REST API reference.

The Gensparkly API is a JSON-over-HTTPS interface for provisioning GPU instances, managing training clusters, and serving inference. All requests are versioned under /v1 and accept & return application/json.

• REST semantics, predictable resource URLs
• HTTPS only, TLS 1.3, mTLS optional on Enterprise
• Cursor pagination on list endpoints
• Idempotency-Key header on all mutating requests
• Server-Sent Events for long-running operations

All requests require a bearer token issued from the console. Tokens are scoped to a project and can be restricted by IP and permission set.

curl https://api.gensparkly.com/v1/instances \
  -H "Authorization: Bearer $GENSPARKLY_API_KEY" \
  -H "Content-Type: application/json"

Fill parameters below and copy the equivalent request in cURL, Python, or Node. Nothing is sent — the console builds the request for you to run locally.

curl -X POST "https://api.gensparkly.com/v1/instances" \
  -H "Authorization: Bearer $GENSPARKLY_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: req_01HRXK7Q8Z3F4M5W6N" \
  -d '{
  "gpu": "h100",
  "count": 8,
  "region": "us-east-1",
  "image": "cuda:12.4-torch:2.4"
}'

Provision single-node GPU instances on-demand.

curl -X POST https://api.gensparkly.com/v1/instances \
  -H "Authorization: Bearer $GENSPARKLY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "gpu": "h100",
    "count": 8,
    "region": "us-east-1",
    "image": "cuda:12.4-torch:2.4",
    "ssh_keys": ["key_01H..."]
  }'

Multi-node training clusters with NVLink and InfiniBand fabric.

S3-compatible object storage plus parallel filesystems for training data.

Deploy and invoke managed inference endpoints.

curl -X POST https://api.gensparkly.com/v1/inference/endpoints/ep_llama3/invoke \
  -H "Authorization: Bearer $GENSPARKLY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Explain NVLink in one sentence.", "max_tokens": 128}'

Read-only endpoints for cost, usage, and invoices.

Errors follow a consistent shape. HTTP status conveys the class.

{
  "error": {
    "type": "capacity_conflict",
    "message": "No H100 capacity in us-east-1 for the requested window.",
    "request_id": "req_01HRXK7Q8Z3F4M5W6N",
    "retry_after": 30
  }
}

• python · pip install gensparkly
• typescript · npm i @gensparkly/sdk
• go · go get github.com/gensparkly/gensparkly-go
• terraform · provider "gensparkly" { ... }