Issues

• Ship an Apple privacy manifest with the XCFramework

  leehack/llamadart-native#97 · Filter by author leehack opened on Oct 2, 2026

• Free remaining llama.cpp objects at process exit before ggml-metal teardown (Flutter quit, hot restart)

  leehack/llamadart-native#96 · Filter by author leehack opened on Oct 1, 2026
  1

• Windows: ship the Visual C++ 2015-2022 runtime app-locally

  leehack/llamadart-native#95 · Filter by author leehack opened on Oct 1, 2026

• Qwen3-TTS code predictor ignores the request seed and carries its RNG across syntheses

  bugpriority:P2
  leehack/llamadart-native#90 · Filter by author leehack opened on Sep 23, 2026

• Export cancel-byte callbacks so llamadart can stop an in-flight llama_decode and clip encode

  enhancementpriority:P3
  leehack/llamadart-native#88 · Filter by author leehack opened on Sep 23, 2026
  1

• Investigate Galaxy S24 Vulkan crash during first GGUF generation

  blocked-upstreambugpriority:P1
  leehack/llamadart-native#79 · Filter by author leehack opened on Sep 17, 2026
  9

• Qualify optimized Windows ARM64 Vulkan release builds on hardware

  enhancementpriority:P2
  leehack/llamadart-native#65 · Filter by author leehack opened on Aug 24, 2026

• Qualify and publish verified Windows CUDA release sidecars

  enhancementpriority:P2
  leehack/llamadart-native#64 · Filter by author leehack opened on Aug 24, 2026

• Consolidate mtp_*/ngram_* speculative APIs onto the generic speculative_* family (ABI-breaking)

  enhancementpriority:P3
  leehack/llamadart-native#51 · Filter by author leehack opened on Aug 21, 2026
  1