Seeding data

Seed your workspaces with data from production dumps, live databases, or S3 backups — at service setup time or any time after.

Seed sources

Source	Form	Notes
PostgreSQL URL	postgresql://user:pass@host:5432/db	live pg_dump from a running server
Local file	./dump.sql, ./backup.dump	.sql via psql, other extensions via pg_restore (custom format)
S3 object	s3://bucket/path/dump.sql	credentials/region from standard AWS env vars

Seeding at service setup

devflow service add app-db --provider local --service-type postgres --from ./backup.sql
devflow service add app-db --provider local --service-type postgres \
  --from postgresql://readonly:pass@replica:5432/mydb
devflow service add app-db --provider local --service-type postgres \
  --from s3://my-bucket/backups/latest.dump

Seeding an existing workspace

devflow service seed main --from dump.sql
devflow service seed feature/auth --from postgresql://readonly:pass@replica:5432/mydb
devflow service seed main --from s3://my-bucket/backups/latest.dump
devflow service seed main -s app-db --from dump.sql      # specific service

Tip

Seed main once, then let branching do the rest — every new workspace clones main’s data via CoW (or Postgres TEMPLATE for shared engines). Re-seed individual workspaces only when they need different data.

How it works

From a PostgreSQL URL — devflow runs pg_dump in an ephemeral Docker container against the source, downloads the dump to a temp file, and restores it into the target container with pg_restore. localhost URLs are rewritten to host.docker.internal automatically so the dump container can reach databases running on your host.

From a local file — .sql is piped through psql; any other extension is treated as custom format and restored with pg_restore.

From S3 — the object is downloaded using standard AWS credentials (AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_DEFAULT_REGION/AWS_REGION), then restored like a local file.

Resetting instead of re-seeding

For “give me a clean copy of the parent again”, prefer:

devflow service reset feature/auth

It re-clones from the parent state — faster than re-restoring a dump, and exactly what agent retry loops want.