Solution Explorer Commands
Experiment orchestration and retention controls for automated variant search.
labgate explore create
Create a new experiment from an existing repository.
labgate explore create --name <name> --repo <path> --eval <command> [options]| Option | Description |
|---|---|
--base-ref <ref> | Base git ref (default HEAD). |
--timeout <sec> | Evaluation timeout in seconds (default 120). |
--epsilon <value> | Exploration epsilon in [0..1] (default 0.15). |
--top-n <count> | Top-N pool size for epsilon exploration (default 5). |
--max-runs <count> | Optional run budget. |
--agent-mode <mode> | stub or claude_headless (default stub). |
--stub-patch <path> | Patch file for stub mode. |
--claude-resume-session <id> | Resume Claude session id in claude_headless mode. |
--claude-timeout <sec> | Claude headless timeout (60..14400). |
--skip-baseline | Skip baseline evaluation check after create. |
--keep-worktrees | Retention: keep worktrees after runs. |
--artifacts-policy | Retention: artifacts policy. |
--keep-last | Retention: keep last N runs. |
--keep-failed-last | Retention: keep last N failed runs. |
--keep-best | Retention: keep the best run. |
--prune-after-days | Retention: prune artifacts after N days. |
--max-delete-runs | Retention: max runs to delete per GC. |
Supports baseline eval, epsilon/top-N strategy, agent mode, and retention policy flags.
labgate explore list
List experiments with pagination.
labgate explore list [--limit <count>] [--offset <count>]| Option | Description |
|---|---|
--limit <count> | Max rows to print (default 50). |
--offset <count> | Pagination offset (default 0). |
labgate explore status
Show experiment state, best run, and recent runs.
labgate explore status <experimentId> [--limit <count>]| Option | Description |
|---|---|
--limit <count> | Recent runs shown (default 10). |
labgate explore pause
Pause an experiment.
labgate explore pause <experimentId>labgate explore resume
Resume a paused experiment.
labgate explore resume <experimentId>labgate explore tick
Run one manual/autopilot tick for an experiment.
labgate explore tick --experiment <id>| Option | Description |
|---|---|
--experiment <id> | Required experiment id. |
labgate explore tree
Render experiment run tree JSON.
labgate explore tree --experiment <id> [--mode best_path|full]| Option | Description |
|---|---|
--mode <mode> | Tree mode: best_path or full. |
labgate explore leaderboard
Show top-scoring runs.
labgate explore leaderboard --experiment <id> [-k|--top <count>]| Option | Description |
|---|---|
-k, --top <count> | Top-K rows (default 10). |
labgate explore gc
Plan or apply retention-based pruning (dry-run by default).
labgate explore gc --experiment <id> [--yes] [retention overrides]| Option | Description |
|---|---|
--yes | Apply deletion actions (without this, dry-run). |
--prune-worktrees / --no-prune-worktrees | Enable or disable worktree pruning. |
--prune-artifacts / --no-prune-artifacts | Enable or disable artifact pruning. |
--artifacts <mode> | Override artifacts mode: all|minimal|none. |
--keep-last | Override keep_last_n. |
--keep-failed-last | Override keep_failed_last_n. |
--keep-worktrees | Override keep_worktrees. |
--keep-best | Override keep_best. |
--after-days | Override prune_artifacts_after_days. |
--max-delete-runs | Override max_delete_runs. |
labgate explore retention show
Print effective retention policy for an experiment.
labgate explore retention show --experiment <id>labgate explore retention set
Patch retention policy fields for an experiment.
labgate explore retention set --experiment <id> [retention flags]| Option | Description |
|---|---|
--keep-worktrees / --no-keep-worktrees | Set keep_worktrees. |
--artifacts <mode> | Set artifacts mode (all|minimal|none). |
--keep-last <count> | Set keep_last_n. |
--keep-failed-last <count> | Set keep_failed_last_n. |
--keep-best / --no-keep-best | Set keep_best. |
--after-days <days> | Set prune_artifacts_after_days. |
--max-delete-runs <count> | Set max_delete_runs. |
labgate explore compare
Compare one run against best/parent/another run.
labgate explore compare --experiment <id> --run <runId> [--to best|parent|<runId>] [--diff]| Option | Description |
|---|---|
--to <target> | best, parent, or explicit run id. |
--diff | Include full patch artifact in output. |
--stat | Include diff stats (enabled by default). |
labgate explore overview
Print aggregate experiment overview JSON.
labgate explore overview --experiment <id>labgate explore run
Print run-level details and artifact paths.
labgate explore run --id <runId>labgate explore mcp
Run the Solution Explorer MCP server over stdio.
labgate explore mcp [--db <path>]| Option | Description |
|---|---|
--db <path> | Override explorer DB path. |
Last updated on