flutter_gpt_engine 0.0.9
flutter_gpt_engine: ^0.0.9 copied to clipboard
Headless Flutter package for running GGUF LLMs locally on Android with streaming, conversation history, web search, device context, thinking support, benchmarking, and CPU/GPU auto-tuning.
0.0.9 #
- Added opt-in Fast, Balanced, and Quality mobile performance presets.
- Added token-aware history budgeting and safe output-token sizing.
- Added warm-up plus median multi-run benchmarking for more stable tuning.
- Added persistent auto-tune profiles restored for the same GGUF model.
- Added CPU-vs-GPU verification diagnostics without changing the active profile.
- Preserved the existing default generation configuration and public APIs.
0.0.8 #
- More optimized.
0.0.7 #
- Reduced token-stream parsing allocations with an incremental reasoning parser.
- Coalesced ChangeNotifier generation updates for smoother Flutter UIs.
- Reused HTTP connections and parallelized web source fetching.
- Added bounded conversation-history context.
- Added local TTFT / token-throughput benchmarking.
- Added opt-in CPU/GPU auto-tuning for the current device and GGUF model.
0.0.8 #
- More optimized.
0.0.7 #
- More optimized.
0.0.6 #
- device state analysis integrated.
0.0.5 #
- Web search added.
0.0.4 #
- Web search added.
0.0.3 #
- More user friendly.
0.0.2 #
- More user friendly.
0.0.1 #
- Initial package.