01 · Warehouse
Configured database adapters
Snowflake, Databricks, and Redshift adapters are present today for schema inspection and query execution. Credential mode, network path, and result retention are documented for the deployment.
Your stack · Integrations
Fit Spotonix to the stack you already have.
Current coverage, labeled clearly.
Spotonix is the AI analyst that interprets before it queries. This page distinguishes repository-evidenced database adapters from connectors that are coming or confirmed only for a specific engagement.
It does not collapse "can be integrated" into "generally available." Credential mode, network path, retained artifacts, and semantic coverage remain deployment-specific.
How it connects
Three connection points, with different maturity.
Spotonix connects to a warehouse, a configured model backend, and—where supported—existing semantic definitions. Each connection has its own implementation and evidence requirements.
02 · Semantic layer
Use governed definitions where supported
Power BI (PBIX) semantic models can be read today. dbt and LookML definitions can also inform the Context Graph, with supported scope confirmed for each deployment.
03 · Model provider
Configure a supported model backend
The current implementation supports Anthropic and OpenAI backends. Provider, model, endpoint, credential owner, and data-usage terms are confirmed for each deployment.
What connecting actually does
1
→
Configure a supported warehouse & model backend
2
→
Load the domain context supported for the deployment
3
→
Ask · Intent Algebra resolves a visible plan
4
Generate SQL · check bindings · execute through the configured connection
The interpretation and closure steps are the product; see Intent Algebra for how a question becomes a plan you can inspect before SQL executes, and Architecture for the deployment decisions that require technical review.
What connects today
Coverage, labeled honestly.
Every connection below carries an explicit status. Anything not yet generally available is marked coming or roadmap — we confirm scope and timing with design partners against your real environment rather than overstate it here.
Semantic layers · read the definitions your team already wrote
Power BI (PBIX)
LiveReads the measures and relationships in your Power BI semantic model. Deeper native DAX ingestion is in private beta — see Power BI.
dbt
EngagementPotential source of governed models and metrics. Supported scope must be demonstrated for the deployment before it is labeled live.
Looker / LookML
EngagementPotential source of governed dimensions and measures. Supported scope must be confirmed for the deployment.
Warehouse adapters
Snowflake
LiveConfigured adapter for schema inspection and query execution. Credential mode and network path are deployment-specific.
Databricks
LiveConfigured adapter for schema inspection and query execution against Databricks.
BigQuery
LiveConnects through the same SQLAlchemy adapter layer as Snowflake, Databricks, and Redshift. Credential mode and network path are deployment-specific.
Amazon Redshift
LiveConfigured adapter for schema inspection and query execution against Redshift.
BI assets · learned as context
Dashboards & metrics as context
Coming soonReading existing BI assets (dashboards, saved metrics, Explores) as additional context for the Context Graph is in development — a source of meaning to learn from, not a surface Spotonix replaces.
Model backends
Anthropic and OpenAI
LiveConfigurable backends with model selection. Credential ownership, endpoint, and provider terms are deployment decisions.
A connector moves to live only when implementation and supported scope are evidenced—not because the integration is strategically useful.
Where the boundary is
What connecting Spotonix does not do.
Fit is easier to judge when the negative space is explicit. These are deliberate boundaries, not gaps we intend to close.
- It is not a new warehouse or a migration. Spotonix sits on top of the warehouse you already run; your data does not move and your platform team does not adopt a new system of record.
- It is not a replacement warehouse. Spotonix queries a configured warehouse. Retained results, caches, logs, and Context Graph artifacts are documented per deployment.
- It does not require you to discard your semantic layer. Where a connector is supported, existing definitions can inform the Context Graph. Coverage must be verified first.
- It does not make a connector equivalent to governance. Definitions still need owners, review rules, and a demonstrated binding path into the analysis plan.
- It does not claim arbitrary-provider support. The current configurable backends are Anthropic and OpenAI.
Deployment controls
Confirm the boundary, then connect.
Connector availability does not answer hosting, identity, permission, or retention questions. See Architecture for the review model.
Hosting topology
EngagementConfirm where the application, Context Graph, logs, caches, and query results are processed and stored for the selected deployment.
Warehouse credential mode
EngagementConfirm whether execution uses a service identity, delegated user identity, or another mode, then test the expected warehouse permissions.
Model credential and endpoint
LiveConfigure Anthropic or OpenAI, then document model, endpoint, credential owner, region, retention, and provider terms.
Auditable by construction
EngagementThe analysis plan is visible in the workflow. Stable plan identity, retention, query linkage, and export are confirmed as evaluation requirements rather than implied guarantees.
Deployment topology and connection specifics are confirmed per engagement. We work these through with design partners against your real environment, not a reference diagram.
Keep reading