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

Chat2DB fixes the invisible half of Oracle EXPLAIN PLAN

The 5.3.7 beta now writes a uniquely tagged plan, reads it back through DBMS_XPLAN and restores the client transaction—turning a success message into an inspectable result.

Before-and-after Oracle EXPLAIN PLAN flow with visible plan output.
Side by side: what changed
By The News Desk· Sep 21, 2026the quick take — two AI hosts go live when you do

Chat2DB 5.3.7-beta.4 fixes a deceptively important Oracle client failure: EXPLAIN PLAN could succeed while showing the user no plan. The September 21 prerelease now returns the generated execution plan as a result set, alongside two desktop-updater reliability fixes.

Why “success” was not enough

Oracle’s EXPLAIN PLAN does not return the plan as a normal query result. Oracle documents that the statement inserts one row per plan step into an output table—PLAN_TABLE by default—and that a separate query is needed to inspect those rows. That behavior is valid at the database layer, but it meant Chat2DB’s generic execution path could report a successful command with nothing useful in the result grid.

The new Oracle-specific path closes that gap. Chat2DB first checks that PLAN_TABLE is available, generates a unique statement identifier, runs the explanation under that identifier, then reads the formatted plan through DBMS_XPLAN.DISPLAY. If the plan table is missing, the client returns an actionable setup message rather than an empty result.

The transaction boundary matters

The implementation also handles a less visible constraint. Oracle classifies EXPLAIN PLAN as DML and does not implicitly commit its plan-table rows. Chat2DB therefore temporarily disables autocommit when needed, writes and reads the plan in one transaction, rolls back those temporary rows, and restores autocommit afterward. Caller-managed transactions are left alone.

That sequence is more than client polish. A database assistant that proposes or explains SQL needs the same diagnostic evidence a human operator would inspect. Returning only “executed successfully” hides whether Oracle chose an index scan, a full table scan or another expensive path. Returning the actual plan makes the optimization step reviewable before anyone treats an AI-generated query as production-ready.

The merged change also avoids wrapping an existing EXPLAIN PLAN inside a second explain command, which previously could fail with ORA-00905. Custom INTO <table> statements retain Oracle’s default path rather than being intercepted.

What is—and is not—proved

The contributor reports 12 parser tests, 11 command-executor tests and 73 tests across the Oracle module, plus manual checks against Oracle AI Database 26ai Free. Those checks produced different visible plans for indexed and non-indexed predicates. The evidence is maintainer-reported, not an independent benchmark, and the release remains a beta.

Even so, the change defines a useful boundary for AI-assisted SQL tooling: generating a query is only the first half of the workflow. The client must also surface the database’s own execution evidence, preserve transaction semantics and fail clearly when the diagnostic substrate is unavailable.

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