Skip to content

UI End-to-End Visual Check

Use this checklist to quickly verify the full backend-to-UI path.

Preconditions

  • API server running on http://127.0.0.1:8000
  • UI dev server running on http://localhost:4173 (use localhost, not 127.0.0.1, for the Vite dev server)

Start both with:

python scripts/ironflow_server.py start

1) Seed test runs

From repo root:

python scripts/ui_e2e_seed.py

Optional custom seed volume:

python scripts/ui_e2e_seed.py --mapped 4 --chained 4 --complexity 10
python scripts/ui_e2e_seed.py --mapped 2 --chained 2 --failing 2 --complexity 8

Wide and long DAG stress (via benchmark API):

curl -X POST http://127.0.0.1:8000/benchmark/run \
  -H 'Content-Type: application/json' \
  -d '{"flavor":"wide","complexity":100}'

curl -X POST http://127.0.0.1:8000/benchmark/run \
  -H 'Content-Type: application/json' \
  -d '{"flavor":"long_chain","complexity":100}'

(mapped / chained flavors are accepted as aliases for wide / long_chain.)

2) Visual checks in UI

Open http://localhost:4173/runs and verify:

  • Top navigation shows Flow Runs, Flows, Deployments, and Work Pools
  • Runs table shows recent runs with state filter chips
  • State badges are visible (COMPLETED expected for benchmark runs)
  • Clicking a run opens run detail page with Cancel/Retry actions when applicable

Deployments

  • /deployments lists deployments with Quick Run
  • Deployment detail shows schedule, parameters, and recent deployment runs

Work Pools

  • /work-pools shows default-process-pool
  • Pool detail lists workers (local worker should appear as ONLINE when server worker is enabled)

Inside run detail verify tabs:

  • Task Runs: multiple task rows present; when tasks used persist_result (or returned None), pretty-printed JSON / null may appear under the row
  • Logs: log entries present
  • Events: state transitions and task events present
  • Artifacts: result artifacts present for completed task events; persisted payloads show as pretty JSON when summary includes result
  • DAG:
  • Aggregated fan-out mode renders the forecast/manifest graph (source: forecast when static compile succeeded; map() fan-out collapsed)
  • Task runs mode renders one node per execution
  • Definition line: dependencies left → right, parallel tasks top → bottom
  • Toolbar: Fit, Reset, Aggregated fan-out / Task runs toggle
  • Search finds a task by id/label/name, zooms to it, and highlights upstream/downstream path; Enter cycles multiple matches
  • Scroll zooms; drag pans the canvas (smooth GPU transform)
  • Wide runs (wide_flow): Aggregated fan-out mode collapses fan-out; Task runs lists individual mapped runs
  • Long chains (long_chain_flow): top-to-bottom dependency layout; zoom/pan to navigate
  • Failing runs: downstream nodes may show NOT_REACHABLE

See DAG and forecast for details.

3) Persistence check

Stop and restart API server, then refresh UI:

  • http://localhost:4173/runs should still show previously seeded runs

This confirms persisted data is being picked up from local storage.