pg-flux

Reference

Dump / verify / pull

Read-only commands for adoption and out-of-band change detection.

These three commands handle the bidirectional flow: live DB → source files.

#pg-flux dump

Shell 
pg-flux dump [--output ./schema] [--layout per-kind|flat] [--force]
Flag Default Description
--output <dir> ./schema Destination
--layout <kind> per-kind per-kind (one file per object kind) or flat (single schema.sql)
--force off Overwrite a non-empty target directory

The dump is round-trip clean: running pg-flux migrate generate immediately after a dump produces zero pending statements. That property is enforced by an integration test gate.

#pg-flux verify

Shell 
pg-flux verify [--strict]

Read-only asymmetric diff: lists every catalog object present in live but not declared in source. --strict exits 1 when any are found — pair with CI.

#Example

Shell 
$ pg-flux verify --strict
verify: 2 undeclared live object(s):

  Tables (1):
    - public.hotfix_overrides

  Indexes (1):
    - public.idx_users_legacy

Run `pg-flux pull` to capture these into schema/_pulled/<ts>.sql for review.

Exit codes: 0 clean, 4 undeclared objects found.

#pg-flux pull

Shell 
pg-flux pull [--dry-run=true] [--output ./schema/_pulled]
Flag Default Description
--dry-run true Print the would-be SQL; don't touch disk
--output <dir> ./schema/_pulled Quarantine directory

Renders every live-only object via the same emitters dump uses, into a single timestamped quarantine file.

#Example

Shell 
$ pg-flux pull --dry-run=false
Wrote 1 object(s) to ./schema/_pulled/20260520_103245_pulled.sql