Data Sources

The Data Quality Agent connects to a data source and runs data quality checks against it. The databaseType setting selects between two families:

FHIR — a FHIR R4 server queried with CQL.
SQL — a relational database or a directory of CSV files queried with SQL.

FHIR servers

When databaseType is FHIR, the agent acts as a FHIR R4 REST client and runs CQL checks against the configured server:

Setting	Description
`fhirUrl`	Base URL of the FHIR R4 server (e.g. `http://localhost:8080/fhir`).
`fhirUsername`	Optional username for basic authentication.
`fhirPassword`	Optional password for basic authentication (Base64 encoded).

Only the databaseType=FHIR configuration is required to use this data source; the agent connects to the server at fhirUrl and authenticates with the provided credentials when given.

SQL driver selection

When databaseType is SQL, the agent looks at the JDBC URL (sqlUrl):

URLs starting with jdbc:calcite: are handled by Apache Calcite. Only two sub-forms are accepted (see below); anything else under that prefix is rejected.
Any other jdbc: URL is passed to the standard JDBC DriverManager and uses whichever driver is on the classpath. The agent ships with these drivers bundled:

JDBC URL prefix	Driver / engine
`jdbc:postgresql://...`	PostgreSQL
`jdbc:mysql://...`	MySQL
`jdbc:sqlite:...`	SQLite
`jdbc:calcite:directory=<dir>`	Apache Calcite / CSV

For the standard JDBC URLs the agent connects using the supplied username/password. Additional database drivers can be supported by adding them to the agent's classpath — no code change is required, since these URLs simply go through DriverManager.

Querying CSV files with Calcite

Only these two Calcite URL forms are supported:

Form	Purpose
`jdbc:calcite:directory=<dir>`	Query a directory of CSV files (auto-generated model)
`jdbc:calcite:model=<model-file>`	Query using a custom Calcite JSON/YAML model file

Example

databaseType: SQL
sqlUrl: jdbc:calcite:directory=/opt/omop/csv

If /opt/omop/csv/person.csv and /opt/omop/csv/specimen.csv exist, checks such as SELECT COUNT(*) FROM person work as expected.

CSV conventions

Header row: the first row must be the column names; column types are inferred from the values.
Table names: each file becomes a table named after the file's base name (e.g. person.csv → person).
Case sensitivity: the directory= form uses lex=MYSQL, so unquoted identifiers are matched case-insensitively and the common OMOP lowercase style (person, specimen) works out of the box.

Custom model file

For more control (custom delimiters, explicit schemas, multiple directories), use a standard Calcite model file instead:

json

{
  "version": "1.0",
  "defaultSchema": "CSV",
  "schemas": [
    {
      "name": "CSV",
      "type": "custom",
      "factory": "org.apache.calcite.adapter.csv.CsvSchemaFactory",
      "operand": { "directory": "/opt/omop/csv" }
    }
  ]
}

sqlUrl: jdbc:calcite:model=/opt/omop/model.json

Note that with a custom model file, identifier quoting and case-sensitivity follow the conventions defined by the model rather than the defaults applied automatically to the directory= form.

Data Sources ​

FHIR servers ​

SQL driver selection ​

Querying CSV files with Calcite ​

CSV conventions ​

Custom model file ​