@vtxmacro/cli 2026.9.67 → 2026.9.69

package/bin/vtx-service-bootstrap.js

@@ -16,7 +16,7 @@ import { fileURLToPath } from "node:url";
 // agent-cli-release.json
 var agent_cli_release_default = {
   package_name: "@vtxmacro/cli",
-  package_version: "2026.9.67",
+  package_version: "2026.9.69",
   codex_package_name: "@openai/codex",
   codex_version: "0.155.0",
   copilot_sdk_package_name: "@github/copilot-sdk",
@@ -133,6 +133,14 @@ var runtime_policy_generated_default = {
       min_delay_milliseconds: 250,
       max_delay_milliseconds: 1e4
     },
+    trade_refresh_policy: {
+      server_config_stream_timeout_ms: 1e4,
+      config_retry_interval_ms: 5e3,
+      copy_trading_status_poll_ms: 5e3,
+      post_execution_refresh_delay_ms: 3e3,
+      account_read_backoff_min_ms: 2e3,
+      account_read_backoff_max_ms: 6e4
+    },
     deployment_reload_observability: {
       bootstrap_max_error_events_per_page: 4,
       bootstrap_max_profiles: 10
@@ -210,10 +218,12 @@ var runtime_policy_generated_default = {
       },
       server_runtime: {
         prompt_rescue: {
+          request_ttl_seconds: 20,
           snapshot_max_age_seconds: 5,
           snapshot_component_max_skew_seconds: 2,
-          client_long_poll_seconds: 20,
-          client_claim_http_timeout_milliseconds: 25e3,
+          client_claim_http_timeout_milliseconds: 5e3,
+          client_claim_retry_delay_milliseconds: 1e3,
+          client_failure_report_timeout_milliseconds: 2e3,
           client_presence_heartbeat_interval_milliseconds: 3e4,
           client_presence_http_timeout_milliseconds: 5e3
         }

package/bin/vtx-service-launcher.js

@@ -16,6 +16,14 @@ var runtime_policy_generated_default = {
       min_delay_milliseconds: 250,
       max_delay_milliseconds: 1e4
     },
+    trade_refresh_policy: {
+      server_config_stream_timeout_ms: 1e4,
+      config_retry_interval_ms: 5e3,
+      copy_trading_status_poll_ms: 5e3,
+      post_execution_refresh_delay_ms: 3e3,
+      account_read_backoff_min_ms: 2e3,
+      account_read_backoff_max_ms: 6e4
+    },
     deployment_reload_observability: {
       bootstrap_max_error_events_per_page: 4,
       bootstrap_max_profiles: 10
@@ -93,10 +101,12 @@ var runtime_policy_generated_default = {
       },
       server_runtime: {
         prompt_rescue: {
+          request_ttl_seconds: 20,
           snapshot_max_age_seconds: 5,
           snapshot_component_max_skew_seconds: 2,
-          client_long_poll_seconds: 20,
-          client_claim_http_timeout_milliseconds: 25e3,
+          client_claim_http_timeout_milliseconds: 5e3,
+          client_claim_retry_delay_milliseconds: 1e3,
+          client_failure_report_timeout_milliseconds: 2e3,
           client_presence_heartbeat_interval_milliseconds: 3e4,
           client_presence_http_timeout_milliseconds: 5e3
         }

package/bin/vtx.js

@@ -77,7 +77,7 @@ var init_agent_cli_release = __esm({
   "agent-cli-release.json"() {
     agent_cli_release_default = {
       package_name: "@vtxmacro/cli",
-      package_version: "2026.9.67",
+      package_version: "2026.9.69",
       codex_package_name: "@openai/codex",
       codex_version: "0.155.0",
