Quickstart

# pip install sgraal
from sgraal import SgraalClient

client = SgraalClient(api_key="sg_demo_playground")
verdict = client.preflight(memory_state=[{"id": "mem1", "content": "user prefers email", "type": "preference"}])
print(verdict["recommended_action"])  # USE_MEMORY / WARN / ASK_USER / BLOCK

// Node 18+ — global fetch, no SDK package needed
const res = await fetch("https://api.sgraal.com/v1/preflight", {
  method: "POST",
  headers: { "Authorization": "Bearer sg_demo_playground", "Content-Type": "application/json" },
  body: JSON.stringify({ memory_state: [{ id: "mem1", content: "user prefers email", type: "preference" }] }),
});
const verdict = await res.json();
console.log(verdict.recommended_action);  // USE_MEMORY / WARN / ASK_USER / BLOCK

curl -X POST https://api.sgraal.com/v1/preflight \
  -H "Authorization: Bearer sg_demo_playground" \
  -H "Content-Type: application/json" \
  -d '{"memory_state": [{"id": "mem1", "content": "user prefers email", "type": "preference"}]}'

Hit POST /v1/preflight on api.sgraal.com — the scoring endpoint that returns the four-band decision plus full analytics. For an unmetered plain-strings preview, see POST /v1/check.

What you'll get back

{
  "decision": "USE_MEMORY",
  "confidence": 0.87,
  "request_id": "req_abc123",
  "reasoning": "Memory aligns with agent context, no provenance conflict detected.",
  "policy_version": "v12"
}