Data connections

Connect object storage, PostgreSQL, Supabase, or upload local files.

A connection stores non-secret location and provider settings. Its credential reference points to a server-managed identity or secret. A dataset selects an object, prefix, or table within that connection and registers it as a logical Space.

Upload

Use uploads for CSV, JSON, JSONL, and NDJSON. Relifold stores the file in workspace-owned physical storage and registers the chosen logical Space. Read the upload guide.

Object storage

• Amazon S3: object or prefix, optional region, JSON/JSONL/CSV.
• Google Cloud Storage: object or prefix using workload identity.
• Azure Blob Storage: container object or prefix using managed identity.

A prefix may contain several supported objects. Browsing a connection shows direct prefixes and objects; selecting a prefix tells the connector to read its compatible objects.

PostgreSQL and Supabase

Select a schema and table or a bounded query. Store the database credential in the configured secret service; do not paste a password into a pipeline or dataset declaration. Read the PostgreSQL and Supabase guide.

Row limits

Rows to read is optional. Leave it empty to read the full selected dataset. Use a limit for planning previews, demos, and bounded research experiments—not for a sparse sample of a prediction target that requires complete daily coverage.

Connection checklist

1. Create the connection and credential reference.
2. Test it.
3. Browse to the exact object, prefix, or table.
4. Preview its schema and rows.
5. Register it under a meaningful Space name.