Skip to content

loupekind reset

What loupekind reset does

Return a shoot, its exports, or a single frame to a known state when a bug or a mistake corrupts it. reset is a command group and always takes a subcommand; running loupekind reset on its own is an error, not a no-op.

Subcommand Scope
loupekind reset frame <name> one frame, back to post-ingest
loupekind reset shoot the whole shoot, back to post-ingest (or a softer "keep my picks" reset)
loupekind reset exports the rendered export/ output only

The shoot must already be registered with loupekind ingest. Every subcommand takes --shoot DIR, defaulting to the current working directory, and honours the global --output text|json flag.

What reset never touches

Reset only ever writes inside the shoot's .loupekind/ sidecar and its export/ directory. No original frame is renamed, rewritten or deleted by any reset, in any mode. Nor is .loupekind/shoot.yaml, the manifest recording which frames belong to the shoot: after any reset, the shoot is still a shoot.

Two things reset also leaves alone that are easy to assume it clears:

  • .loupekind/reviews/: AI critiques survive every reset. They are paid, non-reproducible calls to a third-party model, so they are deliberately kept out of anything a reset can destroy.
  • .loupekind/exports.yaml: the ledger of what was exported and when survives reset exports. Deleting the rendered files means "render these again", not "this delivery never happened".

One more thing no reset cleans up: the .xmp sidecars that loupekind xmp write places beside your originals. They live in your own folder, not in .loupekind/, so removing them is a manual step.

Reset asks for no confirmation

None of the three subcommands prompt, and none take a --force or --dry-run flag. loupekind reset shoot executes immediately and, in its default mode, permanently deletes the object-removal patches under .loupekind/removals/, the only copies of those generative fills. Read reset shoot before running it.

The word "Guarded" in the command's own help text refers to the non-destructive scope (originals are never touched), not to an interactive guard.

Reset is hidden from the MCP tool surface

All three subcommands are excluded from the MCP tool surface, so an AI agent driving LoupeKind over MCP cannot invoke them. That is a rail against agents, not against you: from the CLI or the studio's per-shoot Danger zone, they run immediately.

See also