How to Preview Migrations Without a Live Database

The standard loom stitch --preview workflow needs a live database connection — it inspects information_schema to compute the diff. For brand-new projects (or quick "what would this emit?" sanity checks) that's overkill.

The fix: --no-live

loom stitch --preview --no-live
# or, since --no-live implies --preview:
loom stitch --no-live

This skips the database connection entirely. Loom treats the live schema as empty and shows the plan as if you were running against a fresh database.

When to use it

• Brand-new projects. Before you've provisioned a database, see what the first migration will look like.
• Schema review. Open a PR that adds three new Threads, and reviewers want to see the SQL that'll get applied — paste the --no-live output into the PR description.
• CI smoke checks. Run loom stitch --no-live to confirm Threads are still SQL-able after a refactor, without standing up a database in the pipeline.

Example output

loom stitch (env=dev, --no-live)

Plan (3 step(s)):
  + create_table — CREATE TABLE customers
  + create_table — CREATE TABLE contacts
  + create_table — CREATE TABLE notes

If a Thread has type-drift or nullability issues that would surface against a real database, --no-live won't catch them — by definition, there's no live schema to drift from.

When NOT to use it

• Before applying real migrations. Always loom stitch --preview (without --no-live) against the live schema before loom stitch for real. The diff against your actual database is the one that matters.
• Production schema reviews. --no-live shows what a fresh run would emit, not what your current run will emit.

Combining with --env

The --env label gets logged for audit even in --no-live mode:

loom stitch --no-live --env=preview-pr-123

Useful when you want PR reviewers to see exactly which environment context the plan is for.

See also

• Reference → CLI Command Reference — every flag for loom stitch
• Getting Started → Connecting to PlanetScale — DSN setup for real previews