gsd-pi 2.74.0-dev.b741afb → 2.74.0-dev.b7c5f96

package/src/resources/extensions/gsd/auto/phases.ts

@@ -12,6 +12,7 @@ import { importExtensionModule, type ExtensionAPI, type ExtensionContext } from
 import type { AutoSession, SidecarItem } from "./session.js";
 import type { LoopDeps } from "./loop-deps.js";
 import type { PostUnitContext, PreVerificationOpts } from "../auto-post-unit.js";
+import type { Phase } from "../types.js";
 import {
   MAX_RECOVERY_CHARS,
   BUDGET_THRESHOLDS,
@@ -28,6 +29,8 @@ import { debugLog } from "../debug-logger.js";
 import { PROJECT_FILES } from "../detection.js";
 import { MergeConflictError } from "../git-service.js";
 import { setCurrentPhase, clearCurrentPhase } from "../../shared/gsd-phase-state.js";
+import { pauseAutoForProviderError } from "../provider-error-pause.js";
+import { resumeAutoAfterProviderDelay } from "../bootstrap/provider-error-resume.js";
 import { join, basename, dirname, parse as parsePath } from "node:path";
 import { existsSync, cpSync, readdirSync } from "node:fs";
 import {
@@ -47,6 +50,9 @@ import { withTimeout, FINALIZE_PRE_TIMEOUT_MS, FINALIZE_POST_TIMEOUT_MS } from "
 import { getEligibleSlices } from "../slice-parallel-eligibility.js";
 import { startSliceParallel } from "../slice-parallel-orchestrator.js";
 import { isDbAvailable, getMilestoneSlices } from "../gsd-db.js";
+import { ensurePlanV2Graph } from "../uok/plan-v2.js";
+import { resolveUokFlags } from "../uok/flags.js";
+import { UokGateRunner } from "../uok/gate-runner.js";
 import { resetEvidence } from "../safety/evidence-collector.js";
 import { createCheckpoint, cleanupCheckpoint, rollbackToCheckpoint } from "../safety/git-checkpoint.js";
 import { resolveSafetyHarnessConfig } from "../safety/safety-harness.js";
@@ -55,6 +61,15 @@ import {
   getRequiredWorkflowToolsForAutoUnit,
 } from "../workflow-mcp.js";
 
+// ─── Session timeout auto-resume state ────────────────────────────────────────
+
+let consecutiveSessionTimeouts = 0;
+const MAX_SESSION_TIMEOUT_AUTO_RESUMES = 3;
+
+export function resetSessionTimeoutState(): void {
+  consecutiveSessionTimeouts = 0;
+}
+
 // ─── generateMilestoneReport ──────────────────────────────────────────────────
 
 /**
@@ -77,6 +92,17 @@ export function _resolveDispatchGuardBasePath(
   return s.originalBasePath || s.basePath;
 }
 
+const PLAN_V2_GATE_PHASES: ReadonlySet<Phase> = new Set([
+  "executing",
+  "summarizing",
+  "validating-milestone",
+  "completing-milestone",
+]);
+
+function shouldRunPlanV2Gate(phase: Phase): boolean {
+  return PLAN_V2_GATE_PHASES.has(phase);
+}
+
 /**
  * Generate and write an HTML milestone report snapshot.
  * Extracted from the milestone-transition block in autoLoop.
@@ -202,14 +228,60 @@ export async function runPreDispatch(
   loopState: LoopState,
 ): Promise<PhaseResult<PreDispatchData>> {
   const { ctx, pi, s, deps, prefs } = ic;
+  const uokFlags = resolveUokFlags(prefs);
+  const runPreDispatchGate = async (input: {
+    gateId: string;
+    gateType: string;
+    outcome: "pass" | "fail" | "retry" | "manual-attention";
+    failureClass: "none" | "policy" | "input" | "execution" | "artifact" | "verification" | "closeout" | "git" | "timeout" | "manual-attention" | "unknown";
+    rationale: string;
+    findings?: string;
+    milestoneId?: string;
+  }): Promise<void> => {
+    if (!uokFlags.gates) return;
+    const gateRunner = new UokGateRunner();
+    gateRunner.register({
+      id: input.gateId,
+      type: input.gateType,
+      execute: async () => ({
+        outcome: input.outcome,
+        failureClass: input.failureClass,
+        rationale: input.rationale,
+        findings: input.findings ?? "",
+      }),
+    });
+    await gateRunner.run(input.gateId, {
+      basePath: s.basePath,
+      traceId: `pre-dispatch:${ic.flowId}`,
+      turnId: `iter-${ic.iteration}`,
+      milestoneId: input.milestoneId ?? s.currentMilestoneId ?? undefined,
+      unitType: "pre-dispatch",
+      unitId: `iter-${ic.iteration}`,
+    });
+  };
 
