@duckcodeailabs/dql-cli 1.12.9 → 1.12.10

package/dist/local-runtime.js

@@ -21,7 +21,7 @@ import { fetchLatestPublishedDqlVersion, resolveDqlRuntimeVersionStatus } from '
 import { resolveRetrievalHealthStatus } from './retrieval-health.js';
 import { createDqlAgentProviderRunner, resolveAgentFollowUpContext } from './llm/providers/dql-agent-provider.js';
 import { listRemoteMcpSettings, saveRemoteMcpSettings } from './llm/mcp-config.js';
-import { ClaudeProvider, ConversationStore, advanceThreadState, buildConversationSnapshot, conversationHistoryFromContext, recallRelevantTurns, renderConversationEnvelopeForPrompt, GeminiProvider, MemoryStore, OllamaProvider, OpenAIProvider, buildBlockBusinessFingerprint, buildBlockSqlFingerprints, buildAnalysisQuestionPlan, composeSemanticQueryForQuestion, aggregationIntegrityIssuesForSql, buildLocalContextPack, applyContextPackCompatibility, toAgentRetrievalEvidence, prepareConversationPath, defaultMemoryPath, ensureDefaultMemoryFiles, ensureAgentProjectReady, isAgentProjectIndexReady, currentMetadataFingerprint, ensureMetadataCatalogFresh, readIndexedDomainKnowledge, readIndexedKnowledge360, compactSemanticRuntimeFailure, classifyAnalyticalFailure, propose, proposePlan, recordGovernedCorrection, HintStore, defaultHintIndexPath, ensureHintIndexFresh, listHintsFromGit, getHintEvaluationFromGit, getCorrectionTraceFromGit, inspectGovernedHint, editGovernedHintCandidate, reopenGovernedHint, retireHint, supersedeHint, hintsConflict, mineJoinPatterns, reviewGovernedHint, AgentRunEngine, SqliteAgentRunStore, defaultAgentRunGates, createLlmAgentRunPlanner, createHybridRouter, computeResultStats, buildDeterministicDashboardStory, synthesizeAnswer, streamOrGenerate, narrateResult, buildProposePreview, buildFromPrompt, internalRelationIdsInSql, defaultAgentRunStorePath, defaultAgentRunSqlitePath, resolveLocalOwner, resolveProposeConfig, recordQueryRun, recordRuntimeSchemaSnapshot, latestRuntimeSchemaSnapshotForProject, loadSkills, migrateLegacySkills, configuredSkillsPath, skillsDir, draftDomainSkillBootstrap, buildDomainSkillBootstrapPrompt, mergeDomainSkillBootstrapEnrichment, writeSkill, deleteSkill, deriveGeneratedDraftSlug, deriveAnalyticalRepair, reindexProject, invalidateAgentProjectState, recordAgentRuntimeVersion, resolveDomainContextEnvelope, projectEmbeddingProvider, isHashedEmbeddingProvider, clearProjectEmbeddingCache, upgradeVectorIndexForProject, openMetadataCatalog, defaultKgPath, planAppFromPrompt, KGStore, planResearch, loadSemanticMetrics, cascadeTraceToEvidenceRouteSteps, createCascadeAnswerResult, createCascadeTrace, routeReasoningEffort, routeForCascadeAnswerTier, clampReasoningEffort, bumpReasoningEffort, resolveThinkingMode, coerceThinkingMode, upsertGeneratedDqlArtifactDraft, loadAgentSemanticLayer, isTrustedConversationTurn, resolveInternalRelationIds, analyticalError, withAnalyticalErrorOrigin, withAnalyticalErrorOriginSync, } from '@duckcodeailabs/dql-agent';
+import { ClaudeProvider, ConversationStore, advanceThreadState, buildConversationSnapshot, conversationHistoryFromContext, recallRelevantTurns, renderConversationEnvelopeForPrompt, GeminiProvider, MemoryStore, OllamaProvider, OpenAIProvider, buildBlockBusinessFingerprint, buildBlockSqlFingerprints, buildAnalysisQuestionPlan, composeSemanticQueryForQuestion, aggregationIntegrityIssuesForSql, buildLocalContextPack, applyContextPackCompatibility, toAgentRetrievalEvidence, prepareConversationPath, defaultMemoryPath, ensureDefaultMemoryFiles, ensureAgentProjectReady, isAgentProjectIndexReady, currentMetadataFingerprint, ensureMetadataCatalogFresh, readIndexedDomainKnowledge, readIndexedKnowledge360, compactSemanticRuntimeFailure, classifyAnalyticalFailure, normalizeWarehouseSqlFailure, parseProposal, propose, proposePlan, recordGovernedCorrection, HintStore, defaultHintIndexPath, ensureHintIndexFresh, listHintsFromGit, getHintEvaluationFromGit, getCorrectionTraceFromGit, inspectGovernedHint, editGovernedHintCandidate, reopenGovernedHint, retireHint, supersedeHint, hintsConflict, mineJoinPatterns, reviewGovernedHint, AgentRunEngine, SqliteAgentRunStore, defaultAgentRunGates, createLlmAgentRunPlanner, createHybridRouter, computeResultStats, buildDeterministicDashboardStory, synthesizeAnswer, streamOrGenerate, narrateResult, buildProposePreview, buildFromPrompt, internalRelationIdsInSql, defaultAgentRunStorePath, defaultAgentRunSqlitePath, resolveLocalOwner, resolveProposeConfig, recordQueryRun, recordRuntimeSchemaSnapshot, latestRuntimeSchemaSnapshotForProject, loadSkills, migrateLegacySkills, configuredSkillsPath, skillsDir, draftDomainSkillBootstrap, buildDomainSkillBootstrapPrompt, mergeDomainSkillBootstrapEnrichment, writeSkill, deleteSkill, deriveGeneratedDraftSlug, deriveAnalyticalRepair, reindexProject, invalidateAgentProjectState, recordAgentRuntimeVersion, resolveDomainContextEnvelope, projectEmbeddingProvider, isHashedEmbeddingProvider, clearProjectEmbeddingCache, upgradeVectorIndexForProject, openMetadataCatalog, defaultKgPath, planAppFromPrompt, KGStore, planResearch, loadSemanticMetrics, cascadeTraceToEvidenceRouteSteps, createCascadeAnswerResult, createCascadeTrace, routeReasoningEffort, routeForCascadeAnswerTier, clampReasoningEffort, bumpReasoningEffort, resolveThinkingMode, coerceThinkingMode, upsertGeneratedDqlArtifactDraft, loadAgentSemanticLayer, isTrustedConversationTurn, resolveInternalRelationIds, analyticalError, tagAnalyticalError, withAnalyticalErrorOrigin, withAnalyticalErrorOriginSync, } from '@duckcodeailabs/dql-agent';
