Forklight
INC-2107open

Error spike after the 14:05 deploy

Signals

per-minute · 46 min observed
primary signal

Web error rate

18.3%baseline 0.4%
0102014:05 deploy v8.3.1nowFORECASTSLO 1%13:4715:05 horizon
p95 latency1,799ms
Cache hit ratio40.2%
Requests64/min
Primary DB CPU51.8%

Observed record

deploys and alerts on the incident clock
13:30 → 14:32 · incident opened 14:05
13:30
13:45
14:00
14:15
14:30
api@2.41.0
web@8.3.1
now
14:05deployIntroduce an edge response cache (cache:v2 keys).
14:07pageweb_error_rate above 5% for 2m (threshold 1%).
14:12warnweb_p95_ms above 1500ms for 5m.

Counterfactual timelines

nothing forked yet — the agent branches here, or fork by hand
Observed timelinemain

What production is actually doing. Forks branch from here; only the human switch merges one back.

open · 14:32

start here

Hand this incident to your agent

The page tools in this tab let an agent read the signals, fork counterfactual timelines, rehearse each mitigation against the forecast, and stage one proposal — this canvas updates live as it works. Applying to production is not one of its tools.

“Errors spiked after the 14:05 deploy. Compare rolling back against bypassing the new cache, show me the evidence, and stage the lowest-risk mitigation. Do not apply anything.”

  1. signals.queryRead the evidence
  2. scenario.forkFork a timeline
  3. scenario.simulateRehearse a fix
  4. mitigation.stageStage a proposal
  5. mitigation.applyYou only

Staged for review

human-only

No proposal yet. Once the agent has compared its forks it stages one mitigation here, with its rationale and evidence.

Reviewing it — and shipping it — stays with you.

Activity ledger

agent you

Quiet so far. Every dispatch — an agent tool call or your own click — lands here as it happens, through the same typed contract.