Get Operator Run

get/v1/graphs/{graph_id}/audit/runs/{operation_id}

One operator run with its full paper trail: the history it was seeded with, every model call with the tool calls it made and the results it read, the answer, and the response metadata. Readable by a graph admin or by the user who made the run.

Authentication

Authenticate in any one of these ways — not all of them:

API key in the X-API-Key header.
Bearer token in the Authorization header.

Path parameters

Name	Type	Description
`graph_id`required	string	Graph identifier Constraints: matches `^(kg[a-f0-9]{16,}(?:_[a-zA-Z0-9]{1,20})?\|sec(?:_[a-zA-Z0-9]{1,20})?\|library)$`
`operation_id`required	string	The run's operation id

Example request

curl

curl -X GET "https://api.robosystems.ai/v1/graphs/{graph_id}/audit/runs/{operation_id}" \
  -H "X-API-Key: $ROBOSYSTEMS_API_KEY"

Responses

200 Successful Response

Field	Type	Description
`operation_id`required	string	The run's operation id; every mutation audit entry the run made carries it as `operation_id`
`thread_id`optional	string	The conversation the run belongs to, when the caller threaded it; null for a standalone API call
`user_id`required	string	The user the run ran as
`operator_type`required	string	The operator that ran
`mode`required	string	The execution mode the run used
`status`required	string	How the run ended One of: `completed`, `failed`, `cancelled`
`question`required	string	The message the run was given
`answer`optional	string	The operator's answer; null when the run did not reach one
`error`optional	string	Why a failed run failed
`credits_consumed`optional	number	Credits the run's model calls cost Default: `0`
`call_count`optional	integer	Model calls the run made Default: `0`
`tools_called`optional	string[]	The tools the run called, each named once
`write_count`optional	integer	Writes the run made to the graph Default: `0`
`started_at`required	string (date-time)	When the run started on the worker
`finished_at`required	string (date-time)	When the run ended
`duration_ms`required	number	How long the run took
`history`optional	object[]	The prior turns the run was given, as role/content pairs
`transcript`optional	object[]	One entry per model call, in order: the input added since the previous call (the question, then each turn's tool results), the model's output blocks (text and tool calls), stop reason, token usage and credits. The first entry also carries the system prompt's SHA-256 and size and the tool names offered; later entries repeat them only when they change
`result`optional	object	The response envelope's metadata without its result rows: tokens, credits, the Cypher run, the writes made, confidence and loop flags

Status	Meaning
400	Invalid request
401	Authentication required
403	Access denied
404	Resource not found
422	Validation Error
429	Rate limit exceeded
500	Internal server error