Semantic layer solution

A semantic layer for every way your team uses data.

A semantic layer defines business metrics, dimensions, and relationships once so people, BI tools, and AI agents use the same definitions. DOSI validates compatible Apache Ossie models and compiles supported metric queries into native warehouse SQL. Your team can review the definitions and inspect how results are produced.

Get Started Free See how it works

Define business meaning once. Use it everywhere.

A semantic model brings business logic and its data mappings together. Give teams, BI tools, and agents a shared definition to work from as they ask different questions.

Apache Ossie

Metrics

booked_revenueSUM((lines.gross_sales_cents-lines.discount_cents)*0.01)

order_countCOUNT(DISTINCT orders.order_id)

order_return_ratereturned_order_count*1.0/NULLIF(order_count,0)

Dimensions

citycustomers.cityString

categoryproducts.categoryString

delivered_atorders.delivered_atDate

↔ Swipe to explore the full model.

Select an entity to explore its relationships and data mapping.

Ontology

Customer

Source mappings

crm.dim_customer

Business relationships

placed_byOrder → Customer

held_byMembership → Customer

Trace business meaning to the datasets and fields behind it.

Source	Semantic model	Ontology
crm.dim_customer	customers.customer_id	Customerid
commerce.fct_orders	orders.order_id	Orderid
commerce.fct_order_lines	lines.order_line_id	OrderLineid
catalog.dim_product	products.product_id	Productid
logistics.fct_order_logistics	deliveries.order_id	Deliveryid
finance.fct_order_line_costs	line_costs.order_line_id	LineCostid

Explore metrics and dimensions, business relationships, and source mappings in one model.

Your business definitions. Independent of your stack.

Capture business definitions from existing SQL, supported BI dashboards, and new requirements. Keep them in Apache Ossie, an open, vendor-neutral standard your team can review, maintain, and reuse.

Start with the work you already do.

Your SQL, dashboards, and new business questions already contain modeling knowledge. Datus Agent helps turn it into reusable semantic definitions.

Existing SQL

SELECT customer_region,
  SUM(net_revenue)
FROM sales
GROUP BY customer_region

Extract from existing SQL

Use historical queries and SQL files as evidence. Capture reusable calculations, filters, dimensions, and source mappings in a semantic model.

Calculations, filters & grouping

Dashboard query logic

Build from BI dashboards

Start with selected queries from a supported BI dashboard. Recover the logic behind its charts and bring those definitions into a shared model.

Chart queries & source data

New business requirement

Model as you build

Describe a new business requirement. Work with Datus Agent to define datasets, relationships, and metrics alongside development, then validate the model.

Business requirements & live schema

Shared semantic definitions

Datasets
Metrics
Dimensions
Relationships

Draft with Datus Agent. Review definitions and validate the model before using it.

See the modeling workflow

Open definitions. Native execution.

Apache Ossie is the core specification behind DOSI’s semantic execution. DOSI reads compatible Ossie models, validates their definitions, and compiles semantic queries into native SQL for your configured warehouse.

Select a stage to see its role.

1 / 4

Shared definitions

Start with reviewed definitions drawn from your SQL, dashboard queries, or new business requirements. Keep datasets, dimensions, metrics, and relationships together in a compatible Ossie model.

MetricsDimensionsRelationships

Start Datus with your existing business definitions.

Use your reviewed Ossie model as a starting point in Datus. Give teams and agents an open, vendor-neutral semantic layer to build on.

Get Started Free

Apache Ossie (incubating) is the open specification. DOSI is licensed under Elastic License 2.0. Datus Agent is open source.

Business context your team can review and evolve.

Bring the knowledge in your warehouses, documents, SQL, and ETL into a shared modeling workflow. Datus Agent helps draft definitions; your team uses Datus Studio to inspect, refine, and verify them.

Initialize project context from the connected data source and selected SQL or documents.

Attach project files to a conversation and work with Datus Agent to draft datasets and metrics.

northstar_premier.yamlSource

version: 0.2.0.dev0
semantic_model:
- name: northstar_premier
  # …
  datasets:
  - name: customers
    source: northstar_living.dim_customer
    primary_key:
    - customer_id
    # …
  metrics:
  - name: booked_revenue
    expression:
      dialects:
      - dialect: ANSI_SQL
        expression: SUM((lines.gross_sales_cents-lines.discount_cents)*0.01)
  # …

Trace the full source lineage, review a YAML diff, and approve changes before merging them into the shared model.

