Datus + Google BigQuery

A semantic layer for BigQuery, with business meaning built in.

Keep data in BigQuery while your team reviews metrics in an Apache Ossie model. DOSI compiles a metric request into GoogleSQL and sends it to BigQuery; Datus Agent has a separate native BigQuery adapter for data work.

Get Started Free See the connection steps

How the pieces connect

Existing SQL + business rulesTeam reviewApache Ossie model

Agents + apps · MCP / API / CLIDOSIGoogleSQL via jobs.query

BigQuery project + dataset

Results return through DOSI to your tools

DOSI sends a query to the configured BigQuery project; BigQuery owns execution and billing.

Project + dataset awareNative BigQuery queriesExperimental DOSI execution

Where Datus fits

Two ways to model metrics over BigQuery.

Looker can model BigQuery data in LookML for dashboards and applications. DOSI uses a separately reviewed Apache Ossie model to serve named metrics through agent-friendly interfaces; BigQuery still executes the query.

BigQuery + Looker

BigQuery tables + LookML model

Looker semantic model

Looker dashboards and applications

BigQuery query jobs

Looker is a separate Google BI product that connects to BigQuery.

BigQuery + Datus / DOSI

Team-reviewed Apache Ossie model

DOSI semantic layer

Agents and apps · MCP / API / CLI

BigQuery jobs.query

The model is managed separately; DOSI BigQuery execution is experimental.

Both paths keep query execution in BigQuery. BigQuery also supports SQL views for reusable query logic.Google: Looker with BigQuery DOSI BigQuery connector

Make the definition reviewable

Describe revenue, dimensions and relationships in an Apache Ossie model. Teams can check the business logic before applications or agents reuse it.

Preserve the BigQuery boundary

Queries run in your configured BigQuery project and dataset. DOSI does not copy the underlying tables into another analytics store.

Separate the two connection paths

Datus Agent has a native BigQuery adapter for metadata and data tasks. DOSI has its own BigQuery execution connector for semantic metric queries.

Choosing a BigQuery semantic layer

One semantic model across your BigQuery workflows.

Decide where shared metrics are defined, how existing models contribute, and how BI tools and agents consume them.

Where should business definitions live?

Keep existing dbt transformations and BigQuery tables. Define the metrics you want agents and applications to share in a reviewed model. Check how changes reach each consumer; connecting to the same warehouse does not automatically synchronize LookML, dbt metrics and Ossie.

Explore the semantic layer approach

How do existing dbt models and views fit?

Use curated BigQuery tables or views as the data foundation. Keep transformations in dbt, and describe shared metric expressions, dimensions and relationships in the semantic model. Start with a small set of definitions and validate them against existing reporting logic.

See semantic model examples

Which tools will consume the metrics?

Separate a warehouse connection from a semantic-model integration. An application must call the supported DOSI interface to use its definitions. A BI tool querying BigQuery directly does not automatically use the Ossie model. Prove one complete consumer-to-query path first.

Using DOSI in agents

A concrete workflow

A metric request, an explicit project, an inspectable query.

Provide a reviewed model and a BigQuery connection profile before asking DOSI for a named metric.

dosi-connections.yaml

dosi-connections.yamlYAML

datasources:
  bigquery:
    type: bigquery
    project: ${BIGQUERY_PROJECT}
    dataset: analytics
    credentials_path: /path/to/service-account.json
    location: US
    billing_project_id: ${BIGQUERY_BILLING_PROJECT}

DOSI CLI · example request

dosiCLI

dosi query --model model.yaml \
  --metrics revenue --group-by orders.status \
  --execute --connection bigquery

This is a configuration and command example, not a live BigQuery run. Use your own validated model and service account. The DOSI BigQuery executor is currently experimental.

Connect BigQuery

Set the data project and billing project deliberately.

The DOSI connector uses a service-account JSON key and the BigQuery jobs.query API. A BigQuery project is the catalog; its dataset is the next namespace level.

DOSI BigQuery connector docs

01
Grant the right access
Allow the service account to read the referenced data and create jobs in the billing project. Choose the correct dataset location, such as US or EU.
02
Configure one credential source
Set credentials_path, credentials_info or credentials_base64 in the DOSI profile, but not more than one. DOSI does not use Application Default Credentials for this connector.
03
Validate the model and query
Check the dataset names and metric definitions, then run a bounded request and inspect the SQL, result and BigQuery job cost.

DOSI BigQuery execution is documented as experimental. Datus Agent’s native BigQuery adapter is separate and supports its own authentication options; do not assume a setting from one connection path works in the other.

Datus Agent BigQuery adapter docs

Useful for teams

Share a definition without moving the data.

A common model is useful when multiple people and tools ask the same business question of BigQuery.

Consistent metrics for AI analysis

Give agents a named metric to call, with a definition your team can inspect, rather than leaving them to infer a calculation from table names.

Explicit project boundaries

Keep data-project, dataset, job location and billing-project choices visible in configuration so a query runs in the intended context.

FAQ

BigQuery semantic layer FAQ

How shared semantic definitions fit your BigQuery stack.

How is DOSI different from Looker on BigQuery?

Looker uses a LookML model to serve BI dashboards and applications over BigQuery. DOSI reads a separately managed Apache Ossie model and exposes named metrics through MCP, API and CLI interfaces. Both use BigQuery for query execution.

Does DOSI store a copy of my BigQuery tables?

No. DOSI compiles a metric request and sends SQL to the BigQuery project configured in its connection profile. BigQuery executes the job.

How does an Apache Ossie model connect business metrics to BigQuery?

The model maps business datasets to BigQuery tables or views and defines metric expressions, dimensions and relationships. DOSI uses those definitions to compile a metric request into GoogleSQL, then sends it to the configured BigQuery project. Teams can inspect the shared definition alongside the generated SQL and results.

Do I need Looker to use a semantic layer on BigQuery?

No. Looker is one modeling and consumption path. Datus/DOSI uses compatible Apache Ossie definitions and queries BigQuery separately. Keep Looker for existing dashboards and use DOSI interfaces to serve shared metrics to agents and applications.

Can I reuse dbt models and BigQuery views?

They can provide the data foundation for a semantic model. Start with curated outputs and document their grain, metric expressions and relationships. Keeping dbt transformations does not require moving the data. Review the selected definitions against existing reports before sharing them.

Start with a BigQuery metric your team can verify.

Review the definition, set a bounded query context and inspect what runs in BigQuery.

Get Started Free Read connector docs