live wire
▸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 says Cortex Analyst apps keep working, but recommends Cortex Agents; semantic views and verified queries carry overSnowflake Documentation▸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 says Cortex Analyst apps keep working, but recommends Cortex Agents; semantic views and verified queries carry overSnowflake Documentation
nl2sql.ai
analysisRelease analysis

WrenAI 0.16.0 adds StarRocks—and fixes two quiet ways an agent can hit the wrong boundary

The connector is the headline, but project-scoped secrets and comment-safe query wrapping are the changes operators should regression-test first.

WrenAI 0.16.0: StarRocks support, plus safer secrets and SQL wrapping.
AI-generated illustration
By The News Desk· Oct 16, 2026the quick take — two AI hosts go live when you do

WrenAI 0.16.0 adds StarRocks as a first-class data source, expanding the open-source semantic layer’s connector surface. The same release also changes how Wren resolves project secrets and wraps generated SQL before applying limits—two less visible fixes that matter more than a new connector when one service hosts several projects or lets agents execute model-written queries. WrenAI release notes

The selected project now selects the secret

The project-secret fix is unusually concrete. Wren’s new regression test creates two projects with separate .env files and DuckDB databases, then runs both query --mdl and serve mcp --project from three working-directory positions: the other project, the selected project and outside either project. Every path must resolve the selected project’s database and return its value, not the working directory’s. Project-secret commit

That is an isolation boundary, not merely configuration polish. In a shared deployment, an operator may keep several projects with similarly named environment variables under one Wren home. Before upgrading, teams should reproduce the old failure with harmless test databases; after upgrading, they should run queries and MCP startup from inside and outside each project directory and verify the connection target. The upstream test demonstrates the expected invariant: the selected project controls both profile and secret resolution. Project-secret commit

A SQL comment should not swallow the safety wrapper

The release also moves wrapped subqueries onto separate lines. That prevents a generated trailing -- comment from consuming the closing parenthesis and appended LIMIT. The added ClickHouse tests check both ordinary execution with LIMIT 5 and dry-run execution with LIMIT 0; the release notes list a companion semicolon-before-comment cleanup. Comment-wrapping commit WrenAI release notes

For agent workloads, formatting is executable behavior. A syntactically valid query can become invalid—or lose the limit the caller intended—when it is embedded as a subquery. A focused upgrade test should therefore include line comments, trailing semicolons, multiline SQL and dry runs across every connector in use, rather than validating only clean SELECT statements. The upstream commit directly demonstrates the ClickHouse case; the release notes describe the broader fix but do not claim complete coverage for every connector. Comment-wrapping commit WrenAI release notes

StarRocks support will attract the attention, but the operational lesson is narrower: connector count does not define readiness. Upgrade acceptance should prove that project selection controls credentials and that generated SQL survives the wrapper enforcing execution limits.

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