Kaax
HomeSign In
Guías

Automate a flow

The whole path: from the catalogs to a queryable result, without opening Kaax.

  1. 1

    Read the catalogs first

    The valid reportType values and statistics are edited by the Kaax super admin. Hardcoding them guarantees a break the day someone adds an index. Watch `defaults.mapResolution`: it is a number (72, 150 or 300), and sending it as text is refused for its type.

    javascript
    const types = await kaax("/api/v2/catalogs/report-types");
const vi = await kaax("/api/v2/catalogs/vegetation-index-defaults");
  2. 2

    Build the template from a preset

    Authoring `layouts` by hand means writing nested 0..1 normalised coordinates. The preset already carries the structure; you only change the name.

    javascript
    const { data: presets } = await kaax("/api/v2/templates/presets");
const preset = presets.find((p) => p.key === "counting-density");

const template = await kaax("/api/v2/templates", {
  method: "POST",
  body: { ...preset.seedPayload, name: "Densidad — Banano Álamos" },
});
  3. 3

    Resolve the lot and the configuration

    Only lots with a measured area can run: one without geometry cannot be matched to a raster and the queue skips it. The configuration's `tag` is what a queue step references, and its `modelId` is the detection model bound to that tag.

    javascript
    const { data: lots } = await kaax("/api/v2/lots?withGeometry=true");
const { data: configs } = await kaax("/api/v2/analysis-configs");

const tag = configs.counting[0]?.tag;
  4. 4

    Follow the queue and collect the result

    Kaax sends no push notifications, so status is polled. Once a step is `completed`, its `analysisId` is the bridge to the result.

    javascript
    const { data: runs } = await kaax("/api/v2/queue-runs?status=completed");

const done = runs.flatMap((r) => r.steps).filter((s) => s.analysisId);
const { data: analyses } = await kaax("/api/v2/analyses?limit=100");

for (const step of done) {
  const result = analyses.find((a) => a._id === step.analysisId);
}

Enqueuing stays manual. Creating and running a queue consumes credits and storage, so the public API only reads. It is assembled from the dashboard or by talking to the MCP.