Draft text

Settings module

POST/aurora/draftaurora.draft

Drafts an SOP, advance packet, email, day sheet or change order description in the org's verbiage library. Facts the context lacks are marked as gaps, never invented. The draft is returned, not saved or sent. Requires `Idempotency-Key`; a replay returns the original response. A rule that refuses the request returns 422 with `refusal` and `reason`.

Requires the capability home.dashboard.read.
OAuth scopes: home.
Send an Idempotency-Key header. A replay with the same key returns the original response.
Error statuses: 400, 401, 409, 422, 429. Errors are RFC 9457 problem details.

Request body: AuroraDraftRequest

aurora.draft: AuroraDraftRequest
Field	Type	Required	Description
kind	"sop" \| "advance_packet" \| "email" \| "day_sheet" \| "change_order_description"	Yes	What to draft.
instructions	string	Yes	What the draft should say.Constraints: length 1 to 4000
context	array of object	No	Records to draft from, read with the caller's permissions.
verbiage	array of object	No	The org's verbiage library entries to apply.

{
  "kind": "sop",
  "instructions": "Write the load-in SOP for the main stage.",
  "context": [
    {
      "collection": "venues",
      "id": "0192f1e2-7c3a-7b4d-9e8f-1a2b3c4d5e6f"
    }
  ],
  "verbiage": [
    {
      "term": "loading dock",
      "preferred": "Dock"
    }
  ]
}

Response 200: AuroraResult

aurora.draft: AuroraResult
Field	Type	Required	Description
outcome	"answer" \| "no_answer" \| "failed"	Yes	answer, no_answer (the data does not support an answer; `missing` says why) or failed (the model call failed and was logged).
ai_generated	true	Yes	Always true: the content is AI-generated and is labeled so (EU AI Act Article 50).
request_id	string (uuid) \| null	Yes	Identifier of the logged AI request; null when no model was called.
model	string \| null	Yes	Model that answered.Constraints: length 1 to 100
missing	string \| null	Yes	What the data lacks, for no_answer.Constraints: length 1 to 500
failure_code	string \| null	Yes	Why the request failed.Constraints: length 1 to 64
answer	object	No	The feature's answer for outcome answer, shaped per operation; null otherwise.
suggestions	array of AuroraProposal	Yes	Changes stored for confirmation.

{
  "outcome": "answer",
  "ai_generated": true,
  "request_id": "0192f1e2-7c3a-7b4d-9e8f-1a2b3c4d5e6f",
  "model": "claude-sonnet-5-5",
  "missing": "No budget lines exist yet.",
  "failure_code": "model_error",
  "answer": {
    "text": "One permit is blocked: Fire Marshal Permit.",
    "citations": []
  },
  "suggestions": [
    {
      "suggestion_id": "0192f1e2-7c3a-7b4d-9e8f-1a2b3c4d5e6f",
      "operation_id": "savedViews.create",
      "summary": "Save the view Blocked Miami Permits."
    }
  ]
}

Code samples

curl -X POST "$XOS_BASE_URL/aurora/draft" \
  -H "Authorization: Bearer $XOS_API_KEY" \
  -H "Idempotency-Key: $(uuidgen)" \
  -H "Content-Type: application/json" \
  -d '{"kind":"sop","instructions":"Write the load-in SOP for the main stage.","context":[{"collection":"venues","id":"0192f1e2-7c3a-7b4d-9e8f-1a2b3c4d5e6f"}],"verbiage":[{"term":"loading dock","preferred":"Dock"}]}'

const { data } = await xos.aurora.draft({
    "kind": "sop",
    "instructions": "Write the load-in SOP for the main stage.",
    "context": [
      {
        "collection": "venues",
        "id": "0192f1e2-7c3a-7b4d-9e8f-1a2b3c4d5e6f"
      }
    ],
    "verbiage": [
      {
        "term": "loading dock",
        "preferred": "Dock"
      }
    ]
  });

response = xos.aurora.draft({"kind": "sop", "instructions": "Write the load-in SOP for the main stage.", "context": [{"collection": "venues", "id": "0192f1e2-7c3a-7b4d-9e8f-1a2b3c4d5e6f"}], "verbiage": [{"term": "loading dock", "preferred": "Dock"}]})

xos aurora draft --data '{"kind":"sop","instructions":"Write the load-in SOP for the main stage.","context":[{"collection":"venues","id":"0192f1e2-7c3a-7b4d-9e8f-1a2b3c4d5e6f"}],"verbiage":[{"term":"loading dock","preferred":"Dock"}]}'