@awak-app/simy-cli 0.7.4 → 0.7.5

package/src/sqm/proof.js

@@ -42,7 +42,8 @@ export function buildProof({ organizationId, repository, bundle, minCheck, evalu
   const enforcingFindings = findings.filter((finding) => finding.enforcement === "blocking");
   const statuses = enforcingModules.flatMap((module) => module.scenarios.map((scenario) => scenario.status));
   const failedModules = enforcingModules.filter((module) => module.scenarios.some((scenario) => scenario.status === "failed"));
-  const hasFailure = enforcingFindings.length > 0 || failedModules.length > 0;
+  const localTests = evaluation.localTests;
+  const hasFailure = enforcingFindings.length > 0 || failedModules.length > 0 || Boolean(localTests && localTests.status !== "passed");
   const advisoryFindingCount = findings.length - enforcingFindings.length;
   const requiresHuman = statuses.includes("requires_human");
   const requiresPostDeploy = statuses.includes("requires_post_deploy");
@@ -68,14 +69,20 @@ export function buildProof({ organizationId, repository, bundle, minCheck, evalu
       status: minCheck.status,
       preliminary_finding_count: minCheck.preliminary_findings.length,
     },
-    executions: evaluation.modules,
+    executions: [...evaluation.modules, ...(localTests ? [{
+      kind: 'local_test_profile', source: 'cli_catalog_repository_tests',
+      module_id: `LOCAL-TESTS:${localTests.profile}`, module_version: String(localTests.version),
+      enforcement: 'blocking', rule_results: [], scenarios: [], state_models: [], stressors: [],
+      local_tests: localTests,
+    }] : [])],
     findings,
     outcome: {
       status,
       reasons: [
         ...(hasFailure ? ["one_or_more_blocking_checks_failed"] : []),
         ...(advisoryFindingCount > 0 ? ["advisory_findings_observed"] : []),
-        ...(evaluation.modules.length === 0 ? ["no_applicable_rules"] : []),
+        ...(evaluation.modules.length === 0 ? [localTests ? "no_applicable_cloud_rules" : "no_applicable_rules"] : []),
+        ...(localTests ? ["local_tests_executed"] : []),
         ...(bundle.modules?.length === 0 ? ["no_system_knowledge"] : []),
         ...(semanticResults.length ? ['semantic_assessments_advisory'] : []),
         ...(semanticResults.some(rule => rule.assessment.outcome !== 'scored') ? ['semantic_assessments_inconclusive'] : []),
@@ -141,3 +148,7 @@ export function verifySignedEvidence(proof, evidence) {
 function taggedDigest(value) {
   return `sha256:${createHash("sha256").update(canonicalize(value), "utf8").digest("hex")}`;
 }
+
+export function localTestsFromProof(proof) {
+  return proof?.executions?.find(row => row.kind === 'local_test_profile')?.local_tests || null;
+}

package/src/sqm/report.js

@@ -1,6 +1,9 @@
+import { localTestsFromProof } from './proof.js';
+
 // Presentation only: retain the signed proof and enforcement outcome unchanged.
 export function formatSqmCoverage(proof) {
-  const modules = proof.executions;
+  const modules = proof.executions.filter(row => row.kind !== 'local_test_profile');
+  const local = localTestsFromProof(proof);
   const rules = modules.flatMap(module => module.rule_results);
   const scenarios = modules.flatMap(module => module.scenarios);
   const blocking = modules.filter(module => module.enforcement === "blocking").length;
@@ -13,10 +16,11 @@ export function formatSqmCoverage(proof) {
     `SQM scope: ${modules.length} module(s) · ${rules.length} rule(s) · ${scenarios.length} scenario evaluation(s)`,
     `Enforcement: ${blocking} blocking module(s) · ${advisory} advisory module(s)`,
   ];
+  if (local) lines.push(`Local tests: ${local.executions.length}/${local.planned_suite_count ?? local.executions.length} suite(s) · ${local.executions.reduce((n, r) => n + (r.counts?.tests || 0), 0)} actual tests · ${local.status}. Cloud rules and repository test execution are separate.`);
   if (scenarios.length) lines.push(`Scenario evaluations: ${[...statusCounts].map(([status, count]) => `${status}=${count}`).join(" · ")}`);
   if (proof.outcome.reasons.length) lines.push(`Outcome reasons: ${proof.outcome.reasons.join(", ")}`);
   lines.push(...sqmCoverageWarnings({ moduleCount: modules.length, scenarioCount: scenarios.length,
-    noSystemKnowledge: proof.outcome.reasons.includes("no_system_knowledge"), outcomeStatus: proof.outcome.status }));
+    noSystemKnowledge: proof.outcome.reasons.includes("no_system_knowledge"), outcomeStatus: proof.outcome.status, localTests: local }));
   if (modules.length && !blocking) lines.push("Advisory-only scope: findings do not block this outcome.");
   const semantic = rules.filter(rule => rule.assessment).map(rule => rule.assessment);
   if (semantic.length) lines.push(`Semantic assessments: ${semantic.filter(result => result.executed).length} executed · ${semantic.filter(result => result.outcome === 'scored').length} scored · ${semantic.filter(result => result.outcome !== 'scored').length} inconclusive. Local model judgments are advisory and do not establish independent acceptance.`);
@@ -24,7 +28,8 @@ export function formatSqmCoverage(proof) {
   return lines.join("\n");
 }
 