 import { addSqlResultFilter, filterableResultColumns, replaceBlockStudioSql } from './sql-result-filter.js';
 import { gatherProposeEnrichment } from './propose-enrich.js';
 import { handleAppsApi, proposeAppAiBuild, recommendVisualization } from './apps-api.js';
@@ -403,6 +403,36 @@ export function analyticalFailedRunFromAgentRun(run) {
     }
     return undefined;
 }
+/** Exact failed statement retained on an immutable run, if one exists. */
+export function repairableSqlFromAgentRun(run) {
+    for (const artifact of run.artifacts) {
+        const payload = agentRunRecord(artifact.payload);
+        if (!payload)
+            continue;
+        const dqlArtifact = normalizeDqlArtifactReference(payload.dqlArtifact);
+        const sql = agentRunString(payload.proposedSql)
+            ?? agentRunString(payload.sql)
+            ?? dqlArtifact?.compiledSql;
+        if (sql?.trim())
+            return sql.trim();
+    }
+    return undefined;
+}
+/** Structured producer failure retained on the source run. */
+export function warehouseFailureFromAgentRun(run) {
+    for (const artifact of run.artifacts) {
+        const payload = agentRunRecord(artifact.payload);
+        const failure = agentRunRecord(payload?.warehouseFailure);
+        if (failure?.version !== 1 || typeof failure.redactedMessage !== 'string')
+            continue;
+        return failure;
+    }
+    return undefined;
+}
+/** Permission, authentication, policy and unsafe-query failures are never AI-repaired. */
+export function analyticalFailureAllowsDeterministicRetry(failure) {
+    return !failure || !['permission', 'authentication', 'unsafe', 'cancelled'].includes(failure.category);
+}
 function businessNarrativeGaps(warnings) {
     const businessRelevant = (warnings ?? []).filter((warning) => !(/missing requested output column/i.test(warning)
         || /\bquery plan\b/i.test(warning)
@@ -1996,6 +2026,7 @@ export async function startLocalServer(opts) {
                     sql: exploration.sql,
                     exploratoryCandidate: repairedExploratoryCandidate,
                     executionError: exploration.error,
+                    warehouseFailure: exploration.warehouseFailure,
                     text: explorationAnswer,
                     answer: explorationAnswer,
                     trustLabel: 'Exploratory · DBT-grounded',
@@ -3242,10 +3273,21 @@ export async function startLocalServer(opts) {
             const semanticError = semanticCompose?.diagnostics.find((diagnostic) => diagnostic.severity === 'error')?.message;
             throw analyticalError(semanticError ?? `DQL artifact "${metadata.name ?? 'draft'}" produced no executable SQL.`, { origin: 'dql_compilation', stage: 'compile' });
         }
-        const prepared = prepareLocalExecution(semanticCompose?.sql ?? plan.sql, activeConnection, projectRoot, projectConfig);
-        const executionSql = invocationInput?.rowLimit
-            ? buildAgentPreviewSql(prepared.sql, invocationInput.rowLimit)
-            : prepared.sql;
+        const preparation = await prepareAnalyticalExecutionSql({
+            sql: semanticCompose?.sql ?? plan.sql,
+            subject: 'DQL artifact query',
+            executor,
+            connection: activeConnection,
+            projectRoot,
+            projectConfig,
+            // Preserve the existing contract: previewed Ask artifacts are explicitly
+            // read-only and bounded; saved/manual DQL execution keeps its compiler-
+            // governed behavior when no preview bound was requested.
+            enforceReadOnly: invocationInput?.rowLimit !== undefined,
+            rowLimit: invocationInput?.rowLimit,
+        });
+        const prepared = { sql: preparation.preparedSql, connection: preparation.connection };
+        const executionSql = preparation.executedSql;
         const app = loadRuntimeApp(projectRoot, activePersonaAppId());
         const sourceDomain = metadata.domain ?? source.match(/\bdomain\s*=\s*"([^"]+)"/i)?.[1];
         assertAppAccess({ app, domain: sourceDomain ?? app?.domain, level: 'execute' });
@@ -3454,13 +3496,18 @@ export async function startLocalServer(opts) {
         if (semantic.unresolvedRefs.length > 0) {
             throw analyticalError(`Unknown semantic reference${semantic.unresolvedRefs.length > 1 ? 's' : ''}: ${semantic.unresolvedRefs.join(', ')}`, { origin: 'dql_compilation', stage: 'compile', code: 'semantic_ref' });
         }
-        const prepared = prepareLocalExecution(semantic.sql, activeConnection, projectRoot, projectConfig);
-        // Read-only enforcement runs on the statement that will ACTUALLY execute.
