87 lines (45 loc) · 5.54 KB

Changelog

0.9.1 (2026-05-19)

Dependencies & Chores

surface dependency & chore commits in release notes (d650799)

0.9.0 (2026-05-19)

Features

add cumulative token totals to engine throughput cards (dd80c8e)
add Time Per Output Token (TPOT) to engine latency card (20bc677)
bracket chart units in titles and clean up hover tooltips (d92a975)
cache resolved model info and support per-endpoint engine API keys (18fa2f8)
hide tooltip header on prefill and decode throughput charts too (952d5ee)
move tok/s into engine throughput titles and fix hover header (bc2292f)
plot prefix cache hit rate alongside KV cache over time (4a105ca)
select physical/Wi-Fi network interface instead of loopback (d7cdfbf)
surface cumulative prefix cache queries on engine cache card (0860324)

0.8.0 (2026-05-01)

Features

slo goodput customization (#23) (1576e43)

0.7.0 (2026-04-28)

Features

improvind dashboard and fixing data visibility (#20) (c7851bd)

0.6.0 (2026-04-28)

Features

model detail tags (#18) (c0633bc)

0.5.0 (2026-04-25)

Features

vllm: expand vLLM observability with latency percentiles, SLO goodput, and dashboard polish (082bd17)

Bug Fixes

compute vLLM prefix cache hit rate from counters (#14) (1797e8a)

0.4.0 (2026-04-23)

Features

multi-engine dashboard with auto-rotation and live vLLM metrics (#12) (53b2af8)

0.3.0 (2026-04-22)

Features

make dashboard hardware- and host-agnostic (3b77d5a)
make dashboard hardware- and host-agnostic (9f0e3a6)
surface engine deployment mode (Docker vs Direct) in tabs (6e9de44)

0.2.0 (2026-04-20)

Features

package and distribute via cargo install + systemd service (b2a87d8)

Bug Fixes

dev: prevent local tilde expansion of SPARK_DIR (e1e8350)
install: refuse sudo invocation; let binary self-escalate (942c796)
install: refuse sudo invocation; let binary self-escalate (e778623)
test: align MemoryCard test selectors with current StackedBar markup (e499da9)