How do I submit jobs from an autonomous agent?

Request sandbox API access, then POST job JSON to the jobs endpoint using your sandbox API key. Include client_job_id to support idempotent retries and a callback_url for webhooks if you want push updates.

What is callback_url and how is it used?

Provide a callback_url to receive job status webhooks. SolarCompute will POST status updates (started, completed, failed) to the URL. Validate signatures if you require authenticity.

How should I poll job status if I can't use webhooks?

If polling, use exponential backoff (2s, 5s, 15s, 60s) and cap at 60s. Limit polling concurrency to avoid rate limiting and design idempotent job submissions.

Autonomous Agent Integration

This page provides concise guidance for integrating autonomous agents, CI pipelines, or bots that submit jobs to SolarCompute's developer API. It is intended to reduce integration friction and surface required details when requesting sandbox access.

Recommended workflow

Request sandbox API access (see Developer Quickstart).
Use the sandbox API key to submit test jobs from your agent or pipeline.
Validate outputs and scale concurrency in the sandbox before requesting production quotas.

Sample job JSON (agent submission)

Agents should include a short id, human-readable name, and expected resource sizes. Replace placeholders with sandbox values.

Copy sample JSON Download sample JSON Copied

{
  "client_job_id": "agent-12345-2026-10-20",
  "workload": "inference",
  "instance": "gpu-24gb-v1",
  "concurrency": 4,
  "command": "python serve_model.py --port 8080",
  "container": {
    "image": "registry.example.com/my-model:latest",
    "env": {"MODEL_PATH":"/data/model"}
  },
  "storage": {"type":"s3","path":"s3://my-bucket/input"},
  "callback_url": "https://my-agent.example.com/webhook/job-status",
  "expected_duration_seconds": 3600
}

Submitting a job cURL example

Use your sandbox API key in the Authorization header. This example posts the sample JSON to the jobs endpoint; replace the URL and fields with your sandbox values.

curl -X POST "https://api.solarcompute.net.au/v1/jobs" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "client_job_id": "agent-12345-2026-10-20",
    "workload": "inference",
    "instance": "gpu-24gb-v1",
    "concurrency": 4,
    "command": "python serve_model.py --port 8080",
    "container": {"image": "registry.example.com/my-model:latest"},
    "storage": {"type":"s3","path":"s3://my-bucket/input"},
    "callback_url": "https://my-agent.example.com/webhook/job-status",
    "expected_duration_seconds": 3600
  }'

Webhook callback format

If you provide a callback_url, SolarCompute will POST job status updates. Validate signatures if you require authenticity.

{
  "job_id": "sc-abc123",
  "client_job_id": "agent-12345-2026-10-20",
  "status": "completed",
  "exit_code": 0,
  "started_at": "2026-10-20T10:12:00Z",
  "finished_at": "2026-10-20T11:15:00Z",
  "outputs": [{"type":"s3","path":"s3://my-bucket/output/result.tar.gz"}]
}

Polling and retry guidance

Prefer webhooks for near-real-time updates. If polling, use exponential backoff: start at 2s, then 5s, 15s, 60s, and cap at 60s.
Limit polling concurrency per API key during peak times to avoid rate limiting.
Design idempotent submissions using client_job_id to safely retry on transient failures.

Security & headers

Use Authorization: Bearer <API_KEY> in the Authorization header. Protect keys in your agent's secret store and rotate keys regularly.

Agent best practices

Include a short human-readable description in job metadata to help support triage.
Batch small tasks to reduce scheduling overhead; use concurrency fields for parallelism.
Validate storage paths and region compatibility before submitting large datasets.

When requesting sandbox access, include your agent's intended concurrency, expected job sizes, and callback URL in the request. Contact sales and developer support: hello@solarcompute.net.au.