v0.8.15

What's Changed

• Bump llama.cpp to 19e92c33e (v0.8.15) + fix grammar double-accept crash by @nyo16 in #45
• Optimize Server batch loop, streaming NIF, and embed_batch by @nyo16 in #46

Full Changelog: v0.8.14...v0.8.15

v0.8.14

What's Changed

• Bump llama.cpp to 0d18aaa9d, release v0.8.14 by @nyo16 in #44

Full Changelog: v0.8.13...v0.8.14

v0.8.13

What's Changed

• Bump llama.cpp to b22ff4b7b, release v0.8.13 by @nyo16 in #43

Full Changelog: v0.8.12...v0.8.13

v0.8.12

What's Changed

• Bump llama.cpp to 52fb93a2b (30 commits) by @nyo16 in #42

Full Changelog: v0.8.11...v0.8.12

v0.8.11

What's Changed

• Bump llama.cpp to b28a2f372, release v0.8.11 by @nyo16 in #41

Full Changelog: v0.8.10...v0.8.11

v0.8.10

What's Changed

• Fix Server telemetry: centralize TTFT, add :start/:exception, stop_reason by @nyo16 in #40

Full Changelog: v0.8.9...v0.8.10

v0.8.9

What's Changed

• Add code review findings + experiment scripts (2026-05-16) by @nyo16 in #39

Full Changelog: v0.8.8...v0.8.9

v0.8.8

What's Changed

• Fix Server prefix-cache M-RoPE crash on hybrid GDN models by @nyo16 in #38

Full Changelog: v0.8.7...v0.8.8

v0.8.7

What's Changed

• Make checksum push resilient to concurrent master updates by @nyo16 in #36
• Add Multi-Token Prediction (MTP) speculative decoding by @nyo16 in #37

Full Changelog: v0.8.6...v0.8.7

v0.8.6

Full Changelog: v0.8.5...v0.8.6