       copilot_sdk_package_name: "@github/copilot-sdk",
@@ -15727,6 +15727,8 @@ var init_external_inference_contract = __esm({
       }
     });
     providerWeeklyQuotaReadSchema = external_exports.strictObject({
+      credits_available: external_exports.boolean().nullable().optional(),
+      rate_limit_reached: external_exports.boolean().nullable().optional(),
       source: external_exports.literal("codex_app_server"),
       used_percent: external_exports.number().finite().min(0).max(1e6),
       resets_at: timestampSchema,
@@ -20516,6 +20518,14 @@ var init_runtime_policy_generated = __esm({
           min_delay_milliseconds: 250,
           max_delay_milliseconds: 1e4
         },
+        trade_refresh_policy: {
+          server_config_stream_timeout_ms: 1e4,
+          config_retry_interval_ms: 5e3,
+          copy_trading_status_poll_ms: 5e3,
+          post_execution_refresh_delay_ms: 3e3,
+          account_read_backoff_min_ms: 2e3,
+          account_read_backoff_max_ms: 6e4
+        },
         deployment_reload_observability: {
           bootstrap_max_error_events_per_page: 4,
           bootstrap_max_profiles: 10
@@ -20593,10 +20603,12 @@ var init_runtime_policy_generated = __esm({
           },
           server_runtime: {
             prompt_rescue: {
+              request_ttl_seconds: 20,
               snapshot_max_age_seconds: 5,
               snapshot_component_max_skew_seconds: 2,
-              client_long_poll_seconds: 20,
-              client_claim_http_timeout_milliseconds: 25e3,
+              client_claim_http_timeout_milliseconds: 5e3,
+              client_claim_retry_delay_milliseconds: 1e3,
+              client_failure_report_timeout_milliseconds: 2e3,
               client_presence_heartbeat_interval_milliseconds: 3e4,
               client_presence_http_timeout_milliseconds: 5e3
             }
@@ -21099,6 +21111,7 @@ var init_mcp_client = __esm({
     agentDecisionTimingSchema = external_exports.strictObject({
       analysis_started_at: timestampSchema2,
       model_latency_ms: external_exports.number().int().nonnegative().nullable(),
+      analysis_duration_ms: external_exports.number().int().nonnegative().max(Number.MAX_SAFE_INTEGER).nullable().optional(),
       model_latency_source: external_exports.literal("harness_invocation").nullable()
     }).refine((timing) => timing.model_latency_ms === null === (timing.model_latency_source === null), {
       message: "Model latency and its measurement source must be supplied together."
@@ -21111,7 +21124,7 @@ var init_mcp_client = __esm({
       observed_at: timestampSchema2,
       timing: agentDecisionTimingSchema.optional(),
       provenance: jsonObjectSchema.nullable().optional()
-    }).refine((request) => !request.timing || Date.parse(request.timing.analysis_started_at) <= Date.parse(request.observed_at), {
+    }).refine((request) => !request.timing || Date.parse(request.timing.analysis_started_at) <= Date.parse(request.observed_at) && (request.timing.analysis_duration_ms == null || request.timing.analysis_duration_ms <= Date.parse(request.observed_at) - Date.parse(request.timing.analysis_started_at)), {
       message: "Analysis start must precede decision observation."