   // Resource version guard
   const staleMsg = deps.checkResourcesStale(s.resourceVersionOnStart);
   if (staleMsg) {
+    await runPreDispatchGate({
+      gateId: "resource-version-guard",
+      gateType: "policy",
+      outcome: "fail",
+      failureClass: "policy",
+      rationale: "resource version guard blocked dispatch",
+      findings: staleMsg,
+    });
     await deps.stopAuto(ctx, pi, staleMsg);
     debugLog("autoLoop", { phase: "exit", reason: "resources-stale" });
     return { action: "break", reason: "resources-stale" };
   }
+  await runPreDispatchGate({
+    gateId: "resource-version-guard",
+    gateType: "policy",
+    outcome: "pass",
+    failureClass: "none",
+    rationale: "resource version guard passed",
+  });
 
   deps.invalidateAllCaches();
   s.lastPromptCharCount = undefined;
@@ -225,6 +297,14 @@ export async function runPreDispatch(
       );
     }
     if (!healthGate.proceed) {
+      await runPreDispatchGate({
+        gateId: "pre-dispatch-health-gate",
+        gateType: "execution",
+        outcome: "manual-attention",
+        failureClass: "manual-attention",
+        rationale: "pre-dispatch health gate blocked dispatch",
+        findings: healthGate.reason,
+      });
       ctx.ui.notify(
         healthGate.reason || "Pre-dispatch health check failed — run /gsd doctor for details.",
         "error",
@@ -233,7 +313,23 @@ export async function runPreDispatch(
       debugLog("autoLoop", { phase: "exit", reason: "health-gate-failed" });
       return { action: "break", reason: "health-gate-failed" };
     }
+    await runPreDispatchGate({
+      gateId: "pre-dispatch-health-gate",
+      gateType: "execution",
+      outcome: "pass",
+      failureClass: "none",
+      rationale: "pre-dispatch health gate passed",
+      findings: healthGate.fixesApplied.length > 0 ? healthGate.fixesApplied.join(", ") : "",
+    });
   } catch (e) {
+    await runPreDispatchGate({
+      gateId: "pre-dispatch-health-gate",
+      gateType: "execution",
+      outcome: "manual-attention",
+      failureClass: "manual-attention",
+      rationale: "pre-dispatch health gate threw unexpectedly",
+      findings: String(e),
+    });
     logWarning("engine", "Pre-dispatch health gate threw unexpectedly", { error: String(e) });
   }
 
