db_query

Parameterized SQL query against a configured DSN.

Source	internal/agents/workflow/nodes/db_query.go
When to use	Reading from an external user database — Postgres, SQLite.
Supported drivers	postgres://, sqlite:, file:

Schema

Field	Type	Required	Notes
database	string	✅	Env key whose value is the DSN. Stored in the workflow's env: block.
query	textarea (template)	✅	SQL with $1 / $2 / … placeholders (Postgres style — sqlite driver normalises).
params	list (templated)		Positional params for the placeholders. Each value rendered as a Go template.
timeout_sec	int		Per-call timeout.

Output

Field	Type	What
rows	[]map[string]any	One entry per row, column name → cell value.
row_count	int	Same as len(rows).
columns	[]string	Column names in order.

Example

{
  "env": [
    {"key": "USERS_DB", "desc": "Read-only DSN to the users database", "secret": true}
  ],
  "graph": {
    "nodes": [
      {
        "id": "lookup_user",
        "type": "db_query",
        "database": "USERS_DB",
        "query": "SELECT id, email, plan, created_at\nFROM users\nWHERE id = $1",
        "params": ["{{index .Event.Payload \"user_id\"}}"]
      }
    ]
  }
}

Downstream nodes reach {{index (index .Node.lookup_user.rows 0) "plan"}} (or {{.Node.lookup_user.row_count}} for a quick existence check).

Read-only, by convention

The connector framework's OpDestructive annotation has no analogue here — wick doesn't parse the SQL. Use a read-only DSN when wiring up the env var; if you genuinely need writes, scope the credential and confirm it's intentional.

Pair with

• branch — route on row_count for "found vs not found" flows.
• transform — reshape rows before downstream nodes.