Add data and connections

Upload or connect

Direct upload sends a file into a Catalog collection you select first. A source connection lists objects or rows from S3, GCS, Azure Blob, PostgreSQL, or MySQL. Both still need a stored workspace and document:upload. Non-retaining workspaces do not take connections.

Three states

Uploaded means the document row exists and processing was queued. Extracted means a run produced Smart Bites. Searchable means those bites are indexed with a compatible embedding profile. Retry a failed confirmation with the same document id. Do not create a duplicate to recover a queue error.

Workers

Stored ingest needs object storage and a worker queue. S3_NOT_CONFIGURED and DOCUMENT_PROCESSING_QUEUE_UNAVAILABLE are operator escalations, not user mistakes to hide.