oe resume

Re-run an experience with cached results from a prior run.

Synopsis

oe resume [options] <run-id>

Description

oe resume replays an experience using the event log from a previous run as a cache. Nodes whose outputs were recorded in the original run are restored from the log rather than re-executed. Only nodes that did not complete in the prior run (e.g. because it failed partway through) are dispatched again.

The original args are recovered automatically from the first event (run.started) in the prior run's .jsonl log, so you do not need to pass --args again.

oe resume writes a new run-id and a new event log — the original log is never modified. The new run-id is returned in the resume complete log line.

API key required for agent/skill nodes

oe resume currently hard-wires the Anthropic client (AnthropicLLMClient) for agent and skill nodes. Set ANTHROPIC_API_KEY before resuming an experience that contains those node kinds.

Arguments

Argument	Required	Description
<run-id>	✓	The run-id of the prior run to resume from. Must match a file at .openexpertise/runs/<run-id>.jsonl.

Options

Flag	Description	Default
--experience <path>	Path to the experience directory (or experience.yaml).	.
--log-format <fmt>	Log format: json | pretty	pretty
--log-level <level>	Log verbosity: info | debug | warn | error	info
-h, --help	Display help and exit	—

Exit codes

Code	Meaning
0	Resumed run completed with status: success
1	experience.yaml not found, prior run log not found, or run completed with status: failure

Examples

Resume the most-recently-failed run (copy the run-id from the prior oe run output):

oe resume 4f3a1b2c-0d9e-4f7a-8b6c-1e2d3f4a5b6c

{"level":"info","type":"run.started","runId":"9a8b7c6d-…","msg":"run.started"}
{"level":"info","type":"node.cached","nodeId":"fetch-data","msg":"node.cached"}
{"level":"info","type":"node.completed","nodeId":"analyse","msg":"node.completed"}
{"level":"info","runId":"9a8b7c6d-…","status":"success","originalRunId":"4f3a1b2c-…","msg":"resume complete"}

Resume a run in a specific experience directory:

oe resume 4f3a1b2c-0d9e-4f7a-8b6c-1e2d3f4a5b6c --experience examples/review-branch

Get JSON output:

oe --log-format json resume 4f3a1b2c-0d9e-4f7a-8b6c-1e2d3f4a5b6c | jq '.msg'

Notes / gotchas

• LLM provider is fixed to Anthropic in the current implementation of oe resume. Unlike oe run, there is no --llm flag; the AnthropicLLMClient is always used for any agent or skill nodes that fire.
• Args are recovered from the log, not accepted on the command line. If the original run log's first line does not contain an args field, an empty object is used.
• A new run-id is always issued. Inspect the resumed run with oe inspect <new-run-id>, not the original run-id.
• The cache key used to match nodes from the prior run is based on node id + input hash — the same mechanism described in Cache key + memoization.
• oe resume does not accept --tui, --evolve, or --concurrency. Use oe run if you need those features and are starting fresh.

See also

• oe run — start a new run from scratch
• oe inspect — examine the event log for any run-id
• Resume + cache
• Cache key + memoization