@@ -252,6 +348,32 @@ export async function runPreDispatch(
 
   // Derive state
   let state = await deps.deriveState(s.basePath);
+  if (prefs?.uok?.plan_v2?.enabled && shouldRunPlanV2Gate(state.phase)) {
+    const compiled = ensurePlanV2Graph(s.basePath, state);
+    if (!compiled.ok) {
+      const reason = compiled.reason ?? "Plan v2 compilation failed";
+      await runPreDispatchGate({
+        gateId: "plan-v2-gate",
+        gateType: "policy",
+        outcome: "manual-attention",
+        failureClass: "manual-attention",
+        rationale: "plan v2 compile gate failed",
+        findings: reason,
+        milestoneId: state.activeMilestone?.id ?? undefined,
+      });
+      ctx.ui.notify(`Plan gate failed-closed: ${reason}`, "error");
+      await deps.pauseAuto(ctx, pi);
+      return { action: "break", reason: "plan-v2-gate-failed" };
+    }
+    await runPreDispatchGate({
+      gateId: "plan-v2-gate",
+      gateType: "policy",
+      outcome: "pass",
+      failureClass: "none",
+      rationale: "plan v2 compile gate passed",
+      milestoneId: state.activeMilestone?.id ?? undefined,
+    });
+  }
   deps.syncCmuxSidebar(prefs, state);
   let mid = state.activeMilestone?.id;
   let midTitle = state.activeMilestone?.title;
@@ -297,7 +419,10 @@ export async function runPreDispatch(
             s.basePath,
             mid,
             eligible,
-            { maxWorkers: prefs.slice_parallel.max_workers ?? 2 },
+            {
+              maxWorkers: prefs.slice_parallel.max_workers ?? 2,
+              useExecutionGraph: uokFlags.executionGraph,
+            },
           );
           if (result.started.length > 0) {
             ctx.ui.notify(
@@ -1117,6 +1242,8 @@ export async function runUnitPhase(
   // unit in the same Node process (see workflow-logger.ts module header).
   _resetLogs();
   s.currentUnit = { type: unitType, id: unitId, startedAt: Date.now() };
+  s.lastGitActionFailure = null;
+  s.lastGitActionStatus = null;
   setCurrentPhase(unitType);
   s.lastToolInvocationError = null; // #2883: clear stale error from previous unit
   const unitStartSeq = ic.nextSeq();
@@ -1386,24 +1513,75 @@ export async function runUnitPhase(
       debugLog("autoLoop", { phase: "exit", reason: "provider-pause", isTransient: unitResult.errorContext.isTransient });
       return { action: "break", reason: "provider-pause" };
     }
-    // Session creation timeout (not a structural error): pause auto-mode
-    // and let the provider-error-resume timer handle recovery (#3767). This
-    // matches the provider-pause path — break out cleanly, don't hard-stop.
+    // Timeout category covers two distinct scenarios:
+    //   1. Session creation timeout (120s) — transient, auto-resume with backoff
+    //   2. Unit hard timeout (30min+) — stuck agent, pause for manual review
     // Structural errors (TypeError, is not a function) are NOT transient
     // and must hard-stop to avoid infinite retry loops.
     if (
       unitResult.errorContext?.isTransient &&
       unitResult.errorContext?.category === "timeout"
     ) {
+      const isSessionCreationTimeout = unitResult.errorContext.message?.includes("Session creation timed out");
+
+      if (isSessionCreationTimeout) {
+        consecutiveSessionTimeouts += 1;
+        const baseRetryAfterMs = 30_000;
+        const retryAfterMs = baseRetryAfterMs * 2 ** Math.max(0, consecutiveSessionTimeouts - 1);
+        const allowAutoResume = consecutiveSessionTimeouts <= MAX_SESSION_TIMEOUT_AUTO_RESUMES;
+
+        if (!allowAutoResume) {
+          ctx.ui.notify(
+            `Session creation timed out ${consecutiveSessionTimeouts} consecutive times for ${unitType} ${unitId}. Pausing for manual review.`,
+            "warning",
+          );
+        }
+
+        debugLog("autoLoop", {
+          phase: "session-timeout-pause",
+          unitType, unitId,
+          consecutiveSessionTimeouts,
+          retryAfterMs,
+          allowAutoResume,
+        });
+
+        const errorDetail = ` for ${unitType} ${unitId}`;
+        await pauseAutoForProviderError(
+          ctx.ui,
+          errorDetail,
+          () => deps.pauseAuto(ctx, pi),
+          {
+            isRateLimit: false,
+            isTransient: allowAutoResume,
+            retryAfterMs,
+            resume: allowAutoResume
+              ? () => {
+                  void resumeAutoAfterProviderDelay(pi, ctx).catch((err) => {
+                    const message = err instanceof Error ? err.message : String(err);
+                    ctx.ui.notify(
+                      `Session timeout recovery failed: ${message}`,
+                      "error",
+                    );
+                  });
+                }
+              : undefined,
+          },
+        );
+        await deps.autoCommitUnit?.(s.basePath, unitType, unitId, ctx);
+        await emitCancelledUnitEnd(ic, unitType, unitId, unitStartSeq, unitResult.errorContext);
+        return { action: "break", reason: "session-timeout" };
+      }
+
+      // Unit hard timeout (30min+): pause without auto-resume — stuck agent
       ctx.ui.notify(
-        `Session creation timed out for ${unitType} ${unitId}. Pausing auto-mode (recoverable).`,
+        `Unit timed out for ${unitType} ${unitId} (supervision may have failed). Pausing auto-mode.`,
         "warning",
       );
-      debugLog("autoLoop", { phase: "session-timeout-pause", unitType, unitId });
+      debugLog("autoLoop", { phase: "unit-hard-timeout-pause", unitType, unitId });
       await deps.pauseAuto(ctx, pi);
       await deps.autoCommitUnit?.(s.basePath, unitType, unitId, ctx);
       await emitCancelledUnitEnd(ic, unitType, unitId, unitStartSeq, unitResult.errorContext);
-      return { action: "break", reason: "session-timeout" };
+      return { action: "break", reason: "unit-hard-timeout" };
     }
     // All other cancelled states (structural errors, non-transient failures): hard stop
     if (s.currentUnit) {
@@ -1433,6 +1611,8 @@ export async function runUnitPhase(
   // Guard: stopAuto() may have nulled s.currentUnit via s.reset() while
   // this coroutine was suspended at `await runUnit(...)` (#2939).
   if (s.currentUnit) {
+    // Reset session timeout counter — any successful unit clears the slate
+    consecutiveSessionTimeouts = 0;
     await deps.closeoutUnit(
       ctx,
       s.basePath,
@@ -1625,11 +1805,15 @@ export async function runFinalize(
 
