A data source is an external database table or file that a concept can read from via a bind annotation. The Data Sources API lets you register connections, browse uploaded files, infer Vadalog schemas from tables, and preview raw rows before building logic. All paths are relative to the base URL and require authentication. All responses use the standard envelope unless noted otherwise. Paginated endpoints follow pagination.

Connections

Connect a data source

Register an external database or file-based source. On success the platform stores the connection and returns the full updated source list.
database object fields:

Example

On a failed connection connectionStatus is false and errorMessage contains the driver error.

List data sources

Return all registered data sources for the authenticated user. No parameters required.
data is an array of source objects (id, name, databaseType, host, schema, tables, and row counts if computed).

List sheets

Return the sheet names available in a spreadsheet or multi-tab source (e.g. an Excel file or Google Sheet connection).

Browse a connection

Walk a warehouse connection’s hierarchy without connecting anything yet: catalogs, then the schemas in one, then the tables in one. This is how a connect form fills its dropdowns, and how you discover the table names to pass to /data/connect.
All three take the same shape: a database connection descriptor in the body, narrowed by the level you are asking about. Each returns its own key — catalogs, schemas, or tables.
Not every connector supports discovery. One that does not returns 400 with a message saying so — distinct from a 500, which means the connection itself failed. A connector with no catalog concept (most single-database engines) answers /list-schemas and /list-tables but not /list-catalogs.

Demo sources

Return the pre-built demo data sources shown during onboarding. Each preset includes its full database connection payload so the frontend can pre-populate the connect form.
No parameters. Requires authentication.

Refresh sources

Re-probe every stored connection group and reconcile the source list. Database groups are re-discovered by (type, host, port, database, schema); file-based sources are removed if the file is no longer accessible.

Infer schema

Inspect a connection and generate Vadalog bind annotations (and optionally a concept stub) for every table in the descriptor.

Preview data

Return a paginated sample of raw rows from a bound data source, with optional search and column filters.

Joins

All-pairs join

Discover join relationships across a set of data sources. Given a list of source connection descriptors, the platform analyzes every pair to find candidate join keys, and optionally evaluates the resulting joins.
At least two sources are needed to produce any pairs — a single descriptor in database_payloads yields no joins.

Files

All file operations are scoped to the server-side disk/ storage directory. Paths are relative to that root; path traversal (..) is rejected.

Upload file

Upload a file to disk/. The request must be sent as multipart/form-data. The returned filePath (e.g. disk/reports.csv) can be used directly as the host in a subsequent /data/connect call.
Allowed extensions: .csv, .xlsx, .xls, .json, .parquet, .pdf, .txt, .docx, .yaml, .ttl, .rdf, .owl, .cobol, and others configured by the server. Maximum file size: 10 GB (server-configurable).

List files

List the contents of a directory within disk/. Directories are returned before files; hidden entries (names starting with .) are excluded.

Download file

Download a file from disk/. Returns the raw file bytes, not the standard envelope. The Content-Type is inferred from the file extension.

Make directory

Create a new subdirectory inside disk/. The directory must not already exist; its parent must exist.

Delete files

Delete one or more files or directories within disk/. Non-empty directories require recursive: true.

Move or rename a file

Move or rename a file or directory within disk/. The destination must not already exist and its parent directory must exist.

Download a folder as a zip

Bundle one or more folders and files into a single .zip. Repeat the path parameter to select several. Each selected folder is nested under its own name inside the archive; a selected file sits at the archive root.
The response is the raw zip, not the envelope. An empty selection returns 400. When a single folder is selected the download is named after it; otherwise it is prometheux-download.zip.
This covers your own files. Folders that someone else shared with you are not included — download those individually with /data/files/download.

Upload a zipped folder

Upload a .zip and extract it into disk/, recreating the folder structure inside.
Send as multipart/form-data:
The whole archive is validated before anything is written — file extensions, entry count, path traversal, and your storage quota against the uncompressed size. A rejected archive therefore leaves no partial upload behind, so a 400 here means nothing changed on disk.

File sharing

Share an individual file or folder with another user, independently of ontology sharing. The recipient is granted access on the underlying storage and receives an abfss:// location they connect as a data source of their own. These endpoints apply to cloud (ADLS) storage.

Share a file or folder

data carries the abfss:// location to hand to the recipient, along with the share_id you need to revoke it later.
File sharing requires a subscription that includes it. Without one the endpoint returns a feature-access error rather than creating the share.

List and revoke shares

GET /data/files/shares lists the shares you created — active ones only, unless you pass include_revoked=true. POST /data/files/share/revoke takes a share_id in the body and removes the recipient’s access.

Other operations