-        const readOnlyError = readOnlySqlValidationError(prepared.sql, 'Generated SQL', activeConnection.driver);
-        if (readOnlyError) {
-            throw analyticalError(readOnlyError, { origin: 'governance_gate', stage: 'validation', code: 'unsafe_sql' });
-        }
-        const bounded = buildRowBoundedSql(prepared.sql, rowBound, activeConnection.driver);
+        const preparation = await prepareAnalyticalExecutionSql({
+            sql: semantic.sql,
+            subject: 'Generated SQL',
+            executor,
+            connection: activeConnection,
+            projectRoot,
+            projectConfig,
+            enforceReadOnly: true,
+            rowLimit: rowBound,
+        });
+        const prepared = { sql: preparation.preparedSql, connection: preparation.connection };
+        const bounded = preparation.rowBound;
         const app = loadRuntimeApp(projectRoot, activePersonaAppId());
         // Generated SQL is governed as 'uncategorized' unless its seed declares a
         // domain — the same value `generatedSqlArtifactContract` stamps into the
@@ -3488,7 +3535,7 @@ export async function startLocalServer(opts) {
             })
             : null;
         const rawResult = semanticExecution?.result
-            ?? await executor.executeQuery(bounded.sql, [], runtimeVariables({}), prepared.connection);
+            ?? await executor.executeQuery(preparation.executedSql, [], runtimeVariables({}), preparation.connection);
         const normalized = normalizeQueryResult(rawResult, semantic.semanticRefs);
         // The authoritative bound. `buildRowBoundedSql` may legitimately decline to
         // touch the statement, and only one of 15 drivers honours a maxRows hint, so
@@ -3528,34 +3575,65 @@ export async function startLocalServer(opts) {
      */
     const executeExploratoryCandidate = async (candidate, schemaContext = [], question = '', requestedTopN, executionConnection) => {
         const activeConnection = requireActiveConnection(executionConnection);
-        const preflight = repairExploratorySqlBeforeExecution(candidate.sql, schemaContext, question, activeConnection.driver);
+        const failed = (message, sql, repairs, proofs = [], code = 'exploratory_validation_failed') => ({
+            proofs,
+            sql,
+            repairs,
+            error: message,
+            warehouseFailure: normalizeWarehouseSqlFailure(analyticalError(message, {
+                origin: 'governance_gate',
+                stage: 'validation',
+                code,
+            }), activeConnection.driver),
+        });
+        let normalizedCandidateSql;
+        try {
+            normalizedCandidateSql = (await prepareAnalyticalExecutionSql({
+                sql: candidate.sql,
+                subject: question || 'this exploratory query',
+                executor,
+                connection: activeConnection,
+                projectRoot,
+                projectConfig,
+                enforceReadOnly: true,
+            })).preparedSql;
+        }
+        catch (error) {
+            const failure = normalizeWarehouseSqlFailure(error, activeConnection.driver);
+            return {
+                proofs: [],
+                sql: candidate.sql,
+                repairs: [],
+                error: failure.redactedMessage,
+                warehouseFailure: failure,
+            };
+        }
+        const preflight = repairExploratorySqlBeforeExecution(normalizedCandidateSql, schemaContext, question, activeConnection.driver);
         const boundedSql = applyRequestedTopNToExploratorySql(preflight.sql, requestedTopN);
         const repairs = boundedSql === preflight.sql
             ? [...preflight.repairs]
             : [...preflight.repairs, `Applied the requested overall top-${requestedTopN} bound before exploratory execution.`];
         if (preflight.blockedReason) {
-            return {
-                proofs: [],
-                sql: boundedSql,
-                repairs,
-                error: preflight.blockedReason,
-            };
+            return failed(preflight.blockedReason, boundedSql, repairs, [], 'exploratory_preflight_blocked');
         }
         const analysis = analyzeSqlReferences(boundedSql, activeConnection.driver);
+        if (!analysis.parsed && sqlMayContainJoin(boundedSql)) {
+            return failed('DQL could not parse the exploratory SQL to validate its join predicates. Nothing was sent to the warehouse.', boundedSql, repairs, [], 'exploratory_join_parse_failed');
+        }
         if (!analysis.parsed) {
-            return { proofs: [], sql: boundedSql, repairs, error: 'DQL could not parse the exploratory SQL to validate its join predicates.' };
+            repairs.push('The local DQL parser could not inspect this join-free statement; warehouse read-only execution will provide the authoritative syntax check.');
         }
-        const probeableJoins = probeableExploratoryJoins(analysis.joins, analysis.ctes, analysis.derivedRelations);
+        const probeableJoins = probeableExploratoryJoins(analysis.parsed ? analysis.joins : [], analysis.parsed ? analysis.ctes : [], analysis.parsed ? analysis.derivedRelations : []);
         // A normal customer → order → order-item → product question needs three
         // joins. Keep the lane bounded, but do not reject this common dbt-star path
         // solely because the old two-join demo limit was too small. CTE-internal
         // joins are counted physically (they appear in analysis.joins with their
         // physical endpoints); joins ON a CTE are restructuring, not new paths.
