Leaving Nelcota (no lock-in)

Export your schema, policies, identities and files to another backend. You need owner access, PostgreSQL tools and a verified destination. Database metadata is SQL; file bytes and deployment secrets must be exported separately. Completion time depends on dataset size and destination compatibility.

For the Docker commands below, enter projects/<name>/ inside your host folder so Compose selects that project. Systemd hosts use native PostgreSQL tools. Keep dumps, user/password-hash exports, file mappings and configuration private.

1. The whole database (schema, data, policies, functions, users)#

docker compose exec -T postgres pg_dump -U postgres -Fc postgres > nelcota.dump
# restore into any Postgres 17+:
pg_restore --no-owner -d "postgres://user@new-host/db" nelcota.dump

The roles anon, authenticated, service_role, authenticator and nelcota_auth are global and not part of pg_dump. Create them on the target first (the file migrations/V1__roles_e_auth.sql does that and is idempotent) or use pg_restore --no-acl if you will not use them.

2. Just the users (with their hashes)#

docker compose exec -T postgres psql -U postgres -c "\copy (
  SELECT id, email, encrypted_password, email_confirmed_at,
         raw_user_meta_data, created_at, last_sign_in_at
  FROM auth.users) TO STDOUT WITH CSV HEADER" > users.csv

encrypted_password uses the standard PHC argon2id format ($argon2id$v=19$m=19456,t=2,p=1$...). Any argon2 library verifies these hashes without conversion: users do not need to change their passwords. Details in The auth schema.

3. The JWTs#

The format is in JWT and roles. To switch auth providers without rewriting the policies, the new provider only has to issue tokens with role and sub. The auth.uid()/auth.jwt() functions stay the same.

4. The files#

storage.objects is part of the dump above. The bytes are plain files in projects/<name>/storage/ (disk) or objects in your own bucket (S3), stored as <bucket>/<version>. The name of each one:

docker compose exec -T postgres psql -U postgres -c "\copy (
  SELECT bucket_id || '/' || version AS key, bucket_id || '/' || name AS name
  FROM storage.objects) TO STDOUT WITH CSV HEADER" > files.csv

5. What stays behind#

Only the nelcota binary: the automatic REST API and the panel. Without Nelcota, Postgres keeps serving exactly the same data with the same policies, and you can point PostgREST, Hasura or your own backend at it.