offline_slm 0.0.1

A fast, privacy-focused Flutter package for running offline Small Language Models (SLMs) and drop-in AI chatbots locally on device.

Changelog

0.0.1

• Initial release of offline_slm.
• Core engine interfaces (SLMEngine, LlamaEngine, MockSLMEngine).
• Native C/C++ FFI bridge with CMake and Podspec configurations for Android, iOS, macOS, Windows, and Linux.
• Background isolate execution for zero-jank 60/120fps UI token streaming.
• Chat state controller (OfflineChatController) with history management, sliding context window, cancellation support, and JSON serialization.
• Chat prompt templating for ChatML (SmolLM2, Qwen 2.5), Meta Llama 3.2, Google Gemma 2, Microsoft Phi-3.5, and Mistral.
• Mobile model catalog with presets for SmolLM2 (135M, 360M, 1.7B), Qwen 2.5 (0.5B), Llama 3.2 (1B), and Phi-3.5 Mini.
• Resumable/streaming ModelDownloader with speed calculation, progress tracking, and atomic file saving.
• Complete drop-in UI kit: OfflineChatWidget, OfflineChatTheme, ChatBubble with animated typing cursor and metrics pill, ChatInputBar, ModelStatusBar, and PromptSuggestions.
• Full Flutter example app showcasing drop-in and custom configurations.