         if (probeableJoins.length > 4) {
-            return { proofs: [], sql: boundedSql, repairs, error: 'This exploratory query has more than four joins and requires an analyst to review the relationship path.' };
+            return failed('This exploratory query has more than four joins and requires an analyst to review the relationship path.', boundedSql, repairs, [], 'exploratory_join_limit');
         }
         if (probeableJoins.some((join) => !join.leftRelation || !join.rightRelation)) {
-            return { proofs: [], sql: boundedSql, repairs, error: 'This exploratory query has a join endpoint DQL could not resolve for bounded validation.' };
+            return failed('This exploratory query has a join endpoint DQL could not resolve for bounded validation.', boundedSql, repairs, [], 'exploratory_join_unresolved');
         }
         // Structural binding: when the candidate carries a declared draft join path,
         // every SQL join must ride one of its edges on the declared key pair. A join
@@ -3568,22 +3646,12 @@ export async function startLocalServer(opts) {
                 const edge = declaredEdges.find((value) => declaredExploratoryRelationMatches(join.leftRelation, value.fromRelation) && declaredExploratoryRelationMatches(join.rightRelation, value.toRelation)
                     || declaredExploratoryRelationMatches(join.leftRelation, value.toRelation) && declaredExploratoryRelationMatches(join.rightRelation, value.fromRelation));
                 if (!edge) {
-                    return {
-                        proofs: [],
-                        sql: boundedSql,
-                        repairs,
-                        error: `The join between ${join.leftRelation} and ${join.rightRelation} is not part of the declared relationship path, so DQL did not execute it. Declare the relationship in the DQL model to enable it.`,
-                    };
+                    return failed(`The join between ${join.leftRelation} and ${join.rightRelation} is not part of the declared relationship path, so DQL did not execute it. Declare the relationship in the DQL model to enable it.`, boundedSql, repairs, [], 'undeclared_relationship');
                 }
                 const keyMatches = edge.keys.some((key) => exploratoryColumnsMatch(join.leftColumn, key.from) && exploratoryColumnsMatch(join.rightColumn, key.to)
                     || exploratoryColumnsMatch(join.leftColumn, key.to) && exploratoryColumnsMatch(join.rightColumn, key.from));
                 if (!keyMatches) {
-                    return {
-                        proofs: [],
-                        sql: boundedSql,
-                        repairs,
-                        error: `The join ${join.leftColumn} = ${join.rightColumn} uses different keys than relationship "${edge.relationshipId}" declares (${edge.keys.map((key) => `${key.from}=${key.to}`).join(', ')}). Review the declared keys before executing.`,
-                    };
+                    return failed(`The join ${join.leftColumn} = ${join.rightColumn} uses different keys than relationship "${edge.relationshipId}" declares (${edge.keys.map((key) => `${key.from}=${key.to}`).join(', ')}). Review the declared keys before executing.`, boundedSql, repairs, [], 'relationship_key_mismatch');
                 }
                 joinEdges.set(index, edge);
             }
@@ -3594,7 +3662,7 @@ export async function startLocalServer(opts) {
                 }
             }
         }
-        if (probeableJoins.length < analysis.joins.length) {
+        if (analysis.parsed && probeableJoins.length < analysis.joins.length) {
             repairs.push(`Skipped join probes for ${analysis.joins.length - probeableJoins.length} join(s) on derived (CTE) relations; the final query execution validates them.`);
         }
         const proofs = [];
@@ -3615,14 +3683,21 @@ export async function startLocalServer(opts) {
                 // answer — these gates only inspect the raw key samples.
                 const contradiction = exploratoryProbeContradiction(raw.rows, join, joinEdges.get(index));
                 if (contradiction) {
-                    return { proofs, sql: boundedSql, repairs, error: contradiction };
+                    return failed(contradiction, boundedSql, repairs, proofs, 'relationship_probe_contradiction');
                 }
             }
             const result = await executeGeneratedArtifactForAgent(question, boundedSql, undefined, activeConnection);
             return { result, proofs, sql: boundedSql, repairs };
         }
         catch (error) {
-            return { proofs, sql: boundedSql, repairs, error: error instanceof Error ? error.message : String(error) };
+            const failure = normalizeWarehouseSqlFailure(error, activeConnection.driver);
+            return {
+                proofs,
+                sql: boundedSql,
+                repairs,
+                error: failure.redactedMessage,
+                warehouseFailure: failure,
+            };
         }
     };
     /**
@@ -4784,9 +4859,10 @@ export async function startLocalServer(opts) {
      */
     const MIN_COMPRESS_BYTES = 1024;
     const COMPRESSIBLE = /^(?:text\/(?!event-stream)|application\/(?:json|javascript|xml|wasm)|image\/svg)/i;
-    const withCompression = (request, response) => {
+    const withCompression = (request, rawResponse) => {
+        const response = rawResponse;
         if (!/\bgzip\b/i.test(String(request.headers['accept-encoding'] ?? '')))
-            return response;
+            return rawResponse;
         const writeHead = response.writeHead.bind(response);
         const end = response.end.bind(response);
         const write = response.write.bind(response);
@@ -4857,7 +4933,7 @@ export async function startLocalServer(opts) {
             });
             return end(gz);
         });
-        return response;
+        return rawResponse;
     };
     const server = createServer((req, res) => {
         res = withCompression(req, res);
@@ -5710,7 +5786,7 @@ export async function startLocalServer(opts) {
                     if (!body.expectedSnapshotId || body.expectedSnapshotId !== snapshot.snapshotId)
                         throw Object.assign(new Error('Project sources changed before relationship validation.'), { code: 'SOURCE_CHANGED' });
                     const manifest = snapshot.manifest;
-                    const activeConnection = requireActiveConnection();
+                    const activeConnection = await resolveExecutionConnection(body);
                     const evidence = await validateModelingRelationship(body.relationship, manifest, (sql) => executor.executeQuery(sql, [], {}, activeConnection), (identifier) => getDialect(activeConnection.driver).quoteIdentifier(identifier));
                     res.writeHead(200, { 'Content-Type': 'application/json; charset=utf-8' });
                     res.end(serializeJSON({ requestId, snapshotId: snapshot.snapshotId, evidence }));
@@ -5735,6 +5811,267 @@ export async function startLocalServer(opts) {
                 res.end(serializeJSON({ runs, total: agentRunStore.list().length, limit }));
                 return;
             }
+            /**
+             * One-click bounded execution repair.