-export function sqmCoverageWarnings({ moduleCount, scenarioCount, noSystemKnowledge, outcomeStatus = "passed" }) {
+export function sqmCoverageWarnings({ moduleCount, scenarioCount, noSystemKnowledge, outcomeStatus = "passed", localTests }) {
+  if (localTests && !moduleCount) return ["Warning: no applicable Cloud rules were evaluated. Local regression test results are reported separately; they do not establish deployed acceptance."];
   if (noSystemKnowledge) return [`Warning: the selected system/repository's downloaded SQM knowledge bundle has no published knowledge. ${outcomeStatus === "passed" ? "This is a pass with 0 evaluated cases" : "This result has 0 evaluated cases"}; it does not establish feature or historical incident coverage.`];
   if (!moduleCount) return ["Warning: no applicable rules matched this change (0 evaluated cases). This result does not establish feature or historical incident coverage; it does not mean the system has no past knowledge."];
   if (!scenarioCount) return ["Warning: No scenario evaluations: rule results alone do not establish scenario coverage."];

package/src/sqm/review-worker.js

@@ -442,6 +442,8 @@ async function runSingleProviderReview(
     subscriptionOnly = false,
     authStatusProbe,
     beforeAutomaticRetry = async () => {},
+    onUsage = () => {},
+    onContext = () => {},
   } = {},
 ) {
   assertProvider(provider);
@@ -508,7 +510,7 @@ async function runSingleProviderReview(
           ],
           prompt,
           { spawnImpl, executionTimeoutMs: remaining, stallTimeoutMs: provider === "codex" && readOnlyAnalysis && allowAutomaticRetry && !stallRetried && remaining > codexAttemptTimeoutMs ? codexAttemptTimeoutMs : null, label, cwd: workDir, environment, provider, signal, readOnlyAnalysis, model, retryAttempt,
-            terminationGraceMs: Number.isInteger(terminationGraceMs) ? Math.max(1, Math.min(terminationGraceMs, 5_000)) : 1_000 },
+            terminationGraceMs: Number.isInteger(terminationGraceMs) ? Math.max(1, Math.min(terminationGraceMs, 5_000)) : 1_000, onUsage, onContext },
         );
         selectedModel = model;
         break;
@@ -580,7 +582,7 @@ function codexAnalysisArguments(system) {
     '-c', `developer_instructions=${JSON.stringify(system)}`];
 }
 
-function spawnReview(command, args, prompt, { spawnImpl, executionTimeoutMs, stallTimeoutMs, cwd, environment, provider, signal, readOnlyAnalysis, terminationGraceMs, model, retryAttempt }) {
+function spawnReview(command, args, prompt, { spawnImpl, executionTimeoutMs, stallTimeoutMs, cwd, environment, provider, signal, readOnlyAnalysis, terminationGraceMs, model, retryAttempt, onUsage, onContext }) {
   return new Promise((resolve, reject) => {
     if (signal?.aborted) return reject(providerFailure("analysis_cancelled"));
     let child;
@@ -591,6 +593,7 @@ function spawnReview(command, args, prompt, { spawnImpl, executionTimeoutMs, sta
     } catch { reject(providerFailure("provider_spawn_failed", { model })); return; }
     let stdout = "", stderrPending = "", outputBytes = 0, phase = "before_turn";
     let settled = false, terminalError = null, killTimer = null;
+    let reportedTokens = -1, reportedContext = null;
     let diagnostic = safeProviderDiagnostic({ model, retryAttempt });
     const halt = error => {
       if (settled || terminalError) return;
@@ -653,6 +656,17 @@ function spawnReview(command, args, prompt, { spawnImpl, executionTimeoutMs, sta
         for (const line of stdout.split("\n").filter(Boolean)) {
           let event;
           try { event = JSON.parse(line); } catch { continue; }
+          try {
+            if (event.type === 'thread.started' && typeof event.thread_id === 'string' && event.thread_id !== reportedContext) {
+              reportedContext = event.thread_id; onContext(reportedContext);
+            }
+            if (event.type === 'turn.completed') {
+              const input = event.usage?.input_tokens, output = event.usage?.output_tokens, tokens = input + output;
+              if ([input, output, tokens].every(value => Number.isSafeInteger(value) && value >= 0) && tokens > reportedTokens) {
+                reportedTokens = tokens; onUsage({ source: 'codex_exec', tokens, observedAt: Date.now() });
+              }
+            }
+          } catch { halt(providerFailure('provider_execution_error')); return; }
           if (event.type === "turn.started") observed = observed === "before_turn" ? "turn_started" : observed;
           else if (observed !== "before_turn" && ["turn.completed", "item.completed"].includes(event.type)) observed = "turn_output";
           phase = observed;