API & Integrations

Wire scoring into your agent orchestrator.

A simple JSON API so an IDE plugin, agent runtime, or marketplace can ask: "Given this Spark, what should I run?"

Method	Endpoint	Description
GET	/v1/devices	List every device in the registry with current scores.
GET	/v1/devices/{id}/capabilities	Detailed capability fingerprint for a single device.
POST	/v1/recommend/models	Given a device fingerprint, return the recommended model tier.
POST	/v1/recommend/agents	Suggest an agent stack for a workload + memory budget.
GET	/v1/clusters/topologies	Validated multi-node Spark cluster blueprints.

Example: recommend a model

POST /v1/recommend/models
{
  "device_id": "dell-precision-s16",
  "workload": "multi-agent",
  "concurrency": 3
}

→ {
  "primary":   "llama31-70b",
  "secondary": "qwen3-coder-32b",
  "router":    "qwen3-7b",
  "expected_tokens_per_second": 22
}

Example: device capabilities

GET /v1/devices/dell-precision-s16/capabilities

→ {
  "unified_memory_gb": 128,
  "memory_bandwidth_gbps": 546,
  "scores": { "scs": 94, "ars": 88, "tls": 92 },
  "max_model_params_b": 235,
  "recommended_stack": "openshell-nemo"
}

API in private beta. Request a key at api@rtxsparks.com.