Issues

• Seed-OSS parse takes a leading thought as reasoning when the prompt never opened one

  bugdartpriority:P3
  leehack/llamadart#1561 · Filter by author leehack opened on Oct 28, 2026
  1

• Hook tests poison the checkout's native bundle cache; next dart test crashes

  bugdartpriority:P2
  leehack/llamadart#1558 · Filter by author leehack opened on Oct 28, 2026
  1

• Streaming: closed-thought hold follow-ups from the #1555 audit

  bugdartpriority:P3
  leehack/llamadart#1557 · Filter by author leehack opened on Oct 28, 2026
  1

• Projector refusal: refuse a LiteRT-LM custom backend before IO and tighten wording (#1552 audit)

  enhancementpriority:P3
  leehack/llamadart#1556 · Filter by author leehack opened on Oct 28, 2026
  1

• gpuDevices: pin placement storage on GPU-less CI and harden direct modelLoad splits (#1547 audit)

  enhancementpriority:P3
  leehack/llamadart#1554 · Filter by author leehack opened on Oct 28, 2026
  1

• Streaming: Command R7B, Cohere2 MoE and Seed-OSS stream a thought closed for thinking off as content

  bugdartpriority:P3
  leehack/llamadart#1553 · Filter by author leehack opened on Oct 28, 2026
  1

• Scoring pack: cite the right LiteRT blocker and pin remaining gates (#1543 audit)

  enhancementpriority:P3
  leehack/llamadart#1548 · Filter by author leehack opened on Oct 28, 2026
  1

• Reference-image editing follow-ups from #1515 audit

  enhancementmodel-assetspriority:P3
  leehack/llamadart#1536 · Filter by author leehack opened on Oct 28, 2026
  1

• Reasoning parity follow-ups from #1524/#1525 audit: mutants X2/X7, Ling <think> in args, partial-tag drop

  dartenhancementpriority:P3
  leehack/llamadart#1531 · Filter by author leehack opened on Oct 28, 2026
  6

• gpuDevices: validate multi-GPU layer and row split on real 2-GPU hardware

  diagnosticsenhancementpriority:P3
  leehack/llamadart#1528 · Filter by author leehack opened on Oct 28, 2026
  1

• Cache and hook follow-ups from #1510/#1514 audits

  bugfillerpriority:P3
  leehack/llamadart#1522 · Filter by author leehack opened on Oct 28, 2026
  1

• Pin the default wiring of the llama.cpp library probes (follow-up to #1511)

  diagnosticsenhancementfillerpriority:P3
  leehack/llamadart#1519 · Filter by author leehack opened on Oct 28, 2026
  1

• Model store: chat app integration with app-local fallback

  enhancementmodel-assetspriority:P3
  leehack/llamadart#1500 · Filter by author leehack opened on Oct 28, 2026
  1

• Model store: background download executors for Android and iOS

  enhancementmodel-assetspriority:P3
  leehack/llamadart#1499 · Filter by author leehack opened on Oct 28, 2026
  1

• WebGPU: apply lazy tool-call grammars when the bridge reports lazyGrammar

  enhancementpriority:P3structured-output
  leehack/llamadart#1480 · Filter by author leehack opened on Oct 27, 2026
  1

• Greedy decoding output differs from llama-server on GLM-OCR for the same prompt

  bugdiagnosticspriority:P3
  leehack/llamadart#1478 · Filter by author leehack opened on Oct 27, 2026
  1

• LiteRT-LM auto CPU fallback follow-ups (#1456 audit)

  enhancementpriority:P3
  leehack/llamadart#1458 · Filter by author leehack opened on Oct 27, 2026

• llama.cpp: worker killed during model/context creation leaks the native object

  bugpriority:P2
  leehack/llamadart#1403 · Filter by author leehack opened on Oct 24, 2026
  1

• LiteRT-LM embeddings fail on Linux arm64 and crash on Windows x64 (all v0.18 pins)

  buglitert-lm-paritypriority:P2
  leehack/llamadart#1367 · Filter by author leehack opened on Oct 24, 2026
  4

• Investigate slow Android Vulkan embeddings on Pixel 8a: EG2 Q8_0 and FP16/BF16 operator controls

  diagnosticsenhancementpriority:P2
  leehack/llamadart#1304 · Filter by author leehack opened on Oct 21, 2026
  1

• Propose an opt-in CUDA artifact for the Stable Diffusion runtime

  blocked-runtimeenhancementpriority:P3
  leehack/llamadart#1235 · Filter by author leehack opened on Oct 20, 2026

• Expose opt-in diffusion-step caching and conditioning-cache controls

  enhancementpriority:P3
  leehack/llamadart#1233 · Filter by author leehack opened on Oct 20, 2026
  1

• Expose additional diffusion samplers, schedulers and model-specific guidance

  enhancementfillerpriority:P3
  leehack/llamadart#1232 · Filter by author leehack opened on Oct 20, 2026

• Expose bounded asynchronous image previews during diffusion sampling

  blocked-runtimeenhancementpriority:P3
  leehack/llamadart#1230 · Filter by author leehack opened on Oct 20, 2026

• Add guarded ADetailer detection and cropped-inpainting refinement

  blocked-runtimeenhancementpriority:P3
  leehack/llamadart#1229 · Filter by author leehack opened on Oct 20, 2026