valite
Docs navigation

Docs/API Reference/Workflows API

Workflows API

Read everything the pipeline knows about a workflow: runs, callsites, readiness, and inline artifacts.

List workflows

GET/valite/api/workflows

All workflows observed in the call log, with per-workflow call counts, readiness, pipeline stage progress, and a count of untagged calls.

response
{
  "workflows": [
    {
      "workflow": "support-triage",
      "calls": 4210,
      "readiness": {
        "workflow": "support-triage",
        "complete_runs": 12,
        "open_runs": 1,
        "required_runs": 5,
        "ready": true,
        "untagged_calls": 0
      },
      "last_seen": "2026-07-08T17:58:02Z",
      "stages": ["profile", "triage", "bar", "replay", "plan"]
    }
  ],
  "untagged_calls": 37
}

stages lists which pipeline artifacts exist, in order — a progress bar you can render directly.

Workflow detail

GET/valite/api/workflows/{workflow}

The full picture for one workflow: runs (with token counts, errors, completeness, and whether the run was routed), callsites, readiness, and artifacts.

response (abridged)
{
  "workflow": "support-triage",
  "runs": [
    {
      "run_id": "8c14…", "run_source": "traceparent",
      "n_calls": 6, "input_tokens": 9120, "output_tokens": 1480,
      "errors": 0, "complete": true, "routed": false,
      "first_seen": "…", "last_seen": "…"
    }
  ],
  "callsites": [
    {
      "callsite_id": "cs_9f2ab1",
      "calls": 1890, "models": ["gpt-4o"],
      "endpoint": "/v1/chat/completions",
      "fingerprint": { … },
      "examples": [ … ],     // masked: volatile values redacted
      "mean_latency_ms": 812
    }
  ],
  "readiness": { … },
  "artifacts": { … }
}

Callsite examples are masked — volatile values are redacted so the dashboard can show request shapes without exposing payload content.

Artifacts

The artifacts object inlines each pipeline output that exists: profile, triage, bar, calibration, replay, validation as parsed JSON, plus routing_yaml (raw), routing (parsed), and report_md (the rendered scorecard). This is the same data the dashboard renders — anything you see there, you can fetch here.