     });
     agentDecisionSubmitResultSchema = external_exports.strictObject({
@@ -21464,6 +21477,8 @@ var init_mcp_client = __esm({
       #requestId = 0;
       #initialized = false;
       #providerQuotaRecoverySupported = false;
+      #providerCreditsSupported = false;
+      #agentAnalysisDurationSupported = false;
       #runtimeRecoveryDiagnosticsSupported = false;
       #providerHttpStatusTools = /* @__PURE__ */ new Set();
       #grokToolValidationTools = /* @__PURE__ */ new Set();
@@ -21510,6 +21525,21 @@ var init_mcp_client = __esm({
       supportsRuntimeRecoveryDiagnostics() {
         return this.#runtimeRecoveryDiagnosticsSupported;
       }
+      supportsAgentAnalysisDuration() {
+        return this.#agentAnalysisDurationSupported;
+      }
+      async refreshAgentAnalysisDurationSupport(options = {}) {
+        if (this.supportsAgentAnalysisDuration()) return true;
+        try {
+          await this.initialize(options);
+        } catch {
+          return false;
+        }
+        return this.supportsAgentAnalysisDuration();
+      }
+      supportsProviderCredits() {
+        return this.#providerCreditsSupported;
+      }
       supportsProviderQuotaRecovery() {
         return this.#providerQuotaRecoverySupported;
       }
@@ -21572,6 +21602,35 @@ var init_mcp_client = __esm({
           "inference.host.heartbeat",
           ["provider_quota_recovery"]
         ) && advertisesFields("inference.job.fail", ["provider_retry_at", "provider_retry_source"]);
+        const hostHeartbeatSchema = listed.tools.find(
+          (tool) => tool.name === "inference.host.heartbeat"
+        )?.inputSchema;
+        const advertisesCredits = (value, depth = 0) => {
+          if (depth > 4 || !value || typeof value !== "object" || Array.isArray(value)) return false;
+          const node = value;
+          const properties = node.properties;
+          if (properties && typeof properties === "object" && Object.hasOwn(properties, "credits_available") && Object.hasOwn(properties, "rate_limit_reached")) return true;
+          if (typeof node.$ref === "string" && /^#\/\$defs\/[^/]+$/u.test(node.$ref)) {
+            const definition = node.$ref.slice("#/$defs/".length).replace(/~1/gu, "/").replace(/~0/gu, "~");
+            return advertisesCredits(hostHeartbeatSchema?.$defs?.[definition], depth + 1);
+          }
+          return Array.isArray(node.anyOf) && node.anyOf.some((branch) => advertisesCredits(branch, depth + 1));
+        };
+        this.#providerCreditsSupported = advertisesCredits(hostHeartbeatSchema?.properties?.provider_weekly_quota);
+        const decisionSchema = listed.tools.find(
+          (tool) => tool.name === "inference.agent.decision.submit"
+        )?.inputSchema;
+        const advertisesAnalysisDuration = (value, depth = 0) => {
+          if (depth > 4 || !value || typeof value !== "object" || Array.isArray(value)) return false;
+          const node = value;
+          if (node.properties && typeof node.properties === "object" && Object.hasOwn(node.properties, "analysis_duration_ms")) return true;
+          if (typeof node.$ref === "string" && /^#\/\$defs\/[^/]+$/u.test(node.$ref)) {
+            const definition = node.$ref.slice("#/$defs/".length).replace(/~1/gu, "/").replace(/~0/gu, "~");
+            return advertisesAnalysisDuration(decisionSchema?.$defs?.[definition], depth + 1);
+          }
+          return Array.isArray(node.anyOf) && node.anyOf.some((branch) => advertisesAnalysisDuration(branch, depth + 1));
+        };
+        this.#agentAnalysisDurationSupported = advertisesAnalysisDuration(decisionSchema?.properties?.timing);
