Yes. Adaptrix connects to Snowflake with a read-only role, maps your databases, schemas and views into a semantic layer automatically, and runs the analysis with self-hosted LLMs. Nothing is copied to a third-party AI cloud: your warehouse stays the single source of truth, and prompts and results never go to a third-party AI provider.
Access is scoped entirely by Snowflake grants — Adaptrix sees exactly the objects you grant to its role, and nothing else.
A read-only role is granted usage on the databases and schemas you choose — production marts, a curated analytics schema, or a single project database.
Tables, standard views and secure views the role can select from. Row- and column-level policies you have defined in Snowflake keep applying.
Curated data products — the marts your data team already builds with dbt or SQL — are reused as-is, so their business logic is not duplicated.
Column names, data types and relationships are read from Snowflake's metadata (such as the Information Schema) to build the semantic layer automatically.
A dedicated virtual warehouse can be assigned to Adaptrix so its compute is isolated and its cost is visible as a single line in your Snowflake bill.
Adaptrix follows the value chain from seeing what happened to acting on it — here is what that sounds like on top of your warehouse.
Plain-language access to the models your data team already curates — no waiting for a dashboard request.
Automated root-cause analysis decomposes the change across dimensions instead of leaving you to write exploratory SQL.
Predictions trained on your warehouse history, with explainable drivers rather than a black box score.
Alerts and recommendations route the insight to the owner — computed inside your infrastructure.
An architecture-level view — read-only, incremental, and processed where you decide.
You create a Snowflake role with usage and select grants on the objects you choose. Adaptrix authenticates with that role and can never modify or delete data.
After an initial pass, only changed data is read, so queries stay light on your warehouse and answers reflect the latest loads.
Schemas, tables and columns are mapped to business terms, so a question like 'net revenue by region' resolves to the right mart without hand-built modeling.
The self-hosted open models that interpret questions and explain results run on Adaptrix's sovereign EU infrastructure — warehouse data is never sent to a public AI API.
This is an architecture-level description. Grants, warehouse sizing and sync scope are agreed during onboarding to match your Snowflake setup.
You chose Snowflake to centralize your data. Adaptrix adds AI analysis on top without creating a second copy in someone else's AI cloud.
GDPR-first: processing happens on Adaptrix's EU infrastructure, with a clear data-processing agreement — no transfers to third-country AI services.
Self-hosted open models answer your questions; no prompt, table or result ever reaches a public AI cloud.
Flat-fee licensing with no per-token charges, so exploratory, all-day querying never becomes a metered cost on top of your Snowflake bill.
If you are weighing Snowflake-native AI against a sovereign setup, start with the comparison.
How self-hosted open models on Adaptrix's EU infrastructure compare with AI that runs inside Snowflake's cloud.
The full self-hosted deployment model: self-hosted open models, no data egress, EU hosting (on-premises planned for Enterprise Plus).
Why mapping schemas to business language is the step that makes plain-language questions reliable.
Dashboards, root-cause analysis, predictions and alerts — everything that runs on top of the connector.
Cortex runs AI functions inside Snowflake's cloud and is scoped to Snowflake. Adaptrix runs self-hosted open models on its EU infrastructure and analyzes Snowflake alongside your other sources — ERP, CRM, files — in one semantic layer. See the full comparison page for details.
A read-only role with usage grants on the chosen databases and schemas and select grants on the objects to analyze. No ownership, write or admin privileges — your Snowflake access policies keep applying.
Reads are incremental after the initial pass, and you can assign Adaptrix a dedicated virtual warehouse to isolate and cap its compute. Actual consumption depends on data volume and query patterns, which is sized during onboarding.
Query results are processed by self-hosted open models on Adaptrix's EU infrastructure. Neither raw data nor prompts nor results are sent to a public AI service.
Adaptrix complements them. Marts and models you build with dbt remain the source of business logic — Adaptrix reads them and makes them queryable in plain language, and your existing BI dashboards keep working unchanged.
Bring a real question from your business and watch a live answer — reasoning, sources and math included — or model what it saves you first.