Skip to main content
Seeds a running clone into a known starting state. Fixtures apply deterministic SQL directly; --ai generates realistic data and creates it through the clone’s REST API. For the conceptual difference, see Seeding.

​
Arguments

​
Options

You must pass at least one. With neither, the command prints Nothing to seed. Pass --fixture <name> and/or --ai. and exits 1.

​
Examples

Deterministic fixture:
AI-generated data:
Both at once (fixture backbone + AI on top):
Warnings (prefixed !) are best-effort — rejected records don’t fail the seed. Up to five are shown.

​
AI seeding details

  • Model: claude-haiku-4-5.
  • Auth: reads ANTHROPIC_API_KEY from your environment.
  • Default brief: “a small, fast-moving software startup.”
  • Default volume medium: 6 users, 4 channels, 20 messages. See the volume table.
  • Data is created via real signup → workspace → channels → messages, so auth and referential integrity hold.

​
Seed on spin

To seed at create time instead, pass --seed to spin:

​
Reset replays it

The last seed is recorded on the clone and re-applied automatically by reset.

​
Errors you might see