         this.#grokToolValidationTools.clear();
         if (advertisesFields("inference.job.fail", ["grok_tool_validation"])) {
           this.#grokToolValidationTools.add("inference.job.fail");
@@ -28355,7 +28414,7 @@ async function logoutCodexSubscription(options) {
     await session.close();
   }
 }
-var CODEX_INFERENCE_PERMISSION_PROFILE, CODEX_AGENT_PERMISSION_PROFILE, CODEX_REASONING_CONTENT_MAX_UTF8_BYTES, CODEX_REASONING_SUMMARY_MAX_UTF8_BYTES, CODEX_ACCOUNT_PLAN_TYPES, CODEX_AGENT_DYNAMIC_TOOLS, CodexAppServerError, objectOrNull, finiteToken, tokenUsageFromBreakdown, usageFromNotification, validPlanType, nonnegativeSafeIntegerOrNull, nonnegativeFiniteNumberOrNull, parseRateLimitWindow, RATE_LIMIT_REACHED_TYPES, DEFAULT_CODEX_QUOTA_COOLDOWN_MS, MAX_CODEX_QUOTA_COOLDOWN_MS, CODEX_TRANSIENT_RATE_LIMIT_COOLDOWN_MS, parseRateLimitSnapshot, codexRateLimitRetryAtMs, codexAccountRateLimitReached, forbiddenMethod, forbiddenTerminalItem, forbiddenAgentMethod, forbiddenAgentTerminalItem, createCodexAgentToolSpecs, classifyCodexTurnFailure, scrubbedCodexEnvironment, killWindowsProcessTree, appServerArgs, GUARDIAN_SCRIPT, WINDOWS_RECEIPT_REPLACE_ERROR_CODES, MAX_RETIRED_RESPONSE_IDS, LATE_RESPONSE_SAFE_METHODS, replaceCodexGuardianReceiptFile, writeCodexGuardianSpawnIntent, parseGuardianReceipt, readCodexGuardianReceipt, waitForCodexGuardianState, defaultSpawn, CodexAppServerSession;
+var CODEX_INFERENCE_PERMISSION_PROFILE, CODEX_AGENT_PERMISSION_PROFILE, CODEX_REASONING_CONTENT_MAX_UTF8_BYTES, CODEX_REASONING_SUMMARY_MAX_UTF8_BYTES, CODEX_ACCOUNT_PLAN_TYPES, CODEX_AGENT_DYNAMIC_TOOLS, CodexAppServerError, objectOrNull, finiteToken, tokenUsageFromBreakdown, usageFromNotification, validPlanType, nonnegativeSafeIntegerOrNull, nonnegativeFiniteNumberOrNull, parseRateLimitWindow, RATE_LIMIT_REACHED_TYPES, DEFAULT_CODEX_QUOTA_COOLDOWN_MS, MAX_CODEX_QUOTA_COOLDOWN_MS, CODEX_TRANSIENT_RATE_LIMIT_COOLDOWN_MS, parseCreditsSnapshot, codexCreditsAvailable, parseRateLimitSnapshot, codexRateLimitRetryAtMs, codexAccountRateLimitReached, forbiddenMethod, forbiddenTerminalItem, forbiddenAgentMethod, forbiddenAgentTerminalItem, createCodexAgentToolSpecs, classifyCodexTurnFailure, scrubbedCodexEnvironment, killWindowsProcessTree, appServerArgs, GUARDIAN_SCRIPT, WINDOWS_RECEIPT_REPLACE_ERROR_CODES, MAX_RETIRED_RESPONSE_IDS, LATE_RESPONSE_SAFE_METHODS, replaceCodexGuardianReceiptFile, writeCodexGuardianSpawnIntent, parseGuardianReceipt, readCodexGuardianReceipt, waitForCodexGuardianState, defaultSpawn, CodexAppServerSession;
 var init_codex_app_server = __esm({
   "lib/inference-host/codex-app-server.ts"() {
     "use strict";
@@ -28470,6 +28529,19 @@ var init_codex_app_server = __esm({
     DEFAULT_CODEX_QUOTA_COOLDOWN_MS = 5 * 60 * 1e3;
     MAX_CODEX_QUOTA_COOLDOWN_MS = 8 * 24 * 60 * 60 * 1e3;
     CODEX_TRANSIENT_RATE_LIMIT_COOLDOWN_MS = 30 * 1e3;
+    parseCreditsSnapshot = (value) => {
+      const candidate = objectOrNull(value);
+      if (!candidate || typeof candidate.hasCredits !== "boolean" || typeof candidate.unlimited !== "boolean" || candidate.balance != null && typeof candidate.balance !== "string") return null;
+      return {
+        hasCredits: candidate.hasCredits,
+        unlimited: candidate.unlimited,
+        balance: candidate.balance ?? null
+      };
+    };
+    codexCreditsAvailable = (snapshot) => {
+      const credits = parseCreditsSnapshot(snapshot.credits);
+      return credits !== null && (credits.hasCredits || credits.unlimited);
+    };
     parseRateLimitSnapshot = (value) => {
       const candidate = objectOrNull(value);
       if (!candidate) throw new Error("invalid rate-limit snapshot");
