The MCP server now closes the loop from schema selection to read-only query execution, while explicitly warning that principal and role claims are not authenticated.
The release reports a three-question retrieval gain across 98 tests, while preserving the hashed model for base installs and adding an opt-out for persisted deployments.
Microsoft’s Build Agent with AI preview turns schema and conversation context into instructions, source guidance and examples. Teams should treat that output as proposed configuration, not production policy.
Personal and company-scoped procedures can now steer AI Analyst, while the public marketplace still documents Claude and n8n as today’s execution surfaces.
The new Playground panel turns natural-language requests into runnable Visual Embed SDK code, while leaving execution and conflict resolution with the developer.
Adaptive Instructed-Retriever decides when one pass is enough and when a harder request merits more search, while keeping a fixed step ceiling.
The alpha Python package now derives dimension-like tables from the join graph, while prompt and validator fixes unblock CTEs, window functions and multi-table joins.
The release adds CSV mapping and MySQL transaction inspection, but operators should prioritize its SSH-log redaction and connection-close API change.
The September release adds natural-language filter expressions, generated column descriptions and consumer-facing Auto Insights—but keeps review and model choice with administrators and authors.
The new Apache-2.0 package filters schema objects before they reach a model. That closes one information-leak path while leaving query execution controls firmly in place.
Select AI’s new ADB-D capabilities collapse NL2SQL, feedback, memory and tool orchestration into database objects. The operational burden moves, not disappears.
The release binds local projects to Wren Cloud through ordinary Git remotes, adds a torch-free ONNX embedding path, and fixes two state-management traps.
The 26.16 release note assigns the preview to Looker (original), while Google’s operating guide says it requires Looker (Google Cloud core) and is unavailable on original instances.
The new ADB-D feedback path retrieves up to three prior corrections into each prompt, giving teams a schema-specific repair loop with limits they can inspect.
The release combines one-edit-per-file reconciliation, all-NULL join detection and explicit semantic deletion guards into a sharper maintenance boundary for agent-written dbt projects.
The new workflow generates golden question-SQL pairs, evaluates QueryData, groups failures and writes a revised context set without overwriting the baseline.
The new MIT project combines natural-language analysis, scheduled reports, Skills and MCP—but plainly says its guard is lexical and its permissions stop at the dataset.
The practical change is not another chat box: a plain-language request can now become an audience plan and an externally activated campaign, raising the bar for preflight checks.
The Sept. 4 expansion reaches AWS EU, Japan and Asia-Pacific routing settings—but production teams still have to enable, audit and pay for the control.
The September 2 behavior change preserves role-based access, but moves a missing grant from an HTTP failure into the answer stream — so clients must treat warnings as part of correctness.
The release closes a provenance gap between who originally said a fact and who approved it for use, while narrowing the new approval API to explicit fact IDs.
The new alpha toolkit compares executed results—not SQL strings—but its SQLite-only sandbox and zero-adoption launch make this an early implementation, not yet a production standard.
The public beta carries report permissions and row-level security into generation, but the resulting deck is still a point-in-time file with its own distribution risk.
The maintenance release separates two failure domains: validating imported semantic context, then constraining the SQL execution path.
The practical upgrade question is not which AI feature to enable first, but whether OAuth permissions, tenant audit settings and external credentials still form one governable path.
The release centralizes exhausted-budget handling, makes projects record a session-ceiling decision, and gives refusals historical billing context.
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.
The fresh enterprise text-to-SQL entrant makes MCP its supported agent surface, applies policy before query generation, and prices SaaS by accepted natural-language requests.
Databricks is moving its conversational analytics surface beyond answers and into external writes. The practical rollout boundary is now the combination of Unity AI Gateway, each user’s consented OAuth scopes, and permissions in the destination app.
The beta turns a conversational coding agent into an unattended Lakeflow task. Its most important design fact is not scheduling, but inherited tool reach without a human approval step.
Semi-Persistence cuts measured model swap cycles by 5.6×–19.9× and explicitly targets specialized text-to-SQL serving. The launch’s promised open-source directory is not yet on the public main branch.
The release couples specialized Spotter configurations with connector testing, local-time interpretation and separate raw-data export privileges.
Model choice, delta enrichment and background jobs move Enterprise AI NL2SQL from a demo path toward a service teams can run deliberately.
The new command-line surface keeps large query results outside the model context and gives coding agents and CI one deployment path for queries, pipelines and dashboards.
The open-source release adds reusable metrics, static validation and a CLI workflow that lets teams inspect analytics changes before they reach users.
The preview moves natural-language answers onto Fabric’s shared visualization foundation, with automatic charts across lakehouses, warehouses and Power BI semantic models.
The Rust runtime can coexist with v1, but stricter parsing, removed flags and package compatibility can change the metadata and workflows that downstream AI tools depend on.
The maintenance release hardens legacy-context migration, bounds a growing cache, pushes Spark limits down, preserves PostgreSQL numeric precision and separates timeouts from bad SQL.
The young self-hosted database agent now streams dashboards progressively while moving generation state to the server—but upgrades can require manual schema changes.
The new content-search option raises the file ceiling for Unity Catalog volumes, but it also creates a distinct retrieval and billing path inside Genie.
The coding and data agent can now read PDFs, spreadsheets, images and JSON inside a chat, adding a temporary context path alongside governed catalog metadata.
The release brings a new warehouse connector and a long dependency-security sweep, but it belongs to Wren AI’s archived legacy/v1 branch rather than the consolidated mainline.
A seven-line metadata change shows how multi-package release automation can leave Python users behind even after the underlying engine ships.
The community release adds native MiniMax configuration, an integrated terminal, and a wide set of quoting, validation, and streaming fixes.