LLM OBSERVABILITY / v0.9

Every call, every span, every dollar.

Trace every LLM call. Eval on every commit. Reproduce any prompt, span by span. Spot the regression before the customer tweets.

REQ · LAST 1H
1.24 M
+4.2%
COST · TODAY
$284.18
−12%
p99 · PROD
2.14 s
stable
ERRORS · 15m
7
+3
LIVE — support-bot · prod
streaming · 12 sources
TRACES · LAST 15 MIN5 shown
  • SUCCESSsupport.reply.v7 · v71.82 s1.2kt$0.00190
  • ERRORsupport.reply.v7 · v74.23 s1.4kt$0.00240
  • SUCCESSsupport.reply.v7 · v72.11 s1.3kt$0.00180
  • SUCCESSsupport.reply.v7 · v72.03 s1.1kt$0.00030
  • ERRORsupport.reply.v7 · v73.46 s1.3kt$0.00030
p99 LATENCY · 15m2.14s
15m10m5mnow
EVAL · SUPPORT-QA
92.4%+0.8
pass rate · since PR #482
MODEL MIX
  • claude-opus-4-762%
  • claude-sonnet-4-624%
  • gpt-4.1-mini11%
  • other3%
COST · BY ENV
  • prod$248.50
  • staging$24.18
  • dev$11.50
budget · 28% of $1,000 used
Live numbers for demo purposes. Wrap your SDK; ship in 10 minutes.READ THE SDK DOCS →