@@ -28483,6 +28555,7 @@ var init_codex_app_server = __esm({
         limitId,
         limitName,
         planType,
+        credits: parseCreditsSnapshot(candidate.credits),
         primary: parseRateLimitWindow(candidate.primary ?? null),
         secondary: parseRateLimitWindow(candidate.secondary ?? null),
         rateLimitReachedType: reached,
@@ -28490,6 +28563,7 @@ var init_codex_app_server = __esm({
       };
     };
     codexRateLimitRetryAtMs = (accountRateLimits) => {
+      if (!codexAccountRateLimitReached(accountRateLimits)) return null;
       const snapshots = [accountRateLimits.rateLimits];
       const exhaustedResets = [];
       const reachedResets = [];
@@ -28510,7 +28584,12 @@ var init_codex_app_server = __esm({
       const candidates = exhaustedResets.length > 0 ? exhaustedResets : reachedResets.length > 0 ? reachedResets : [];
       return candidates.length > 0 ? Math.max(...candidates) : null;
     };
-    codexAccountRateLimitReached = (accountRateLimits) => [accountRateLimits.rateLimits].some((snapshot) => snapshot.rateLimitReachedType !== null || snapshot.spendControlReached === true || snapshot.primary?.usedPercent !== void 0 && snapshot.primary.usedPercent >= 100 || snapshot.secondary?.usedPercent !== void 0 && snapshot.secondary.usedPercent >= 100);
+    codexAccountRateLimitReached = (accountRateLimits) => [accountRateLimits.rateLimits].some((snapshot) => (
+      // The account snapshot reports generic included-usage exhaustion even when
+      // purchased credits remain. Workspace restrictions are distinct hard stops;
+      // actual turn failures remain authoritative in classifyCodexTurnFailure.
+      snapshot.rateLimitReachedType !== null && (snapshot.rateLimitReachedType !== "rate_limit_reached" || !codexCreditsAvailable(snapshot)) || snapshot.spendControlReached === true || !codexCreditsAvailable(snapshot) && (snapshot.primary?.usedPercent !== void 0 && snapshot.primary.usedPercent >= 100 || snapshot.secondary?.usedPercent !== void 0 && snapshot.secondary.usedPercent >= 100)
+    ));
     forbiddenMethod = (method) => method.startsWith("mcpServer/") || method.startsWith("serverRequest/") || method.startsWith("item/mcpToolCall") || method.startsWith("item/commandExecution") || method.startsWith("item/fileChange") || method.startsWith("item/dynamicToolCall") || method.startsWith("item/collabAgentToolCall");
     forbiddenTerminalItem = (item) => ![
       "userMessage",
@@ -30850,7 +30929,7 @@ async function clearInferenceAgentNextState(statePath) {
     "Agent-driven inference next recovery state"
   );
 }
-var inferenceAgentAttemptStatePath, inferenceAgentNextStatePath, assertDecisionTiming, inferenceAgentRuntimeStatePath, inferenceAgentFailureFencePath, isIsoTimestamp, assertPlainObject, hasExactKeys, assertInferenceAgentFailureFence, assertInferenceAgentRuntimeState, readCodexAgentRuntimeState, clearCodexAgentRuntimeState, foregroundAgentControlStatePath, assertRuntimeAssignment, assertForegroundAgentControlState, assertNextState, assertAttemptState;
+var inferenceAgentAttemptStatePath, inferenceAgentNextStatePath, measureAgentReviewTiming, assertDecisionTiming, inferenceAgentRuntimeStatePath, inferenceAgentFailureFencePath, isIsoTimestamp, assertPlainObject, hasExactKeys, assertInferenceAgentFailureFence, assertInferenceAgentRuntimeState, readCodexAgentRuntimeState, clearCodexAgentRuntimeState, foregroundAgentControlStatePath, assertRuntimeAssignment, assertForegroundAgentControlState, assertNextState, assertAttemptState;
 var init_agent_state = __esm({
   "lib/inference-host/agent-state.ts"() {
     "use strict";
@@ -30861,9 +30940,20 @@ var init_agent_state = __esm({