Select an object to inspect its definition

Scroll sideways to explore the full lineage.

Source tablesSemantic datasetsAtomic metricsDerived metrics

Recorded Northstar lineage · 9 tables · 11 datasets · 12 metrics

contribution_margin

Derived metricsDataset mapping

booked_revenue

refund_amount

product_cost

recovered_product_cost

outbound_shipping

payment_fees

return_handling

Metrics

COALESCE(booked_revenue,0)-COALESCE(refund_amount,0)-COALESCE(product_cost,0)+COALESCE(recovered_product_cost,0)-COALESCE(outbound_shipping,0)-COALESCE(payment_fees,0)-COALESCE(return_handling,0)

Recorded Northstar lineage · 9 tables · 11 datasets · 12 metrics

Align return handling with currency units

Open

MR #12 · Model author · northstar_premier

subject/semantic_models/northstar_premier.yaml

Merge blockedAt least one approvalAll conflicts resolvedAll checks pass or waived

northstar_premier / contribution_marginModified

1 name: contribution_margin2 expression:3   dialects:4   - dialect: ANSI_SQL5     expression: >-6       COALESCE(SUM((lines.gross_sales_cents-lines.discount_cents)*0.01),0)7       -COALESCE(SUM(refunds.merchandise_refund_cents*0.01),0)8       -COALESCE(SUM(line_costs.cogs_cents*0.01),0)9       +COALESCE(SUM(returns.recoverable_cogs_credit_cents*0.01),0)10       -COALESCE(SUM(deliveries.shipping_cost_cents*0.01),0)11       -COALESCE(SUM(line_costs.payment_fee_cents*0.01),0)12−Previous definition:       -COALESCE(SUM(returns.processing_cost_cents),0)12+Proposed definition:       -COALESCE(SUM(returns.processing_cost_cents*0.01),0)

All conflicts resolved

All checks pass or waived

Explore Datus Studio Explore Datus Agent

One semantic model for BI, applications, and AI agents.

Make reviewed definitions available where your team works. Choose the supported interface that fits each consumer.

northstar_premier.yamlExcerpt

1version: 0.2.0.dev02semantic_model:3- name: northstar_premier4  # …5  datasets:6  - name: customers7    source: northstar_living.dim_customer8    primary_key:9    - customer_id10    # …11  metrics:12  - name: booked_revenue13    expression:14      dialects:15      - dialect: ANSI_SQL16        expression: SUM((lines.gross_sales_cents-lines.discount_cents)*0.01)17  # …

DOSISemantic execution

BI & analytics

SQL & results

Metabase

Superset

Tableau

AI agents

MCP

Claude Code

Codex

Datus Agent

Applications

REST / Python

Python

JavaScript

Datus Studio

FAQ

Questions about your semantic layer

How does a semantic layer differ from a semantic model?

A semantic model records business definitions, including datasets, dimensions, relationships, and metrics. A semantic layer makes those definitions usable by consumers. In Datus, DOSI validates compatible Apache Ossie models and compiles supported metric queries into SQL, while interfaces make that functionality available to teams, applications, and agents.

Where does DOSI execute a semantic query?

DOSI validates the compatible model and compiles the requested metrics into native SQL. The configured warehouse executes that SQL against its data. Your team can inspect the generated query and source mappings, while CLI, REST, Python, or MCP interfaces provide access to the resulting data.

What can I reuse when I change warehouses?

Compatible Apache Ossie definitions can be reused while DOSI compiles queries for a supported target dialect. Check connection settings, source mappings, SQL expressions, and feature support for the new source. Sharing definitions across warehouses does not mean that a query automatically joins live data from several warehouses.

How can an AI agent use the semantic layer?

An agent can use MCP tools to discover metrics and dimensions, compile supported queries, and retrieve results from a configured source. Your team reviews the underlying business definitions and can inspect the generated SQL. The MCP guide covers connection details and the tools available to a client.

Is DOSI open source?

DOSI is licensed under Elastic License 2.0. Apache Ossie (incubating) provides the open specification for compatible semantic models, and Datus Agent is open source. The open model format and the software licenses apply to different parts of the system; they should be evaluated separately when choosing your deployment and workflow.

Explore the ideas. Put them into practice.

What is a semantic layer?Semantic model examples Apache Ossie and DOSI quickstart

Put shared business context to work.

Give your team and agents a semantic model they can understand, review, and reuse.

Get Started Free