sgl-project/SpecForge — GitHub trend

◂ BACK TO LEADERBOARD

FINDING #7395 · UNIT ID 998697045

sgl-project/SpecForge

Train speculative decoding models effortlessly and port them smoothly to SGLang serving.

[ PYTHON ][ ORG ][ GITHUB ↗ ]

SURPRISE SCORE

0.00

Score Breakdown

SURPRISE0.000265

ENGAGEMENT0.57

FRESHNESS1.51

SCORE = SURPRISE × ENGAGEMENT^0.7 × FRESHNESS × VISIBILITY × CONFIDENCE

SURPRISE = WINDOW STARS / DAYS / (AUDIENCE + FLOOR)

8% OF STARS IN ARCHIVE

Growth Telemetry

VELOCITY /D

3.00

ACCEL

-0.05

RETENTION

26.8%

PEAK 2026-08-19 · FORK-RETENTION 71.4% · 90 STARS / WINDOW

Author Audience

AUDIENCE

11,274

FOLLOWERS

1,120

OWNER ★

45,170

Engagement Signals

FORKS

333

ISSUE AUTH

PR AUTH

UNIQUE STARGAZERS 90 / 90 (DIVERSITY 1.00)

Why This Is A Finding

sgl-project/SpecForge собрал 90 звёзд за окно, тогда как у автора всего 1,120 подписчиков — эффективная аудитория ≈ 11,274. Это даёт surprise-индекс 0.000265 (звёзды относительно охвата автора, а не в абсолюте). Удержание форков 71.4% и 0 внешних контрибьюторов отделяют реальный инструмент от разовой вспышки. Акселерация отрицательная — внимание остывает после пика.

Related Findings

RANKS ABOVE 23% OF 9588 FINDINGS

#04guillaumemeyer/watermarks-remover[ PYTHON ]0.45 #11awesome-dsh-plugin/awesome-dsh-plugin[ PYTHON ]0.38 #16debpalash/VoiceStudio[ PYTHON ]0.34 #24wang2122/sprix-sage-router[ PYTHON ]0.28 #27AdamPlatin123/dsh-plugin-radar[ PYTHON ]0.26

METRICS IN CONTEXT

MEDIAN ACROSS ALL 9588 FINDINGS · Δ vs MEDIAN · PERCENTILE = SHARE RANKED BELOW

METRICVALUEMEDIANΔ MEDPERCENTILE

SCORE0.000.00-0.00ABOVE 23%

VELOCITY3.003.77-0.77ABOVE 40%

RETENTION26.8%25.0%+1.8 PPABOVE 53%

FORKS333114+219ABOVE 76%

SURPRISE0.000.01-0.01ABOVE 8%