+             *
+             * The source run is immutable. First, retry its exact SQL through the
+             * shared deterministic preparation boundary (this repairs internal IDs and
+             * dialect-safe row bounds without an LLM). Only a connector-classified
+             * model-repair failure may spend one provider call to rewrite the SQL. The
+             * derived SQL is still read-only, bounded, and executed on the source run's
+             * exact target before a replacement card is returned.
+             *
+             * Acceptance: API-003, API-007, UI-012, UI-013, SEC-004, E2E-014.
+             */
+            if (req.method === 'POST' && /^\/api\/agent-runs\/[^/]+\/repair-execution$/.test(path)) {
+                const match = path.match(/^\/api\/agent-runs\/([^/]+)\/repair-execution$/);
+                const sourceRunId = decodeURIComponent(match?.[1] ?? '');
+                const sourceRun = await agentRunStore.get(sourceRunId);
+                if (!sourceRun) {
+                    res.writeHead(404, { 'Content-Type': 'application/json; charset=utf-8' });
+                    res.end(serializeJSON({ error: 'Agent run not found.' }));
+                    return;
+                }
+                const derivedRunId = `${sourceRun.id}:repair:1`;
+                const existing = await agentRunStore.get(derivedRunId);
+                if (existing) {
+                    res.writeHead(200, { 'Content-Type': 'application/json; charset=utf-8' });
+                    res.end(serializeJSON({ run: existing }));
+                    return;
+                }
+                const sourceSql = repairableSqlFromAgentRun(sourceRun);
+                const retainedFailure = warehouseFailureFromAgentRun(sourceRun);
+                if (!sourceSql || !analyticalFailureAllowsDeterministicRetry(retainedFailure)) {
+                    res.writeHead(409, { 'Content-Type': 'application/json; charset=utf-8' });
+                    res.end(serializeJSON({
+                        error: sourceSql
+                            ? 'This failure requires authorized access or a policy change and cannot be repaired by AI.'
+                            : 'This run has no retained SQL statement to repair.',
+                        ...(retainedFailure ? { warehouseFailure: retainedFailure } : {}),
+                    }));
+                    return;
+                }
+                const body = agentRunRecord(await readJSON(req).catch(() => null));
+                const threadId = agentRunString(body?.threadId);
+                let targetConnection;
+                try {
+                    targetConnection = await resolveAgentRunExecutionConnection({
+                        question: sourceRun.question,
+                        executionTarget: sourceRun.executionTarget,
+                    });
+                }
+                catch (error) {
+                    res.writeHead(409, { 'Content-Type': 'application/json; charset=utf-8' });
+                    res.end(serializeJSON({ error: apiErrorMessage(error) }));
+                    return;
+                }
+                const targetConnectionName = sourceRun.executionTarget?.target === 'connection'
+                    ? sourceRun.executionTarget.connectionName
+                    : sourceRun.executionTarget?.target === 'local'
+                        ? 'local'
+                        : projectConfig.defaultConnectionName;
+                let result;
+                let repairedSql = sourceSql;
+                let repairMode = 'deterministic';
+                let firstFailure;
+                try {
+                    result = await executeGeneratedSqlDirect(sourceRun.question, sourceSql, undefined, targetConnection);
+                }
+                catch (error) {
+                    firstFailure = normalizeWarehouseSqlFailure(error, targetConnection.driver);
+                }
+                if (!result) {
+                    if (firstFailure?.retryDisposition !== 'model_repair') {
+                        res.writeHead(422, { 'Content-Type': 'application/json; charset=utf-8' });
+                        res.end(serializeJSON({
+                            error: firstFailure?.redactedMessage ?? 'The deterministic repair did not produce an executable query.',
+                            ...(firstFailure ? { warehouseFailure: firstFailure } : {}),
+                        }));
+                        return;
+                    }
+                    const provider = await createBlockStudioAssistProvider(projectRoot);
+                    if (!provider) {
+                        res.writeHead(409, { 'Content-Type': 'application/json; charset=utf-8' });
+                        res.end(serializeJSON({ error: 'Configure an available AI provider in Settings to repair this query.' }));
+                        return;
+                    }
+                    let schemaContext;
+                    try {
+                        schemaContext = await getSchemaContextForAgent(sourceRun.question, undefined, targetConnection, targetConnectionName);
+                    }
+                    catch (error) {
+                        const failure = normalizeWarehouseSqlFailure(error, targetConnection.driver);
+                        res.writeHead(422, { 'Content-Type': 'application/json; charset=utf-8' });
+                        res.end(serializeJSON({ error: failure.redactedMessage, warehouseFailure: failure }));
+                        return;
+                    }
+                    const boundedSchema = schemaContext.slice(0, 16).map((table) => ({
+                        relation: table.relation,
+                        columns: table.columns.slice(0, 40).map((column) => ({ name: column.name, type: column.type })),
+                    }));
+                    let raw;
+                    try {
+                        raw = await provider.generate([
+                            {
+                                role: 'system',
+                                content: [
+                                    'Repair one failed read-only analytical SQL statement.',
+                                    'Return exactly one JSON object in a ```json fence with keys summary, sql, viz, outputs.',
+                                    'Preserve the user question, selected grain, filters, and aggregation semantics.',
+                                    'Use only relations and columns present in the supplied schema context.',
+                                    'Never emit internal DQL relation identities such as source::, model::, seed::, or snapshot::.',
