> ## Documentation Index
> Fetch the complete documentation index at: https://www.agentworldprotocol.com/llms.txt
> Use this file to discover all available pages before exploring further.

# session-telemetry.schema.json

> Canonical schema for session.telemetry notifications.

<Note>Generated from [`schemas/v0.1/session-telemetry.schema.json`](https://github.com/Hyperduality/agent-world-protocol/blob/main/schemas/v0.1/session-telemetry.schema.json) (`$id`: `https://agentworldprotocol.com/schemas/v0.1/session-telemetry.schema.json`). Normative text: [/spec/loop/time-models](/spec/loop/time-models). Every tagged example on this site is validated against this schema in CI.</Note>

World-side latency measurements aggregated over the preceding window. Each quantity is present whenever at least one sample fell in the window.

<ResponseField name="window_ms" type="integer" required>Measurement window the statistics cover.</ResponseField>
<ResponseField name="observation_latency_ns" type="latency_stats">Pipeline latency, ts\_send\_ns − ts\_mono\_ns, over every observation frame sent in the window.</ResponseField>
<ResponseField name="admission_latency_ns" type="latency_stats">Admission acknowledgement transmitted − received\_ts\_mono\_ns, over every action.submit in the window.</ResponseField>
<ResponseField name="observation_to_action_ns" type="latency_stats">received\_ts\_mono\_ns − basis\_ts\_mono\_ns, over every submission in the window that carried a basis.</ResponseField>
<ResponseField name="command_latency_ns" type="latency_stats">World receipt − ts\_mono\_ns, over every command frame received in the window.</ResponseField>
<ResponseField name="channels" type="per_channel">Pipeline latency per observation channel, for every channel that sent frames in the window.</ResponseField>