Skip to content
Design System
GitHub

The steps of a run with status and duration: missions, agent tool calls, rollouts, CI runs, audit trails. Steps open to show input, output or sub-steps; an optional waterfall bar.

  1. Load site maphall-b-v12, Done420 ms
  2. Plan routeplanner, Done1.3 s
  3. Drive to door 3nav, Running
  4. Take photo, Waiting
  5. Return to dock, Waiting

Trace default

Install
npx shadcn add @syncrobotic/trace

Also installs: lucide-react, radix-ui, @syncrobotic/cn, @syncrobotic/spinner

Overview

The steps of a run in order, each with a status and a duration, that open to show input, output and sub-steps.

When to use

  • What a robot did on a mission.
  • An AI agent's tool calls and reasoning.
  • Firmware rollouts, map builds and other CI-like runs.
  • Audit trails of who changed what.

Use something else

  • Events at points in time with no duration: Timeline.
  • A sequence a person goes through: Stepper.
  • Thousands of log lines: Data grid with virtual rows.

More examples

States and real uses. Each example is also a Storybook story and a visual regression test.

  1. Load site map, Done420 ms
  2. Plan route, Done1.3 s
  3. Drive to door 3, FailedPath blocked by a pallet at corridor B-4 for 30 s.34 s
  4. Take photo, Skipped
  5. Return to dock, Done52 s

Trace failed step

  1. Input
    From dock 2 to door 3, avoiding zone C
    Output
    { "waypoints": 14, "length_m": 212.4, "eta_s": 168 }

Trace expandable

    1. Gate A, Done5 m 12 s
    2. Gate B, WarningDoor left open; photo sent to security.8 m 18 s
    3. Parking lot, Done10 m 02 s
  1. Night patrol · South, Running

Trace nested

  1. search_events{ site: "hall-b", since: "22:00" }, Done180 ms
  2. get_snapshot{ robot: "7F3A", camera: "front" }, Done940 ms
  3. Three door events at gate B after 22:00 and the snapshot shows the door ajar. Likely the latch, not an intrusion.

  4. create_work_order{ door: "B", priority: "normal" }, Running

Trace agent tool calls

  1. Authenticate, Done120 ms
  2. Fetch fleet, Done680 ms
  3. Fetch maps, Done1.8 s
  4. Fetch cameras, Warning2 of 14 slow2.9 s
  5. Render, Done500 ms

Trace waterfall

Firmware 2.4.1 rollout, Warning32 m 00 s5 done1 warning
  1. Robot 7F3A2.4.0 → 2.4.1, Done5 m 22 s
  2. Robot DAA62.4.0 → 2.4.1, Done5 m 27 s
  3. Robot 9E372.4.0 → 2.4.1, Done5 m 23 s
  4. Robot 17152.4.0 → 2.4.1, Done5 m 20 s
  5. Robot B54C2.4.0 → 2.4.1, Done5 m 26 s
  6. Robot 53842.4.0 → 2.4.1, WarningUpdated, but the lidar self-test took twice as long.6 m 24 s

Trace with header

  1. Boot, Done1.8 s
  2. Self-test, Done4.2 s
  3. Localize, Done2.6 s
  4. Load mission, Done300 ms
  5. Arm, Done120 ms

Trace dense

  1. Map editedlin@site, DoneZone C marked no-go.09:41
  2. Route publishedlin@site, Done09:44
  3. Retention changedadmin, WarningCamera clips kept 7 days instead of 30.10:02

Trace audit log

  1. Waiting, Waiting
  2. Running, Running
  3. Done, Done1.2 s
  4. Warning, Warning2.4 s
  5. Failed, Failed800 ms
  6. Skipped, Skipped

Trace all statuses

Map build #312, Failed1 m 36 s3 done1 failed1 skipped
  1. Checkout, Done4.0 s
  2. Merge scans, Done38 s
  3. Build occupancy grid, Done41 s
  4. zone "C-dock" overlaps wall at (41.2, 8.0)
    1 error, 0 warnings
  5. Publish, Skipped

Trace ci

  1. Charge to 80 %, Done1 h 04 m
  2. Full map scan, Done2 h 01 m
  3. Ping, Done12 ms

Trace long durations

Usage rules

  • Done steps are a neutral check; only warnings and failures are coloured.
  • A failed step says why in its description; the raw error goes in a TraceCode inside it.
  • Open the step that needs attention by default, nothing else.
  • total plus start and duration draw the waterfall; leave them out for a plain list.

Changelog

  • 0.27.02026-11-21A span with nothing under it opens and closes in place instead of as tall as its parent span.
  • 0.23.02026-11-20Every word it says can be translated without editing the source: Trace (inherited by nested traces) and TraceHeader take labels: the six status words, the counts and how durations are written.
  • 0.17.32026-11-19Tracks and quiet shapes use the new --track token, so they read in dark mode (1.5:1 on their surface; they were near invisible).

How to upgrade