EXAMPLE — config migration, exported mid-run (synthetic demo)

session ses_demointerrupted agent build model test-provider/test-model/default export v1.18.31

Started 2026-06-15 05:06:40 UTC · last update 2026-06-15 05:36:40 UTC

messages

tool calls

completed

failed

unfinished

repeated ops

recovered

unresolved

Findings

No repeated identical operations, recovered failures, unresolved failures, or provider errors were detected in this export.

Tool	Calls	Completed	Failed	Avg	Max
bash	1	0	0	?	?
read	1	1	0	70 ms	70 ms

Tool	Call	Status	Duration
read	#0	completed	70 ms

Tool	Call	Status	Output size
read	#0	completed	15 chars

Reported by OpenCode in the export; not an invoice.

input 100 · output 50 · reasoning 10 · cache read 0 · cache write 0 · cost 0.01

user · 2026-06-15 05:06:40 UTC · #0

Migrate the config loader to the new schema.

assistant · 2026-06-15 05:06:40 UTC · finish: tool-calls · #1

read completed call #0 · 70 ms

input

{
 "path": "config/loader.py"
}

(loader source)

bash running call #1

input

{
 "command": "pytest tests/test_config.py -q",
 "description": "run config tests"
}

This report describes what the exported data shows. It does not prove that a run was correct, complete, or caused any particular outcome.
"Identical" means canonical-JSON-equal tool input. Calls that differ only in volatile values (timestamps, random ids) are not grouped.
An identical call repeated after a success is not necessarily a bug; it may be polling, verification, or a re-read.
Missing completion evidence is reported as unknown, not as failure.
Token counts and cost come from the export as reported by OpenCode; they are not an invoice.
One export file covers one session. Subagent/child sessions appear as separate exports and need their own run.
Attached files and patches are summarized (names, sizes, paths); embedded file content is not expanded into the report.
The tool reads a local JSON file only: no network requests, and the input file is never modified.