+                                    'Do not add writes, DDL, multiple statements, or commentary outside the JSON fence.',
+                                ].join(' '),
+                            },
+                            {
+                                role: 'user',
+                                content: JSON.stringify({
+                                    question: sourceRun.question,
+                                    driver: targetConnection.driver,
+                                    failure: firstFailure.redactedMessage,
+                                    sql: sourceSql,
+                                    schema: boundedSchema,
+                                }),
+                            },
+                        ], { maxTokens: 1200, temperature: 0 });
+                    }
+                    catch {
+                        res.writeHead(502, { 'Content-Type': 'application/json; charset=utf-8' });
+                        res.end(serializeJSON({ error: 'The AI provider could not complete this repair. Check Provider Settings and retry.' }));
+                        return;
+                    }
+                    const proposal = parseProposal(raw);
+                    if (!proposal.sql?.trim()) {
+                        res.writeHead(422, { 'Content-Type': 'application/json; charset=utf-8' });
+                        res.end(serializeJSON({ error: 'The AI repair did not return an executable SQL statement.' }));
+                        return;
+                    }
+                    repairedSql = proposal.sql.trim();
+                    repairMode = 'ai';
+                    try {
+                        result = await executeGeneratedSqlDirect(sourceRun.question, repairedSql, undefined, targetConnection);
+                    }
+                    catch (error) {
+                        const failure = normalizeWarehouseSqlFailure(error, targetConnection.driver);
+                        res.writeHead(422, { 'Content-Type': 'application/json; charset=utf-8' });
+                        res.end(serializeJSON({
+                            error: failure.redactedMessage,
+                            warehouseFailure: failure,
+                        }));
+                        return;
+                    }
+                }
+                const completedAt = new Date().toISOString();
+                const preview = agentResultToSynthesisPreview(result);
+                const synthesis = await synthesizeAnswer({
+                    question: sourceRun.question,
+                    audience: 'stakeholder',
+                    resultPreview: preview,
+                    sql: repairedSql,
+                    draftText: `The repaired query returned ${result.rowCount} row${result.rowCount === 1 ? '' : 's'}.`,
+                    ...(preview ? { columnFormats: agentColumnDisplayFormats(projectRoot, preview.columns) } : {}),
+                });
+                const sourceFailure = analyticalFailedRunFromAgentRun(sourceRun)?.failure;
+                const artifact = agentRunArtifact('answer', 'Repaired analytical answer', {
+                    kind: 'uncertified',
+                    certification: 'analyst_review_required',
+                    reviewStatus: 'analyst_review_required',
+                    trustLabel: 'AI-generated · review required',
+                    answer: synthesis.text,
+                    text: synthesis.text,
+                    result,
+                    proposedSql: repairedSql,
+                    sql: result.sql ?? repairedSql,
+                    dqlArtifact: result.dqlArtifact,
+                    previousAttempt: {
+                        sourceRunId: sourceRun.id,
+                        sql: sourceSql,
+                        failure: retainedFailure ?? firstFailure,
+                    },
+                    repair: { attempt: 1, mode: repairMode },
+                }, undefined, 'review_required');
+                const evaluation = {
+                    id: `${derivedRunId}:execution`,
+                    label: 'Repaired query execution',
+                    passed: true,
+                    severity: 'info',
+                    message: `The ${repairMode === 'ai' ? 'AI-repaired' : 'deterministically normalized'} read-only query executed on the original target.`,
+                };
+                const repairStep = {
+                    id: `${derivedRunId}:step`,
+                    index: 0,
+                    route: 'generated_answer',
+                    resolvedRoute: 'generated_answer',
+                    goal: 'Repair and execute the failed analytical query on its original target.',
+                    successCriteria: ['Read-only validation passes.', 'The original execution target returns a bounded result.'],
+                    status: 'repaired',
+                    attempts: 1,
+                    summary: evaluation.message,
+                    evaluations: [evaluation],
+                    artifacts: [artifact],
+                };
+                const events = [
+                    { id: `${derivedRunId}:repair`, runId: derivedRunId, type: 'repair.attempted', at: completedAt, message: 'Started one bounded query repair.', route: 'generated_answer' },
+                    { id: `${derivedRunId}:artifact`, runId: derivedRunId, type: 'artifact.created', at: completedAt, message: 'Created the repaired result artifact.', route: 'generated_answer', trustState: 'review_required' },
+                    { id: `${derivedRunId}:complete`, runId: derivedRunId, type: 'run.completed', at: completedAt, message: 'The repaired query executed successfully.', route: 'generated_answer', status: 'needs_review', trustState: 'review_required' },
+                ];
+                const derivedRun = {
+                    id: derivedRunId,
+                    question: sourceRun.question,
+                    requestedMode: sourceRun.requestedMode,
+                    route: 'generated_answer',
+                    status: 'needs_review',
+                    trustState: 'review_required',
+                    stopReason: 'generated_review_required',
+                    startedAt: completedAt,
+                    completedAt,
+                    selectedObject: sourceRun.selectedObject,
+                    executionTarget: sourceRun.executionTarget,
+                    routeDecision: sourceRun.routeDecision,
+                    plan: sourceRun.plan,
+                    steps: [repairStep],
+                    summary: synthesis.text,
+                    answer: synthesis.text,
+                    answerKind: 'governed',
+                    artifacts: [artifact],
+                    evaluations: [evaluation],
+                    events,
