qita Guide¶
Goal¶
Use qita as your default run inspection tool: board, view, replay, and export.
UI Snapshot¶
qita board¶

qita trajectory view¶

0) First create at least one run¶
Run any example with trace enabled (default):
This usually creates run artifacts under ./runs/<run_id>/ with:
manifest.jsonevents.jsonlsteps.jsonl
1) Start board¶
Open URL printed by CLI (default: http://127.0.0.1:8765/).
Board provides:
- run list and search/filter
- run-level metrics and status
- quick buttons:
view,replay,export raw,export html
2) View one run (readable cards)¶
From board click view, or open directly:
In view page:
Trajtab for step-by-step cardsManifesttab for run metadata- timeline and event sections
- font scaling and fold/unfold controls
3) Replay one run in browser¶
CLI mode:
This opens focused replay mode at:
Use replay when you need to inspect temporal order of events and failures.
4) Export run artifacts¶
Export raw JSON bundle¶
From board/view click export raw, or open:
Export standalone HTML¶
From board/view click export html, or CLI:
Use this to share one run snapshot without requiring a live server.
5) Recommended workflow for debugging¶
- run agent once
- open board and locate failed run by
stop_reason - open
viewand find first abnormal phase/event - use
replayto confirm event ordering - export HTML for issue discussion or PR evidence
6) Common problems¶
- board shows no runs
- ensure
--logdirpoints to the directory containing run subfolders -
ensure each run folder has
manifest.json -
replay reports run not found
-
pass full path to
--run, e.g.runs/<run_id> -
malformed event line
- parser tolerates invalid lines, but fix upstream event writer if this repeats
If qita command is not found¶
If you installed QitOS without editable install, or your shell can't find console scripts, use: