rollout-49calls._expectations.md
Source: <HOME>/.codex/sessions/2026/09/24/rollout-2026-09-24T16-12-01-<UUID>.jsonl,
approximately 1.2 MB and 545 lines. Extracted on 2026-09-24.
Structural expectations
- All 545 lines parse. cli_version=0.155.0-alpha.16.3; originator=codex_vscode, anonymized in the test sample.
- Record counts: response_item/message=101, event_msg/thread_settings_applied=50, event_msg/task_started=49, turn_context=49, event_msg/user_message=49, event_msg/agent_message=49, token_usage_record=49, event_msg/token_count=49, event_msg/task_complete=49, response_item/reasoning=48, session_meta=1, response_item/compaction=1, world_state=1.
- 49 model calls, equal to the token_usage_record count; all models are codex-auto-review.
Usage expectations
Sum the per-call token_usage_record.payload.usage or token_count last_token_usage:
| Field | Sum |
|---|---|
| input_tokens | 4,184,537 |
| cached_input_tokens | 4,022,016 |
| cache_write_input_tokens | 0 |
| output_tokens | 4,051 |
| reasoning_output_tokens | 725 |
| total_tokens | 4,188,588 |
- All 98 per-call records satisfy total=input+output.
- The final token_count total_token_usage snapshot matches every summed per-call field.
- Cache-read/input ratio: 4,022,016 / 4,184,537 ≈ 96.1%, reflecting the auto-review loop.
total_token_usage is cumulative. Summing all 49 snapshots, yielding 81,558,231 input, is incorrect. Use the final value or sum per-call last_token_usage/token_usage_record.