Data connections
Connect existing systems while keeping credentials and execution boundaries explicit.
On this page
Choose a connector
nondro includes connectors for PostgreSQL, SQL databases through installed SQLAlchemy drivers, Google Drive, local files, delimited tabular files, and HTTP services. The connection form shows each connector's supported configuration. Available operations depend on the connector, source permissions, and installed drivers.
A connection is a named reference to a source, separate from the datasets and business objects modeled on top of it. Choose a stable reference such as OPERATIONS_DB or TEAM_DRIVE so downstream definitions do not need to carry connection details.
Connect, test, and discover
Use Connections to add a system. The setup tests the connection before saving it. Credentials are resolved on the server; catalog specifications refer to the connection name through conn_ref. Do not include passwords, tokens, or connection strings in prompts or artifact definitions.
Explore data runs discovery where the connector supports it. Review the proposed datasets, objects, and relationships before creating drafts. Discovery can identify candidates, but your team decides which concepts belong in its ontology.
Work with files and Drive
A tabular-file connection presents a delimited file as a pageable table. Configure its delimiter and encoding explicitly. Rows are streamed, malformed rows are counted, and paging follows row order. Select a stable source file when you need reproducible results.
For Google Drive, folder sharing and credential scopes both matter. Reading existing files requires the appropriate read scope. Writing reports requires a write-capable reference and folder access. Use separate references for incoming source files and report destinations when they need different permissions.
Reach systems inside your network
An operator can deploy a connector agent near private systems. The agent makes outbound requests to the instance and resolves connection references locally. Source credentials stay on that host.
The host's allowlists control writable references, permitted materialization schemas, and database functions. SQL systems reached through this bridge can require explicitly declared datasets when remote schema discovery is unavailable. Verify a connection with the Check connection control or the operator CLI.
nondro check-source OPERATIONS_DB