v0.9.0-dev.8

v0.9.0-dev.8 — MTP speculative decoding, IQ4_NL KV cache

v0.9.0-dev.7

0.9.0-dev.7: b9360, embeddings, speculative decoding, broad API coverage

v0.9.0-dev.6

params: expose missing llama.cpp options across sampler/context/model

- SamplerParams: Mirostat (v1/v2), grammar (incl. lazy patterns), DRY,
  XTC, dynamic temperature, adaptive-P, top-n-sigma, infill, logit
  bias, shared min_keep. SamplerFactory.build now accepts model: for
  vocab-dependent stages.
- ContextParams: RoPE scaling/freq, full YaRN knobs, pooling and
  attention type, defrag threshold, no_perf, op_offload, swa_full,
  kv_unified.
- ModelParams: split mode, main GPU, tensor split, device list, GGUF
  kv overrides (int/float/bool/str), use_direct_io, use_extra_bufts,
  no_host, no_alloc.
- README: link to aichat sample app.
- Bump to 0.9.0-dev.6 + CHANGELOG.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

v0.9.0-dev.5

0.9.0-dev.5: probe fix follow-up to dev.4 (analyzer-clean)

v0.9.0-dev.2

0.9.0-dev.2: M10 gaps closed (KV quant, log capture, primary-accel, A…

v0.9.0-dev.1

v0.9.0-dev.1: M8.5 Hexagon AAR + CI permission fix

Adds:
- Android Hexagon NPU + OpenCL AAR (build via Snapdragon Docker
  toolchain). Six HTP DSP variants (v68 through v81) covering
  Snapdragon 865 -> 8 Elite. ~3.7 MB stripped.

Fixes:
- Workflow GITHUB_TOKEN didn't have contents:write so the v0.9.0-dev.0
  build failed at the release-asset upload step. Added explicit
  permissions block.

NOT yet validated on a Snapdragon device — Hexagon AAR ships
experimental until tested end-to-end on hardware.

v0.2.0

llama.cpp b8595b16e69e3029e06be3b8f6635f9812b2bc3f

v0.1.0

llama.cpp @ 98bab638

vision support