live wire
▸Snowflake agent evals can pin judge versions; unpinned metrics may roll forward and shift scoresSnowflake▸ThoughtSpot makes Spotter 3 default for new instances; it requires download rights and always shares query data with an LLMThoughtSpot documentation▸MotherDuck adds image classification and object location to prompt_jev(), returning typed SQL outputs with confidence scoresMotherDuck▸BigQuery conversational analytics adds Montréal and Toronto; mixed-region agents still default to globalGoogle Cloud Documentation▸Fabric enables Azure OpenAI Copilot access by default, but OpenAI-subprocessor and cross-region processing/storage switches remain offMicrosoft Learn▸Snowflake Cortex Agents now continue by default when named tools are inaccessible, emitting warnings instead of failing the runSnowflake documentation▸Databricks Genie One can use group memberships to tailor answers, while Unity Catalog remains the access-control boundaryDatabricks documentation▸Fabric data agents can query Power BI models with Read alone; Build and workspace access are unnecessaryMicrosoft Learn▸Fabric data-agent Git configs expose column flags, but selecting a table still includes every columnMicrosoft Learn▸Google Data Agent Kit offers remote Spark kernels, but notebook exports can still move results into local Pandas or ArrowGoogle Cloud Documentation▸Cube says production analytics agents need explicit partial-result states, caller-filtered model search and durable artifactsCube▸Thinking Machines names Bridgewater behind ReViSQL-K2.6, citing 92.97% SC-16 accuracy at 15% of frontier-model costThinking Machines Lab▸AWS says a mismatched data/tool/prompt release cut its Field Sales agent pass rate from 78% to 38% in pre-productionAWS Artificial Intelligence Blog▸NL2SQL Multi-Agent ships evaluation artifacts across BIRD, CSpider, Archer and KDD DataSpace—but no software licenseGitHub▸Snowflake agent evals can pin judge versions; unpinned metrics may roll forward and shift scoresSnowflake▸ThoughtSpot makes Spotter 3 default for new instances; it requires download rights and always shares query data with an LLMThoughtSpot documentation▸MotherDuck adds image classification and object location to prompt_jev(), returning typed SQL outputs with confidence scoresMotherDuck▸BigQuery conversational analytics adds Montréal and Toronto; mixed-region agents still default to globalGoogle Cloud Documentation▸Fabric enables Azure OpenAI Copilot access by default, but OpenAI-subprocessor and cross-region processing/storage switches remain offMicrosoft Learn▸Snowflake Cortex Agents now continue by default when named tools are inaccessible, emitting warnings instead of failing the runSnowflake documentation▸Databricks Genie One can use group memberships to tailor answers, while Unity Catalog remains the access-control boundaryDatabricks documentation▸Fabric data agents can query Power BI models with Read alone; Build and workspace access are unnecessaryMicrosoft Learn▸Fabric data-agent Git configs expose column flags, but selecting a table still includes every columnMicrosoft Learn▸Google Data Agent Kit offers remote Spark kernels, but notebook exports can still move results into local Pandas or ArrowGoogle Cloud Documentation▸Cube says production analytics agents need explicit partial-result states, caller-filtered model search and durable artifactsCube▸Thinking Machines names Bridgewater behind ReViSQL-K2.6, citing 92.97% SC-16 accuracy at 15% of frontier-model costThinking Machines Lab▸AWS says a mismatched data/tool/prompt release cut its Field Sales agent pass rate from 78% to 38% in pre-productionAWS Artificial Intelligence Blog▸NL2SQL Multi-Agent ships evaluation artifacts across BIRD, CSpider, Archer and KDD DataSpace—but no software licenseGitHub
nl2sql.ai
releaseRelease analysis

Snowflake’s semantic contract can now ship through the same deployment gate as the database

DCM Projects now manages metrics, relationships, AI instructions and verified queries declaratively—but its governance envelope still has edges.

Semantic view code on one side, governed warehouse objects on the other.
AI-generated illustration
By The News Desk· Sep 30, 2026the quick take — two AI hosts go live when you do

Snowflake has made DEFINE SEMANTIC VIEW generally available in DCM Projects. The change lets teams manage a semantic view’s tables, relationships, facts, dimensions, metrics, AI instructions and verified queries as a declarative definition rather than treating the agent-facing model as a separately edited artifact. Snowflake’s release note is short, but the operational consequence is larger: the contract that grounds Cortex Analyst and Cortex Agents can now move through a plan-and-deploy workflow.

The semantic layer enters the deployment pipeline

DCM Projects stores desired state in SQL and YAML files, supports Git-backed workflows, previews changes with PLAN, and applies them with DEPLOY. Snowflake describes the system as a version-controlled, repeatable route across development, staging and production environments. Its DCM overview also says deployment artifacts and definition files are retained by the DCM project object.

That matters because a Snowflake semantic view is not just documentation. It holds the logical tables, joins, facts, dimensions and metrics that translate business language into warehouse logic. Snowflake says Cortex Agents read the semantic-view definition and generate SQL against the physical tables, while BI tools and analysts can reuse the same metric definitions. The semantic-view documentation positions the object as the shared business abstraction for both AI and conventional analytics.

A team can therefore review a metric formula, relationship or verified query in the same change process as the underlying warehouse objects. DCM reconciles the full semantic-view definition on each definition change. The supported-entities guide also makes an important limitation explicit: constraints and relationships must be named, and tags on a semantic view or its members cannot be added or changed through CREATE OR ALTER SEMANTIC VIEW; existing tags are preserved. Those constraints are documented here.

Governance is moving into code, but not all at once

Snowflake also released ATTACH MASKING POLICY for DCM Projects in public preview. It can declaratively attach a masking policy to columns in tables, Iceberg tables, views and dynamic tables, with DCM managing that attachment through the deployment lifecycle. The masking-policy release note does not list semantic-view members as attachment targets.

The practical rollout is therefore two tracks: semantic definitions are GA, while this new policy-attachment mechanism remains preview. Teams should promote semantic-view changes through PLAN and environment-specific tests, but keep separate checks for the policies on the physical objects those views reference. The new release makes the agent’s vocabulary deployable as code; it does not make every governance control part of one atomic semantic-view definition.

sources

Filed by The News Desk. Corrections: desk@nl2sql.ai · Our standards →

comments · 0

    Comments are moderated before they appear. Your email is used once to confirm it is you — never shown, never sold. Corrections and questions get an answer from the desk when we have one.

    more from the desk