     init_external_inference_contract();
     inferenceAgentAttemptStatePath = (statePath) => `${statePath}.agent-attempt.json`;
     inferenceAgentNextStatePath = (statePath) => `${statePath}.agent-next.json`;
+    measureAgentReviewTiming = (analysisStartedAt, observedAt, measureDuration = true) => ({
+      analysis_started_at: analysisStartedAt,
+      model_latency_ms: null,
+      model_latency_source: null,
+      ...measureDuration ? { analysis_duration_ms: Date.parse(observedAt) - Date.parse(analysisStartedAt) } : {}
+    });
     assertDecisionTiming = (value, observedAt, message) => {
       const timing = assertPlainObject(value, message);
-      if (!hasExactKeys(timing, ["analysis_started_at", "model_latency_ms", "model_latency_source"]) || !isIsoTimestamp(timing.analysis_started_at) || Date.parse(timing.analysis_started_at) > Date.parse(String(observedAt)) || (timing.model_latency_ms === null ? timing.model_latency_source !== null : !Number.isSafeInteger(timing.model_latency_ms) || Number(timing.model_latency_ms) < 0 || timing.model_latency_source !== "harness_invocation")) throw new Error(message);
+      if (!hasExactKeys(timing, [
+        "analysis_started_at",
+        "model_latency_ms",
+        "model_latency_source",
+        ..."analysis_duration_ms" in timing ? ["analysis_duration_ms"] : []
+      ]) || !isIsoTimestamp(timing.analysis_started_at) || Date.parse(timing.analysis_started_at) > Date.parse(String(observedAt)) || timing.analysis_duration_ms != null && (!Number.isSafeInteger(timing.analysis_duration_ms) || Number(timing.analysis_duration_ms) < 0 || Number(timing.analysis_duration_ms) > Date.parse(String(observedAt)) - Date.parse(timing.analysis_started_at)) || (timing.model_latency_ms === null ? timing.model_latency_source !== null : !Number.isSafeInteger(timing.model_latency_ms) || Number(timing.model_latency_ms) < 0 || timing.model_latency_source !== "harness_invocation")) throw new Error(message);
     };
     inferenceAgentRuntimeStatePath = (statePath) => `${statePath}.codex-agent-runtime.json`;
     inferenceAgentFailureFencePath = (statePath) => `${statePath}.agent-failure-fence.json`;
@@ -31514,12 +31604,16 @@ var init_runner = __esm({
       started_at: startedAt
     });
     isoAt = (milliseconds) => new Date(milliseconds).toISOString();
-    providerWeeklyQuotaFromRateLimits = (rateLimits, observedAtMs, currentTimeMs = observedAtMs) => {
+    providerWeeklyQuotaFromRateLimits = (rateLimits, observedAtMs, currentTimeMs = observedAtMs, includeCredits = true) => {
       if (!rateLimits) return null;
       for (const window2 of [rateLimits.rateLimits.primary, rateLimits.rateLimits.secondary]) {
         if (window2?.windowDurationMins !== 7 * 24 * 60 || window2.resetsAt === null || window2.resetsAt * 1e3 <= currentTimeMs) continue;
         return {
           source: "codex_app_server",
+          ...includeCredits ? {
+            credits_available: codexCreditsAvailable(rateLimits.rateLimits),
+            rate_limit_reached: codexAccountRateLimitReached(rateLimits)
+          } : {},
           used_percent: window2.usedPercent,
           resets_at: isoAt(window2.resetsAt * 1e3),
           observed_at: isoAt(observedAtMs)
@@ -31531,7 +31625,7 @@ var init_runner = __esm({
       const window2 = rateLimits.rateLimits[key];
       const known = previous?.[key];
       if (known && (!window2 || window2.windowDurationMins !== known.windowDurationMins)) return false;
-      return window2 === null || window2.resetsAt !== null && window2.resetsAt * 1e3 > nowMs && window2.usedPercent < 100;
+      return window2 === null || window2.resetsAt !== null && window2.resetsAt * 1e3 > nowMs && (window2.usedPercent < 100 || codexCreditsAvailable(rateLimits.rateLimits));