+                    nextActions: sourceRun.nextActions,
+                    repairAttempts: 1,
+                    lifecycle: {
+                        version: 1,
+                        state: 'terminal',
+                        phase: 'run.completed',
+                        revision: 1,
+                        eventCursor: events.length,
+                        startedAt: completedAt,
+                        updatedAt: completedAt,
+                        completedAt,
+                    },
+                    diagnosticReceipt: {
+                        version: 1,
+                        runId: derivedRunId,
+                        phase: 'run.completed',
+                        route: 'generated_answer',
+                        plan: sourceRun.plan,
+                        steps: [repairStep],
+                        artifacts: [artifact],
+                        evaluations: [evaluation],
+                    },
+                    derivation: {
+                        version: 1,
+                        kind: 'analytical_repair',
+                        sourceRunId: sourceRun.id,
+                        ...(sourceFailure?.failureId ? { sourceFailureId: sourceFailure.failureId } : {}),
+                        attempt: 1,
+                    },
+                };
+                await agentRunStore.save(derivedRun);
+                recordConversationTurn(threadId ? getConversationStore() : null, threadId, derivedRun);
+                res.writeHead(201, { 'Content-Type': 'application/json; charset=utf-8' });
+                res.end(serializeJSON({ run: derivedRun }));
+                return;
+            }
             if (req.method === 'POST' && /^\/api\/agent-runs\/[^/]+\/analytical-repair$/.test(path)) {
                 const match = path.match(/^\/api\/agent-runs\/([^/]+)\/analytical-repair$/);
                 const id = decodeURIComponent(match?.[1] ?? '');
@@ -12005,7 +12342,15 @@ export async function startLocalServer(opts) {
                         }));
                         return;
                     }
-                    const prepared = prepareLocalExecution(semantic.sql, executionConnection, projectRoot, projectConfig);
+                    const preparation = await prepareAnalyticalExecutionSql({
+                        sql: semantic.sql,
+                        subject: 'Notebook query',
+                        executor,
+                        connection: executionConnection,
+                        projectRoot,
+                        projectConfig,
+                    });
+                    const prepared = { sql: preparation.preparedSql, connection: preparation.connection };
                     const app = loadRuntimeApp(projectRoot, typeof body.appId === 'string' ? body.appId : activePersonaAppId());
                     const domain = typeof body.domain === 'string' ? body.domain : app?.domain;
                     assertAppAccess({ app, domain, level: 'execute' });
@@ -12033,8 +12378,8 @@ export async function startLocalServer(opts) {
                         ...normalized,
                         executionContext: notebookExecutionIdentity(execContext),
                         executionTarget: executionTargetDescriptor(body),
-                        compiledSql: semantic.sql,
-                        executedSql: prepared.sql,
+                        compiledSql: preparation.decodedSql,
+                        executedSql: preparation.executedSql,
                     });
                     res.writeHead(200, { 'Content-Type': 'application/json; charset=utf-8' });
                     res.end(payload);
@@ -12705,6 +13050,11 @@ export async function startLocalServer(opts) {
             if (req.method === 'POST' && path === '/api/dql/artifacts/execute') {
                 try {
                     const body = await readJSON(req);
+                    const targetConnection = await resolveExecutionConnection(body);
+                    const targetDescriptor = executionTargetDescriptor(body);
+                    const targetConnectionName = targetDescriptor.target === 'connection'
+                        ? targetDescriptor.connectionName
+                        : 'local';
                     const requestedArtifact = normalizeDqlArtifactReference(body.artifact);
                     const requestedBlockName = typeof body.blockName === 'string' ? body.blockName.trim() : '';
                     let source = requestedArtifact?.source ?? '';
@@ -12762,7 +13112,7 @@ export async function startLocalServer(opts) {
                         parameters,
                         question: typeof body.question === 'string' ? body.question : undefined,
                         rowLimit: requestedArtifact?.limit,
-                    });
+                    }, targetConnection, targetConnectionName);
                     const actualReceipt = result.executionReceipt;
                     // Same reasoning: if the canonical block was reloaded, its compiled SQL
                     // is EXPECTED to differ, so this determinism check would fire on a
@@ -12917,26 +13267,16 @@ export async function startLocalServer(opts) {
                         res.end(serializeJSON({ cellType: cell.type, result: null }));
                         return;
                     }
-                    // DECODE, THEN VALIDATE. A leaked `source::db.schema.table` is an internal
-                    // retrieval key whose suffix IS the physical relation, so rejecting the
-                    // cell and asking the user (or another AI round trip) to repair it was
-                    // busywork for a mechanical substitution. Anything still unresolvable
-                    // after this — a BARE `source::orders`, which names no database or
-                    // schema — genuinely cannot be decoded and still fails here.
-                    let decodedSql = resolveInternalRelationIds(executableSql, EMPTY_SCHEMA_GROUNDING);
-                    let bareResolution = null;
-                    if (internalDqlRelationIdValidationError(decodedSql, 'x')) {
-                        bareResolution = await resolveBareInternalRelationIds(decodedSql, executor, cellConnection);
-                        decodedSql = bareResolution.sql;
-                    }
-                    const internalRelationError = internalDqlRelationIdValidationError(decodedSql, cell.type === 'dql' ? 'DQL query' : 'Notebook query', 
-                    // A probe that came back empty is PROOF the relation is not visible
-                    // to this connection, so the refusal states that instead of sending
-                    // the user to an AI repair that cannot conjure a missing table.
