #96
• @atdrendel made their first contribution in #100
• @eastriverlee made their first contribution in #75
• @ajason made their first contribution in #99

Full Changelog: 0.5.4...0.6.0

Contributors

• @mattt
• @atdrendel
• @ajason
• @chenyunguiMilook
• @eastriverlee
• @noorbhatia

What's Changed

• Fix PartiallyGenerated for array properties by @ethanhuang13 in #88
• Fix session transcript not being passed into APIs by @nathanborror in #87
• Fix mlx-swift-lm dependency to use semantic versioning by @alexey1312 in #91
• Implement NSCache to cache MLX ModelContext by @noorbhatia in #90
• Add lint step to CI workflow by @mattt in #92

New Contributors

• @alexey1312 made their first contribution in #91

Full Changelog: 0.5.3...0.5.4

Contributors

• @mattt
• @nathanborror
• @ethanhuang13
• @alexey1312
• @noorbhatia

What's Changed

• Add custom JSON mode generation option for Gemini language model by @mattt in #80
• Add Identifiable conformance to PartiallyGenerated struct in GenerableMacro by @ethanhuang13 in #73
• Document @Generable macro visibility inside #Playground by @mattt in #81
• Update CI workflow to build on macOS with all traits enabled by @XMLHexagram in #72
• Revert "Add custom JSON mode generation option for Gemini language model" by @mattt in #83
• Update CI workflow to add --build-tests to build step by @mattt in #85

New Contributors

• @ethanhuang13 made their first contribution in #73

Full Changelog: 0.5.2...0.5.3

Contributors

• @mattt
• @ethanhuang13
• @XMLHexagram

What's Changed

• Update llama.swift dependency and rename import to LlamaSwift by @BarredEwe in #68
• Use convertTranscriptToMLXChat in MLXLanguageModel streamResponse by @noorbhatia in #67

New Contributors

• @BarredEwe made their first contribution in #68

Full Changelog: 0.5.1...0.5.2

Contributors

• @BarredEwe
• @noorbhatia

What's Changed

• Fix missing iOS availability annotations in SystemLanguageModel extensions by @westonwestenborg in #63
• Refactor message history construction from Transcription in MLXLanguageModel by @noorbhatia in #66
• Implement streaming response for MLXLanguageModel by @XMLHexagram in #64
• Pass entire transcript to Foundation Models session by @mattt in #61

New Contributors

• @westonwestenborg made their first contribution in #63
• @XMLHexagram made their first contribution in #64

Full Changelog: 0.5.0...0.5.1

Contributors

• @mattt
• @westonwestenborg
• @XMLHexagram
• @noorbhatia

What's Changed

• Add custom generation options for language models by @mattt in #50
• expose HubApi and model directory configuration for MLX by @tpae in #44
• Fix non-streaming respond() does not add .response() entry to transcript by @noorbhatia in #46
• Fix invalid construction of greedy sampler in LlamaLanguageModel by @mattt in #55
• Add configuration option for llama.cpp logging by @mattt in #54
• Update LlamaLanguageModel to replace implementation-specific properties with custom generation options by @mattt in #56
• Implement chat templates for LlamaLanguageModel by @mattt in #57
• Add logic to handle encoder-only llama models by @mattt in #53
• Improve migration for deprecated llama model APIs by @mattt in #58
• OpenAILanguageModel + SystemLanguageModel tools calling support by @airy10 in #49
• Pass instructions as system prompt for MLXLanguageModel by @noorbhatia in #48

New Contributors

• @tpae made their first contribution in #44
• @airy10 made their first contribution in #49

Full Changelog: 0.4.5...0.5.0

Contributors

• @mattt
• @airy10
• @tpae
• @noorbhatia

What's Changed

• Re-scope image validation method for Core ML adapter by @mattt in #42

Full Changelog: 0.4.4...0.4.5

Contributors

• @mattt