live wire
IBM makes watsonx Orchestrate AgentOps, custom LLM judging and Bedrock-agent discovery generally availableIBMSchemaGate 0.1.45 fixes broken Oracle ADB wallet connections and an OCI stack pinned 28 releases behindSchemaGatePDI’s Amazon Quick procurement agent grounds spend answers in vendor, category and contract contextAWS Business Intelligence BlogBigQuery’s ML.METRICS example returns 0.84 accuracy but 0.30 macro-F1 on the same 100-row classification queryGoogle Cloud BigQuery docsSchemaGate 0.1.44 auto-selects sentence embeddings, lifting bundled-schema retrieval from 90/98 to 93/98SchemaGateSchemaGate 0.1.43 adds read-only SQL execution with per-principal table checks—and documents unauthenticated client assertionsSchemaGateDatabox adds reusable AI Analyst Skills with personal/company scope, auto-matching and marketplace installsDataboxFabric previews an AI builder for data-agent instructions, source guidance and example queriesMicrosoft FabricDatabricks trains data-agent retriever to stop early or spend bounded extra search steps, reporting 5.8-second latencyDatabricksThoughtSpot adds SpotterCode coding agent to its Visual Embed PlaygroundThoughtSpotLongMemEval-S audit: 67–73% of restore-fixable 80k-budget errors came from evicted evidence under three policiesarXivSchemaGate 0.1.42 adds dimension-aware retrieval and fixes complex multi-table SQL promptsSchemaGateSnowflake agent toolsets can silently drop inherited tools when callers lack accessSnowflake DocumentationLooker’s VS Code extension reaches GA with MCP-assisted LookML generation, editing and validationGoogle Cloud Looker release docsIBM makes watsonx Orchestrate AgentOps, custom LLM judging and Bedrock-agent discovery generally availableIBMSchemaGate 0.1.45 fixes broken Oracle ADB wallet connections and an OCI stack pinned 28 releases behindSchemaGatePDI’s Amazon Quick procurement agent grounds spend answers in vendor, category and contract contextAWS Business Intelligence BlogBigQuery’s ML.METRICS example returns 0.84 accuracy but 0.30 macro-F1 on the same 100-row classification queryGoogle Cloud BigQuery docsSchemaGate 0.1.44 auto-selects sentence embeddings, lifting bundled-schema retrieval from 90/98 to 93/98SchemaGateSchemaGate 0.1.43 adds read-only SQL execution with per-principal table checks—and documents unauthenticated client assertionsSchemaGateDatabox adds reusable AI Analyst Skills with personal/company scope, auto-matching and marketplace installsDataboxFabric previews an AI builder for data-agent instructions, source guidance and example queriesMicrosoft FabricDatabricks trains data-agent retriever to stop early or spend bounded extra search steps, reporting 5.8-second latencyDatabricksThoughtSpot adds SpotterCode coding agent to its Visual Embed PlaygroundThoughtSpotLongMemEval-S audit: 67–73% of restore-fixable 80k-budget errors came from evicted evidence under three policiesarXivSchemaGate 0.1.42 adds dimension-aware retrieval and fixes complex multi-table SQL promptsSchemaGateSnowflake agent toolsets can silently drop inherited tools when callers lack accessSnowflake DocumentationLooker’s VS Code extension reaches GA with MCP-assisted LookML generation, editing and validationGoogle Cloud Looker release docs
nl2sql.ai
releaseRELEASE

Dex 1.9.1 turns ClickHouse Cloud cost controls and query shape into an agent contract

The release adds a guarded ClickHouse Cloud deployment and labels every query result as columnar, two small-looking changes aimed at making warehouse actions safer for coding agents.

Guarded ClickHouse Cloud cost checks contrasted with self-describing query results.
AI-generated illustration
By The News Desk· Aug 31, 2026the quick take — two AI hosts, this story only

What shipped

Exmergo released Dex 1.9.1 on August 31 with a ClickHouse Cloud connector and a self-describing query envelope. The release also includes fixes for seeded-schema exploration, tighter inventory payload caps, a warning for unrecognized dialects that would otherwise silently full-scan, and changes that let Dex skills run without depending on the caller’s Python project. The release notes list all ten merged changes.

The headline addition is not just another connection string. Dex now treats ClickHouse Cloud as a distinct guarded deployment rather than reusing its self-hosted ClickHouse accounting. In cloud mode it reports compute-seconds, derives approximate compute-unit hours from live replica memory and can translate that estimate into dollars when a real CU price is configured. If capacity information is missing, denied, partial or inconsistent, Dex refuses before billed work; self-hosted ClickHouse retains its database-load model. That behavior is documented in the merged connector change.

Why the guardrails matter

The connector work also adds a narrow live-cloud integration suite. The project says its test workflow checks usage before opening a SQL connection, uses a service-scoped API key plus read-only database credentials, and refuses once a protected daily threshold is reached. Query execution is still bounded with ClickHouse’s execution-time and bytes-read limits. The pull request and its CI changes describe the protected environment and preflight admission check.

For teams letting coding agents inspect warehouses, that separation is useful: “read only” does not mean “cost free.” A valid analytical query can still wake cloud capacity or scan more than intended. Dex’s design makes estimated consumption part of the confirmation and ledger path instead of leaving cost as an operational afterthought. That is an interpretation of the release’s documented controls, not a claim that the controls eliminate billing risk. Dex describes its connector budgets and local spend ledger in the project documentation updated by the same change.

A result that explains itself

The second change is smaller in code but directly relevant to agent integrations. Single-query responses, batch responses and each batch result now carry "shape": "columnar"; the existing columns, types and cells layout does not change. That lets a caller discover how to decode the payload from the response itself rather than relying on separate documentation or source inspection. The merged query-envelope change documents the contract and regression coverage.

This is the practical theme of Dex 1.9.1: make both sides of an automated warehouse action explicit. Before execution, the tool identifies the deployment’s cost model and can refuse when it cannot establish safe bounds. After execution, the result states its own data shape. Neither feature improves text-to-SQL accuracy by itself, but together they reduce two integration ambiguities that become expensive when software—not a person—is deciding the next step. Both changes ship in the v1.9.1 release.

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.