Skip to content

rollout-single-call._expectations.md

Source: <HOME>/.codex/sessions/2026/09/24/rollout-<ts>-<UUID>.jsonl, 183 KB, 17 lines, 2026-09-24T08:10Z. Extracted on 2026-09-24; the active session could append later.

Structural expectations

  • All 17 lines parse; parseErrors=0.
  • Counts: session_meta=1, turn_context=1, world_state=1, response_item/message=4, response_item/reasoning=1, response_item/compaction=1, token_usage_record=1; event_msg contains two thread_settings_applied and one each of task_started, user_message, agent_message, token_count and task_complete.
  • session_meta.cli_version=“0.155.0-alpha.16.3”. Original originator=codex_vscode, anonymized to anon-N in the sample.
  • turn_context.model=codex-auto-review: an auto-review sub-agent session, without a user-message-associated interactive model.
  • Top-level timestamp uses ISO 8601 UTC with milliseconds, such as 2026-09-24T08:10:24.567Z. Some payload fields, including started_at/completed_at/create_time, use epoch seconds.

Usage expectations for allowlisted native fields

Exactly one model call appears in three usage records, with identical values:

Record Path input_tokens cached_input_tokens cache_write_input_tokens output_tokens reasoning_output_tokens total_tokens
token_usage_record payload.usage 25558 4864 0 131 65 25689
event_msg/token_count payload.info.last_token_usage 25558 4864 0 131 65 25689
event_msg/token_count payload.info.total_token_usage 25558 4864 0 131 65 25689

Every relationship holds in this sample:

  • total_tokens=input_tokens+output_tokens: 25689=25558+131.
  • cached_input_tokens⊆input_tokens: 4864≤25558.
  • reasoning_output_tokens⊆output_tokens: 65≤131.
  • last==total in this one-call session; the cumulative snapshot equals the individual call.

There are 16 anonymized IDs, anon-1…anon-16.