Skip to content

Web console

The console is a React SPA served by the engine at /console. Sign-in is the server’s own username/password flow (POST /token); everything it shows comes from the documented HTTP API. All surfaces live under a workspace path (/w/<workspace>/…) — tenancy is in the URL.

SurfaceWhat it shows
Workflowsinstalled templates you can run now; each opens as tabs — Recipe · Graph · Configuration · Run · Runs. The Run tab validates inputs and runs a live preflight as you type
Playgrounda governed sandbox for trying workflows against an isolated environment: upload a sample dataset, profile it against the workflow’s expected data roles, run what-if on gate thresholds against real metric history, and file a threshold change as a governed suggestion
Runsrun history with status buckets; the awaiting filter is the gate queue — approve/reject inline. Each run links to its Pilot report (trace, cost, provenance, governance receipt, scorecard) and a “why” diagnosis view
Pipelinesmulti-step pipelines, their gates, and paused approvals
Insightsworst-first triage across workflows: failure clusters, cost vs baseline, health rates, per-workflow drill-down
Libraryinstall / uninstall / enable distribution packs from the registry
Environmentsthe wiring view over profiles: capability→operator bindings, backends, policy and knowledge attach-lists
GatewayLLM spend and routing: daily cap, calls, providers, and the model-roles table (fast/strong/router → provider + model)
Auditthe hash-chained audit log — filter by category, actor, run, time; CSV export

Results (what each run produced — reports, artifacts, registered models) · Jobs (remote GPU/sandbox jobs with approve/reject/cancel) · Suggestions (the governed improvement queue) · Knowledge (RAG corpora per environment) · Policies (author action-gating rules as YAML) · Connectors (provider catalog; connections are health-probed before saving) · Triggers (event triggers + cron schedules, each schedule can pin a config preset) · API (SDK/API usage reference) · Settings (workspaces, environments, users).

  • Test-before-save connections — a connection’s credentials are verified with the provider’s authenticated health probe before the connection is created; secret values are stored by reference and never displayed.
  • Governance made visible — governance config knobs render locked with an env-owned badge; preset saves that touch them are rejected; every gate decision, approval, and policy hit lands in the Audit view.
  • Live runs — run detail streams events over SSE, resumable after a disconnect.