genkit_flutter_gemma 0.3.1

Genkit Dart plugin for flutter_gemma - local on-device AI inference via Google Gemma models.

0.3.1

• Bump flutter_gemma dependency to ^0.15.1 (LiteRT-LM 0.11.0, MTP speculative decoding for Gemma 4, multi-image input, Android GPU fix, desktop storage path fix)
• Add enableSpeculativeDecoding config option for Gemma 4 E2B/E4B MTP toggle (null = model default, true/false = force on/off)

0.3.0

• Bump flutter_gemma dependency to ^0.14.2 (dart:ffi rewrite on desktop, ~5× faster cold start; fixes macOS flutter test install_name_tool failure)
• Add maxFunctionBufferLength config option for large tool-call argument payloads
• Update example/macos/Podfile post_install for flutter_gemma 0.14.2 framework bundling

0.2.2

• Bump flutter_gemma dependency to ^0.13.2
• Update README: document Gemma 4 and Phi-4 support, clarify thinking mode availability

0.2.1

• Bump flutter_gemma dependency to ^0.13.1 (LiteRT-LM 0.10.0, Gemma 4 thinking mode fix)

0.2.0

• Breaking: Upgrade flutter_gemma dependency to ^0.13.0
• Breaking: System messages are now passed natively via createChat(systemInstruction:) instead of being prepended to the first user message
• Add systemInstruction config option for explicit system-level instructions
• Support ModelFileType.litertlm for LiteRT-LM models (Gemma 4)
• Advertise systemRole: true in Genkit model metadata
• Throw on system-only requests (at least one user/model message required)
• Throw on system messages with non-text content parts

0.1.1

• Bump flutter_gemma dependency to ^0.12.8
• Add toolChoice config option ('auto', 'required', 'none') passed to model chat session
• Support ParallelFunctionCallResponse — multiple tool calls in a single model response
• Add latencyMs to ModelResponse for generation profiling
• Fix FakeEmbeddingModel compatibility with flutter_gemma 0.12.8 taskType parameter

0.1.0

• Initial release
• Genkit model provider wrapping flutter_gemma
• Text generation (blocking and streaming)
• Embeddings via FlutterGemmaEmbedder
• Multimodal input (images, audio)
• Function calling / tool use
• Thinking mode (DeepSeek-style reasoning)
• Configurable via @Schema()-annotated options
• Example app with Chat, Embeddings, Tools, Settings tabs