Introduction

The Pathwize API lets you run labeling and evaluation tasks without touching the dashboard after setup. You create a task once (in the dashboard), then your pipeline pushes new items whenever they appear and pulls results as experts finish them. Everything is JSON over HTTPS.

Base URL

https://www.gopathwize.com/api/v1

All endpoints live under this base. Requests and responses are UTF-8 JSON, except results in format=jsonl which stream one JSON object per line. Timestamps are ISO 8601 in UTC.

Quickstart

1. 1
   Create a task in the dashboard under AI tasks. Choose the type, the expert pools and define your label classes. Copy the task ID from the task page (also shown under the API section there).
2. 2
   Create an API key under Developers. It is shown once. Store it server-side, never in a browser or mobile app.
3. 3
   Push items. Send up to 1,000 items per request with a URL we can fetch, or upload a file directly.
4. 4
   Pull results. Poll the results endpoint with since or page through with starting_after. Each result carries your item ID and your meta untouched.

curl https://www.gopathwize.com/api/v1/tasks/TASK_ID/items \
  -H "Authorization: Bearer pw_live_..." \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: 2026-09-25-batch-07" \
  -d '{
    "items": [
      {
        "id": "werk-nord-2026-09-25-000871",
        "url": "https://storage.example.com/loads/2026-09-25/000871.jpg",
        "mime": "image/jpeg",
        "checksum": "sha256:9f3a...",
        "width": 4032,
        "height": 3024,
        "captured_at": "2026-09-25T07:12:04+02:00",
        "group": "werk-nord",
        "meta": { "site": "Werk Nord", "scale_kg": 12480 }
      }
    ]
  }'

curl "https://www.gopathwize.com/api/v1/tasks/TASK_ID/results?since=2026-09-25T00:00:00Z&format=jsonl" \
  -H "Authorization: Bearer pw_live_..."

How a task runs

After you create a task, Pathwize invites experts from the pools you picked, qualifies them on your test set and assigns the passing ones. Items you push enter their queue in the order of priority and creation time. Every item gets one expert; the share you configured as quality overlap (10% by default) gets a second one, and those results carry the agreement so you can watch quality without paying for double labeling on everything. A task keeps running as long as you push items, so daily batches go into the same task.