edge_gen_ai 0.3.1

A Flutter plugin for on-device generative AI, using Apple's Foundation Models (iOS) and Gemini Nano via ML Kit GenAI (Android).

0.3.1

• Fixes Android builds when the plugin is installed from pub.dev by removing its dependency on an unpublished, machine-local android/local.properties file.

0.3.0

• Adds tool (function) calling through EdgeGenAITool.

0.2.0

• Adds Swift Package Manager support for iOS (in addition to CocoaPods).
• Migrates the Android build to built-in Kotlin on Android Gradle Plugin 9+, while still applying the kotlin-android plugin on older AGP versions, so the minimum supported Flutter/Dart version is unchanged.

0.1.1

• Fix README screenshots not rendering on pub.dev by using absolute GitHub raw URLs instead of relative paths.

0.1.0

Breaking changes

• Renamed the package from edge_ai to edge_gen_ai. Update your dependency and imports (e.g. package:edge_gen_ai/edge_gen_ai.dart).
• Renamed the public API from the EdgeAi prefix to EdgeGenAI:
  • EdgeAi is now EdgeGenAIPrompt.
  • EdgeAiAvailability, EdgeAiDownloadProgress, EdgeAiDownloadStatus, and EdgeAiGenerationOptions are now EdgeGenAIAvailability, EdgeGenAIDownloadProgress, EdgeGenAIDownloadStatus, and EdgeGenAIGenerationOptions.
  • The platform interface (EdgeGenAIPlatform) and method channel implementation (MethodChannelEdgeGenAI) were renamed accordingly, and checkAvailability/downloadModel now take the EdgeGenAIFeature they apply to.

New features

• EdgeGenAISummarizer, EdgeGenAIProofreader, EdgeGenAIRewriter, and EdgeGenAIImageDescriber: one-shot text summarization, proofreading, rewriting (in a chosen EdgeGenAIRewriteStyle), and image description. Backed by ML Kit GenAI's dedicated APIs on Android and task-specific Foundation Models prompts on iOS, each with its own independent availability check and model download.
• EdgeGenAIPrompt.generateContent accepts an optional single image (encoded bytes) alongside the prompt. On iOS this requires iOS 27+ (Foundation Models Attachment, Beta) and building with Xcode 27+.
• Conversation memory is now isolated per EdgeGenAIPrompt instance instead of shared per app process.

0.0.1

• Initial release: on-device availability check, model download with progress, streaming text generation with optional generation options, and optional conversation memory, backed by Apple's Foundation Models on iOS and ML Kit GenAI (Gemini Nano) on Android.