Job results - IMP Docs

Job results require an Enterprise Edition license (queues entitlement). job_result records are written when a BullMQ job completes. They capture the job’s final state, result payload, and any error information for audit and debugging purposes. See Evals overview and Routines for product context.

type job_result {
  job_id: String
  state: String
  error: JSON
  label: String
  tries: Float
  result: JSON
  metadata: JSON
  item: String
  context: String
  type: String
  trigger: String
  trigger_metadata: JSON
  session: String
  workflow: String
  last_processed_at: Date
  embeddings_updated_at: Date
  createdAt: Date
  updatedAt: Date
  id: ID!
}

Field notes

Field	Notes
`job_id`	BullMQ job ID; correlates with the `jobs` query on a specific queue.
`state`	Run state string. Non-terminal (live) states: `waiting`, `active`, `delayed`, `paused`, `waiting_approval` (persisted — run paused on a tool-approval request; resumes after the approval turn). Terminal states: `completed`, `failed`, `filtered` (an email-trigger guard rejected the message), `cancelled` (admin-cancelled). The `stuck` state is defined but not currently written.
`error`	JSON error payload if the job failed; `null` on success.
`result`	JSON result payload returned by the job worker.
`metadata`	Supplementary key-value data recorded alongside the result (e.g. token counts, agent name).
`item`	ID string of the primary entity this job operated on (e.g. an eval_run ID).
`context`	Context slug or ID when the job was scoped to a specific knowledge context.
`type`	Job type discriminator (e.g. `"eval_run"`, `"workflow_execution"`).
`trigger`	How the run was initiated: `"manual"` (UI or direct API call), `"api"` (external API trigger), `"schedule"` (cron routine), or `"email"` (inbound email trigger). `null` on pre-migration rows.
`trigger_metadata`	JSON envelope with trigger-specific detail. For email triggers this contains the inbound message envelope (e.g. `message_id`, `from`, `subject`); for other trigger types it may be `null`.
`session`	ID of the run-backed chat session created for this job, if any. Cross-links the `agent_sessions` table.
`workflow`	ID of the owning routine (workflow). Replaces label-substring filtering for run-to-routine association. `null` on pre-migration rows.
`tries`	Number of execution attempts before the job reached its final state.
`last_processed_at`	Timestamp of the last knowledge-embedding pass over this record.
`embeddings_updated_at`	Timestamp when vector embeddings were last regenerated.

Example query

query {
  job_resultsPagination(
    limit: 20
    filters: [{ type: { eq: "eval_run" }, state: { eq: "completed" } }]
    sort: { field: "createdAt", direction: DESC }
  ) {
    items {
      id
      job_id
      type
      state
      result
      item
      createdAt
    }
    pageInfo {
      currentPage
      hasNextPage
    }
  }
}