AIChatGetRun¶

GET ai/chat/run

Returns the state of a run and, when it is finished, the answer. This is both the polling path for interfaces without streaming and the recovery path after a lost connection — the full answer is never sent through the event channel.

Access Control¶

Allowed: client, manager, admin. Only runs started by the current session.

Request¶

RESTTCPSDK

GET request

GET https://{broker_domain}/ai/chat/run?runId=run_55b…
Authorization: <JWT_TOKEN>

{
  "command": "AIChatGetRun",
  "extID": "1",
  "data": { "runId": "run_55b…" }
}

const run = await platform.AIChatGetRun({ runId });

Parameters¶

Parameter	Type	Description
`runId`	string	Run id returned by AIChatSendMessage
`accountLogin`	int	Account context. Required for a client session of an account owner

Response Data¶

{
  "runId": "run_55b…",
  "conversationId": "cnv_8f3a…",
  "state": "completed",
  "model": "anthropic/claude-sonnet-4.5",
  "result": {
    "content": "The position was closed by stop out…",
    "sources": [
      { "id": "trading.trade-close-reasons", "title": "Why a position was closed", "section": "Trading", "scope": "global", "score": 14.2 }
    ],
    "actions": [
      { "type": "navigate", "route": "/backoffice/trades", "label": "Open trades" }
    ],
    "tools": [ { "tool": "get_open_trades", "state": "ok" } ],
    "messageId": "msg_1c7…"
  },
  "usage": { "inputTokens": 1840, "outputTokens": 120, "cost": 0.0042 },
  "errorCode": null,
  "errorMessage": null,
  "startedAt": "2026-09-17 10:12:21",
  "completedAt": "2026-09-17 10:12:40"
}

`state`	Meaning
`queued`	waiting for a slot
`running`	the model is answering
`completed`	the answer is in `result`
`failed`	generation failed; see `errorCode`
`cancelled`	stopped by AIChatCancelRun
`interrupted`	the service restarted mid-generation; the question has to be asked again

sources are the knowledge documents behind the answer, actions are buttons defined by those documents (never invented by the model), tools is the trace of platform data the assistant read. What to render from each is described in Chat integration.

Errors¶

Code	Error	Description
`403`	`AI_ACCESS_DENIED`	The session may not use the assistant
`404`	`AI_RUN_NOT_FOUND`	Unknown run, or a run of another session