Jobs¶

Predictions, designs and embeddings all create a job. This is how you track it, read its output, and clean it up. All job endpoints are keyless; you can only reach jobs you own (see Authentication).

The Job object¶

{
  "object": "job",
  "id": "6f1069cb2d665cb939f8baaa3cd261a6",
  "kind": "predict",              // "predict" | "design" | "embed"
  "status": "running",            // queued | running | succeeded | failed | canceled
  "name": "myprotein",
  "model": "boltz2",              // null for designs
  "protocol": null,               // set for designs
  "progress": 0.5,                // 0..1, or null when indeterminate
  "stage": "msa",                 // human-readable current step
  "done": 0, "total": 1,          // sub-units completed / total
  "params": {},                   // the params the job was submitted with
  "error": null,                  // set when status=failed
  "created_at": "2026-07-09T12:16:06Z",
  "started_at":  "...",
  "finished_at": null,
  "results_ready": false,
  "links": { "self": "...", "results": "...", "archive": "...", "logs": "..." }
}

Poll one job¶

curl -s https://api.japanfold.com/v1/jobs/$JOB

Add Prefer: wait=<seconds> to block until the job finishes instead of returning immediately (see Predictions):

curl -s -H 'Prefer: wait=60' https://api.japanfold.com/v1/jobs/$JOB

A simple poll loop:

until curl -s https://api.japanfold.com/v1/jobs/$JOB \
  | grep -qE '"status":"(succeeded|failed|canceled)"'; do sleep 5; done

List your jobs¶

Paginated (cursor-based). limit defaults to 20 (max 100).

curl -s "https://api.japanfold.com/v1/jobs?limit=20"

{ "object": "list", "data": [ { "object": "job", ... } ],
  "has_more": false, "next_cursor": null }

Pass next_cursor back as ?cursor=... to fetch the next page.

Cancel, delete¶

curl -s -X POST   https://api.japanfold.com/v1/jobs/$JOB/cancel   # stop a queued/running job
curl -s -X DELETE https://api.japanfold.com/v1/jobs/$JOB          # delete the job and its data

Results¶

Once results_ready (or status is succeeded), read the scores and the list of downloadable artifacts:

curl -s https://api.japanfold.com/v1/jobs/$JOB/results

Ask earlier and you get {"object":"results","job_id":"...","ready":false,"status":"running"}, so ready is safe to poll on directly.

A prediction result:

{
  "object": "results",
  "job_id": "...",
  "kind": "predict",
  "ready": true,
  "rows": [
    { "id": "target_1", "status": "ok", "n_chains": 1, "n_residues": 33,
      "samples": 1, "msa": false, "plddt": 0.6675, "ptm": 0.3226,
      "runtime_s": 28.3 }
  ],
  "artifacts": [
    { "path": "target_1.cif", "target": "target_1", "type": "structure",
      "url": "/v1/jobs/.../artifacts/target_1.cif" }
  ],
  "archive_url": "/v1/jobs/.../archive"
}

rows: one row per target, with confidence scores. Fields depend on the model and inputs: plddt/complex_plddt, ptm/iptm, and on Boltz-2 affinity runs the affinity fields. A rough read: interface iptm > 0.5, fold complex_plddt > 0.7.
A design result carries designs instead of rows, ranked for BoltzGen and unranked for RFdiffusion3. An embed result carries sequences; see Embeddings → Results.
artifacts[].url and archive_url are paths under the base URL. Prefix them with https://api.japanfold.com.

Download artifacts¶

One file:

curl -s https://api.japanfold.com/v1/jobs/$JOB/artifacts/target_1.cif -o target_1.cif

Everything as a zip:

curl -s https://api.japanfold.com/v1/jobs/$JOB/archive -o results.zip
unzip -oq results.zip -d results

Logs¶

Plain-text run log, useful while a job runs or to debug a failure:

curl -s https://api.japanfold.com/v1/jobs/$JOB/logs