     });
     finitePositiveOption = (value, fallback, label) => {
       const result2 = value ?? fallback;
@@ -32563,7 +32657,8 @@ var init_runner = __esm({
                 provider_weekly_quota: providerWeeklyQuotaFromRateLimits(
                   providerRateLimits,
                   providerRateLimitsObservedAtMs,
-                  now()
+                  now(),
+                  mcp.supportsProviderCredits?.() === true
                 ),
                 ...mcp.supportsProviderQuotaRecovery?.() === true && providerQuotaRecoveryObservedAtMs !== null && status === "healthy" && active.size === 0 && (providerRetryAtMs ?? 0) <= now() && providerFailureRetryAtMs <= now() && now() - providerQuotaRecoveryObservedAtMs <= 2 * settings.providerRateLimitRefreshMs && providerRateLimits !== null && usableProviderQuotaSnapshot(providerRateLimits, knownProviderQuotaSnapshot, now()) ? { provider_quota_recovery: {
                   source: "codex_app_server",
@@ -33181,14 +33276,7 @@ var init_runner = __esm({
                     providerRetryAtMs = Math.max(providerRetryAtMs ?? 0, refreshedRetryAt);
                     providerCooldownReason = "provider_limit_snapshot";
                     retainQuotaSnapshot(providerRateLimits);
-                  } else if (providerRetryAtMs !== null && !codexAccountRateLimitReached(providerRateLimits) && providerRateLimits.rateLimits.limitId === providerQuotaSnapshot?.limitId && [providerRateLimits.rateLimits.primary, providerRateLimits.rateLimits.secondary].some((window2) => window2 !== null) && ["primary", "secondary"].every((key) => {
-                    const window2 = refreshedRateLimits.rateLimits[key];
-                    const previous = providerQuotaSnapshot?.[key];
-                    if (previous && (!window2 || window2.windowDurationMins !== previous.windowDurationMins)) {
-                      return false;
-                    }
-                    return window2 === null || window2.resetsAt !== null && window2.resetsAt * 1e3 > refreshedAt && window2.usedPercent < 100;
-                  })) {
+                  } else if (providerRetryAtMs !== null && usableProviderQuotaSnapshot(providerRateLimits, providerQuotaSnapshot, refreshedAt)) {
                     emitDiagnostic("provider_cooldown_cleared", {
                       reason: providerCooldownReason,
                       recovery: "provider_quota_restored",
@@ -34560,6 +34648,7 @@ var init_runner = __esm({
       }
     };
     createInferenceAgentControlClient = (mcp) => ({
+      supportsAgentAnalysisDuration: async (options = {}) => mcp.supportsAgentAnalysisDuration?.() === true || await mcp.refreshAgentAnalysisDurationSupport?.(options) === true,
       supportsProviderHttpStatusDiagnostics: async (options = {}) => mcp.supportsProviderHttpStatusDiagnostics?.("inference.agent.assignment.heartbeat") === true || await mcp.refreshProviderHttpStatusSupport?.("inference.agent.assignment.heartbeat", options) === true,
       supportsGrokResponseIdentityDiagnostics: async (options = {}) => mcp.supportsGrokResponseIdentityDiagnostics?.("inference.agent.assignment.heartbeat") === true || await mcp.refreshGrokResponseIdentitySupport?.("inference.agent.assignment.heartbeat", options) === true,
       supportsGrokToolValidationDiagnostics: async (options = {}) => mcp.supportsGrokToolValidationDiagnostics?.("inference.agent.assignment.heartbeat") === true || await mcp.refreshGrokToolValidationSupport?.("inference.agent.assignment.heartbeat", options) === true,
@@ -34979,15 +35068,19 @@ var init_runner = __esm({
