`recordings_get_person_detections`

Get recording person detections

Read raw person-detection evidence for one source-time window of a recording. Returns detector boxes normalized to the recording's display frame, each with its score, whether it was accepted or retained as sub-threshold, an optional track-continuity id, a tracker-ambiguity flag, and an inferred speaker attribution carrying the basis that justified it (face_anchor: a reviewed face sat inside this box at this instant; track_propagated: carried along a track from an anchor elsewhere in it; appearance_reassociation: re-established across a track boundary by relative appearance ranking; escalation_arbitrated: committed by the reviewed escalation lane; voice_turn_binding: the person layer bound this stretch of the track to the voice speaking while it held the picture). This is unreviewed machine output, not identity: a track id is continuity, and an unattributed box is honestly unassigned. Reviewed, human-scoped subject geometry stays in recordings_get_speaker_evidence.

Surface: App-only

Product release 1dce717408a55fac1047e7d779177fa3ef8ede51
Public contract commit 1ce47f7857ae3d9b6933fe83d5c33de3ae5efc64
Public contract SHA-256 241d47d25e4b4c1dae9d2263793874d06fc4ba7bc3c78d4cf2bc63fbb32ed7dc
Captured 2026-09-24T16:39:30.100Z. Review or improve this contract on GitHub.

These blocks show the complete MCP descriptors captured from serving Rails. Security schemes and resource URIs can differ by connected host.

App descriptor

{
  "name": "recordings_get_person_detections",
  "title": "Get recording person detections",
  "description": "Read raw person-detection evidence for one source-time window of a recording. Returns detector boxes normalized to the recording's display frame, each with its score, whether it was accepted or retained as sub-threshold, an optional track-continuity id, a tracker-ambiguity flag, and an inferred speaker attribution carrying the basis that justified it (face_anchor: a reviewed face sat inside this box at this instant; track_propagated: carried along a track from an anchor elsewhere in it; appearance_reassociation: re-established across a track boundary by relative appearance ranking; escalation_arbitrated: committed by the reviewed escalation lane; voice_turn_binding: the person layer bound this stretch of the track to the voice speaking while it held the picture). This is unreviewed machine output, not identity: a track id is continuity, and an unattributed box is honestly unassigned. Reviewed, human-scoped subject geometry stays in recordings_get_speaker_evidence.",
  "inputSchema": {
    "type": "object",
    "properties": {
      "recording_id": {
        "type": "string",
        "description": "Recording public ID."
      },
      "window_start_seconds": {
        "type": "number",
        "description": "Window start on the recording's source-time axis."
      },
      "window_end_seconds": {
        "type": "number",
        "description": "Window end on the recording's source-time axis. Spans are clamped; the served window is reported back."
      },
      "limit": {
        "type": "integer",
        "description": "Maximum detections to return for the window."
      }
    },
    "required": [
      "recording_id"
    ],
    "additionalProperties": false
  },
  "annotations": {
    "readOnlyHint": true,
    "destructiveHint": false,
    "idempotentHint": true,
    "openWorldHint": false
  },
  "securitySchemes": [
    {
      "type": "noauth"
    }
  ],
  "_meta": {
    "securitySchemes": [
      {
        "type": "noauth"
      }
    ],
    "ui": {
      "visibility": [
        "app"
      ]
    },
    "openai/widgetAccessible": true,
    "openai/visibility": "private"
  },
  "outputSchema": {
    "type": "object",
    "properties": {
      "recording_id": {
        "type": "string"
      },
      "person_detections": {
        "type": "object"
      }
    },
    "required": [],
    "additionalProperties": false
  }
}

Errors

[
  "recording_not_found",
  "invalid_input"
]

Examples

[]