Resumable Runs¶
Runs checkpoint after every iteration. If something goes wrong — network error, rate limit, or you just hit Ctrl+C — pick up right where you left off. No wasted API calls.
Run Lifecycle¶
stateDiagram-v2
[*] --> running: generate
running --> paused: interrupted
running --> completed: target reached / max iterations
running --> failed: unrecoverable error
paused --> running: resume
failed --> running: resume
completed --> [*]
| Status | Meaning |
|---|---|
running |
Loop is actively iterating |
paused |
Interrupted (Ctrl+C or signal) |
completed |
Coverage target reached or max iterations used |
failed |
Unrecoverable error during an iteration |
Info
Only paused and failed runs can be resumed. Completed runs are final.
What Gets Saved¶
Full run state is written to disk after each iteration at ~/.daystrom/runs/{runId}.json:
| Field | What it is |
|---|---|
runId |
Unique identifier for this run |
status |
Current lifecycle status |
userInput |
Your original configuration (topic, intent, provider, etc.) |
iterations |
All completed iteration results with metrics |
bestIteration |
Index of the iteration with highest coverage |
currentTopic |
The latest topic definition |
createdAt / updatedAt |
Timestamps |
Tip
Because state is saved after every iteration, you never lose more than the current in-progress iteration on failure.
Resuming¶
The resumed run continues with:
- The same topic name (locked after iteration 1)
- The latest topic definition from the last completed iteration
- Full iteration history for LLM context
- Inherited settings (including
accumulateTests) from the originalUserInput
Add More Iterations¶
# Resume with up to 10 more iterations from current position
daystrom resume abc123 --max-iterations 10
Viewing Results¶
List All Runs¶
Summary table with run ID, status, topic name, iterations completed, best coverage, and timestamps.
View a Specific Run¶
# Best iteration (highest coverage)
daystrom report <runId>
# A specific iteration
daystrom report <runId> --iteration 3
Shows topic definition, test results, metrics, and analysis for the selected iteration.