litertlm 0.0.11

Native LiteRT-LM bindings for Flutter across mobile, desktop, and web.

0.0.11

• Updated the iOS and macOS dependencies after upstream replaced the LiteRT-LM v0.14.0 binary artifacts in place.
• Aligned iOS and macOS streaming callbacks with the replacement artifacts.

0.0.10

• Added Web support for conversation token counts and benchmark information.
• Added opt-in native tool-call token streaming.
• Added native CPU thread-count configuration.
• Added maxOutputTokens to Conversation.sendMessage, Conversation.sendMessageStream, and Conversation.sendMessageWithCallback.
• Added Conversation.renderPreface for native platforms.
• Fixed FFI serialization of ConversationConfig.systemMessage.

0.0.9

• Updated native runtime dependencies to LiteRT-LM v0.14.0.
• Added platform-specific FFI compatibility for the v0.14.0 streaming callback ABI differences.

0.0.8

• Improved WASM compatibility.

0.0.7

• Added MessageCallback.onMessageDone for message boundaries.
• Updated Conversation.sendMessageStream to emit an empty message at native message-generation boundaries.
• Added SessionConfig.maxOutputTokens and SessionConfig.applyPromptTemplateInSession for supported platforms.
• Added ExperimentalFlags.filterChannelContentFromKvCache and ExperimentalFlags.visualTokenBudget.

0.0.6

• Added ExperimentalFlags with enableBenchmark, enableSpeculativeDecoding, enableConversationConstrainedDecoding.

0.0.5

• Added callback-style conversation streaming.

0.0.4

• Added maxNumImages for EngineConfig.
• Added Channel for supported platforms.
• Added Session, SessionConfig, and advanced session control.
• Added model capabilities check.
• Added benchmark support.
• Improved Backend control.

0.0.3

• Improved ConversationConfig, Tool with Swift/Java matching contract.
• Improved tool call support.
• Added document for public contract.

0.0.2

• Initial release with experimental support for all platforms.