                 if (state.pending_decision) {
                   return { success: false, value: { error: "decision_outcome_unresolved" } };
                 }
+                const measureDuration = state.pending_turn.decision_operation_ids.length === 0 && await this.options.controlClient.supportsAgentAnalysisDuration?.({ signal: turnAbort.signal }) === true;
+                const observedAt = nowIso();
                 const pending3 = {
                   operation_id: randomUUID(),
                   candidate,
-                  observed_at: nowIso(),
-                  timing: {
-                    analysis_started_at: state.pending_turn.started_at,
-                    model_latency_ms: null,
-                    model_latency_source: null
-                  },
+                  observed_at: observedAt,
+                  // A later decision in the same turn has no independent review
+                  // boundary. Do not report the whole turn again as its duration.
+                  timing: measureAgentReviewTiming(
+                    state.pending_turn.started_at,
+                    observedAt,
+                    measureDuration
+                  ),
                   provenance: {
                     source: "external_agent",
                     agent_run_id: liveThread.threadId,
@@ -42642,6 +42735,14 @@ var init_runtime_policy_generated2 = __esm({
           min_delay_milliseconds: 250,
           max_delay_milliseconds: 1e4
         },
+        trade_refresh_policy: {
+          server_config_stream_timeout_ms: 1e4,
+          config_retry_interval_ms: 5e3,
+          copy_trading_status_poll_ms: 5e3,
+          post_execution_refresh_delay_ms: 3e3,
+          account_read_backoff_min_ms: 2e3,
+          account_read_backoff_max_ms: 6e4
+        },
         deployment_reload_observability: {
           bootstrap_max_error_events_per_page: 4,
           bootstrap_max_profiles: 10
@@ -42719,10 +42820,12 @@ var init_runtime_policy_generated2 = __esm({
           },
           server_runtime: {
             prompt_rescue: {
+              request_ttl_seconds: 20,
               snapshot_max_age_seconds: 5,
               snapshot_component_max_skew_seconds: 2,
-              client_long_poll_seconds: 20,
-              client_claim_http_timeout_milliseconds: 25e3,
+              client_claim_http_timeout_milliseconds: 5e3,
+              client_claim_retry_delay_milliseconds: 1e3,
+              client_failure_report_timeout_milliseconds: 2e3,
               client_presence_heartbeat_interval_milliseconds: 3e4,
               client_presence_http_timeout_milliseconds: 5e3
             }
@@ -46747,11 +46850,11 @@ Waiting for approval...
               operation_id: agentOperationId("agent-decision-submit"),
               candidate: input.candidate,
               observed_at: observedAt,
-              ...state.review ? { timing: {
-                analysis_started_at: state.review.analysis_started_at,
-                model_latency_ms: null,
-                model_latency_source: null
-              } } : {},
+              ...state.review ? { timing: measureAgentReviewTiming(
+                state.review.analysis_started_at,
+                observedAt,
+                session.client.supportsAgentAnalysisDuration?.() === true
+              ) } : {},
               provenance: input.provenance
             },
             first_transmit_at: null,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vtxmacro/cli",
-  "version": "2026.9.67",
+  "version": "2026.9.69",
   "description": "VTX Macro CLI, MCP server, and durable external inference host.",
   "type": "module",
   "license": "UNLICENSED",