Database connections

Database connections

A Connection is a read-only DB credential Pollenix uses to answer questions in the Workbench. Admins create connections; per-user / per-role / per-workspace grants decide who can query through them and which tables they’re allowed to see.

Open Connections in the navbar.

Supported drivers

Driver	Notes
Postgres	Standard.
MySQL / MariaDB	mysql+aiomysql.
SQL Server	mssql+aioodbc. ODBC driver must be installed in the cluster.
Snowflake	OAuth or key-pair.
BigQuery	Service account JSON.
Redshift	Postgres wire-compatible.
SQLite	Mostly for local dev.

Add a connection

1. Click New connection.
2. Pick the driver kind.
3. Fill in host, port, database name, schema (optional).
4. Pick a secret backend — Pollenix’s local vault, or your AWS Secrets Manager / Azure Key Vault — then paste the read-only DB credentials.
5. Click Test connection — runs SELECT 1 plus a small introspection query.
6. Save.

Connection editor — driver, host, secret backend.

:::caution Use a read-only role Pollenix’s SQL writer is constrained to SELECT and the executor refuses anything else, but defense-in-depth: give the connection a read-only role at the DB. :::

Schema introspection

After saving, Pollenix runs a one-shot schema catalog crawl: it walks the connection’s tables, columns, types, and primary/foreign keys. The catalog drives autocomplete, the table picker, and the NL→SQL writer’s understanding of what’s queryable.

You can trigger a refresh from the connection row’s menu (Re-introspect) whenever you add or remove tables in the source DB.

Schema catalog — every table, its columns, when last seen.

What’s next

A connection alone doesn’t let anyone query — you also need grants to bind it to users, roles, or workspaces.