-                    bareResolution ? { probed: true, ambiguous: bareResolution.ambiguous } : undefined);
-                    if (internalRelationError)
-                        throw new DqlInternalRelationIdError(internalRelationError);
-                    const prepared = prepareLocalExecution(decodedSql, cellConnection, projectRoot, projectConfig);
+                    const preparation = await prepareAnalyticalExecutionSql({
+                        sql: executableSql,
+                        subject: cell.type === 'dql' ? 'DQL query' : 'Notebook query',
+                        executor,
+                        connection: cellConnection,
+                        projectRoot,
+                        projectConfig,
+                    });
+                    const decodedSql = preparation.decodedSql;
+                    const prepared = { sql: preparation.preparedSql, connection: preparation.connection };
                     const app = loadRuntimeApp(projectRoot, typeof body.appId === 'string' ? body.appId : activePersonaAppId());
                     assertAppAccess({ app, domain: resolved.domain ?? app?.domain, level: 'execute' });
                     const semanticExecution = semanticCompose?.compiled
@@ -15281,6 +15621,60 @@ export function prepareLocalExecution(sql, connection, projectRoot, projectConfi
         connection: normalizedConnection,
     };
 }
+/**
+ * One host-owned SQL preparation boundary shared by Ask and Notebook.
+ *
+ * Surface-specific compilation happens before this function and surface policy
+ * (for example exploratory join proof) may run around it, but no analytical
+ * entrypoint may decode internal IDs, resolve dbt paths, enforce read-only
+ * execution, or append a preview bound differently anymore.
+ *
+ * Acceptance: API-003, API-006, API-007, EXP-001, E2E-014.
+ */
+export async function prepareAnalyticalExecutionSql(input) {
+    const sourceSql = input.sql.trim();
+    const rewrites = [];
+    let decodedSql = resolveInternalRelationIds(sourceSql, EMPTY_SCHEMA_GROUNDING, 'qualified', rewrites);
+    let bareResolution = null;
+    if (internalDqlRelationIdValidationError(decodedSql, 'x')) {
+        bareResolution = await resolveBareInternalRelationIds(decodedSql, input.executor, input.connection);
+        decodedSql = bareResolution.sql;
+        rewrites.push(...bareResolution.resolved);
+    }
+    const internalRelationError = internalDqlRelationIdValidationError(decodedSql, input.subject, bareResolution
+        ? { probed: true, ambiguous: bareResolution.ambiguous }
+        : undefined);
+    if (internalRelationError) {
+        throw tagAnalyticalError(new DqlInternalRelationIdError(internalRelationError), {
+            origin: bareResolution?.ambiguous.length ? 'ambiguity' : 'retrieval_gap',
+            stage: 'grounding',
+            code: bareResolution?.ambiguous.length ? 'ambiguous_relation' : 'unknown_relation',
+        });
+    }
+    const prepared = prepareLocalExecution(decodedSql, input.connection, input.projectRoot, input.projectConfig);
+    if (input.enforceReadOnly) {
+        const readOnlyError = readOnlySqlValidationError(prepared.sql, input.subject, input.connection.driver);
+        if (readOnlyError) {
+            throw analyticalError(readOnlyError, {
+                origin: 'governance_gate',
+                stage: 'validation',
+                code: 'unsafe_sql',
+            });
+        }
+    }
+    const rowBound = input.rowLimit === undefined
+        ? undefined
+        : buildRowBoundedSql(prepared.sql, input.rowLimit, input.connection.driver);
+    return {
+        sourceSql,
+        decodedSql,
+        preparedSql: prepared.sql,
+        executedSql: rowBound?.sql ?? prepared.sql,
+        connection: prepared.connection,
+        rewrites,
+        ...(rowBound ? { rowBound } : {}),
+    };
+}
 export function dashboardRuntimeVariables(dashboard, overrides = {}) {
     const variables = {};
     for (const param of dashboard.params ?? []) {
@@ -23332,6 +23726,35 @@ ${predicates.map((predicate) => `    ${predicate}`).join(' OR\n')}
 ORDER BY table_schema, table_name, ordinal_position
 LIMIT 400`;
 }
+/**
+ * Conservative fallback for deciding whether an unparsed statement still
+ * needs DQL's structural join validation. Join-free SELECT syntax may be
+ * newer than the local parser and can safely continue to the read-only
+ * warehouse boundary; an unparsed join must never bypass relationship proof.
+ */
+export function sqlMayContainJoin(sql) {
+    const normalized = stripSqlStringsAndComments(sql)
+        .replace(/\s+/g, ' ')
+        .trim();
+    if (/\bjoin\b/i.test(normalized))
+        return true;
+    const fromMatch = /\bfrom\b/i.exec(normalized);
+    if (!fromMatch)
+        return false;
+    const tail = normalized.slice(fromMatch.index + fromMatch[0].length);
+    const clauseMatch = /\b(where|group\s+by|having|qualify|order\s+by|limit|fetch|union|intersect|except)\b/i.exec(tail);
+    const fromClause = clauseMatch ? tail.slice(0, clauseMatch.index) : tail;
+    let depth = 0;
+    for (const char of fromClause) {
+        if (char === '(')
+            depth += 1;
+        else if (char === ')')
+            depth = Math.max(0, depth - 1);
+        else if (char === ',' && depth === 0)
+            return true;
+    }
+    return false;
+}
 /**
  * Resolve a BARE internal graph identity (`source::orders` — a name with no
  * database or schema) against the warehouse.