   const preResult = preResultGuard.value;
   if (preResult === "dispatched") {
+    const dispatchedReason = s.lastGitActionFailure
+      ? "git-closeout-failure"
+      : "pre-verification-dispatched";
     debugLog("autoLoop", {
       phase: "exit",
-      reason: "pre-verification-dispatched",
+      reason: dispatchedReason,
+      gitError: s.lastGitActionFailure ?? undefined,
     });
-    return { action: "break", reason: "pre-verification-dispatched" };
+    return { action: "break", reason: dispatchedReason };
   }
   if (preResult === "retry") {
     if (sidecarItem) {

package/src/resources/extensions/gsd/auto/session.ts

@@ -106,6 +106,8 @@ export class AutoSession {
 
   // ── Current unit ─────────────────────────────────────────────────────────
   currentUnit: CurrentUnit | null = null;
+  currentTraceId: string | null = null;
+  currentTurnId: string | null = null;
   currentUnitRouting: UnitRouting | null = null;
   currentMilestoneId: string | null = null;
 
@@ -137,6 +139,10 @@ export class AutoSession {
   /** Set when a GSD tool execution ends with isError due to malformed/truncated
    *  JSON arguments. Checked by postUnitPreVerification to break retry loops. */
   lastToolInvocationError: string | null = null;
+  /** Set when turn-level git action fails during closeout. */
+  lastGitActionFailure: string | null = null;
+  /** Last turn-level git action status captured during finalize. */
+  lastGitActionStatus: "ok" | "failed" | null = null;
 
   // ── Isolation degradation ────────────────────────────────────────────
   /** Set to true when worktree creation fails; prevents merge of nonexistent branch. */
@@ -219,6 +225,8 @@ export class AutoSession {
 
     // Unit
     this.currentUnit = null;
+    this.currentTraceId = null;
+    this.currentTurnId = null;
     this.currentUnitRouting = null;
     this.currentMilestoneId = null;
 
@@ -250,6 +258,8 @@ export class AutoSession {
     this.rewriteAttemptCount = 0;
     this.consecutiveCompleteBootstraps = 0;
     this.lastToolInvocationError = null;
+    this.lastGitActionFailure = null;
+    this.lastGitActionStatus = null;
     this.isolationDegraded = false;
     this.milestoneMergedInPhases = false;
     this.checkpointSha = null;

package/src/resources/extensions/gsd/auto-dispatch.ts

@@ -53,6 +53,8 @@ import {
   checkNeedsRunUat,
 } from "./auto-prompts.js";
 import { resolveModelWithFallbacksForUnit } from "./preferences-models.js";
+import { resolveUokFlags } from "./uok/flags.js";
+import { selectReactiveDispatchBatch } from "./uok/execution-graph.js";
 
 // ─── Types ────────────────────────────────────────────────────────────────
 
@@ -584,12 +586,20 @@ export const DISPATCH_RULES: DispatchRule[] = [
         // Only activate reactive dispatch when >1 task is ready
         if (readyIds.length <= 1) return null;
 
-        const selected = chooseNonConflictingSubset(
-          readyIds,
-          graph,
-          maxParallel,
-          new Set(),
-        );
+        const uokFlags = resolveUokFlags(prefs);
+        const selected = uokFlags.executionGraph
+          ? selectReactiveDispatchBatch({
+              graph,
+              readyIds,
+              maxParallel,
+              inFlightOutputs: new Set(),
+            }).selected
+          : chooseNonConflictingSubset(
+              readyIds,
+              graph,
+              maxParallel,
+              new Set(),
+            );
         if (selected.length <= 1) return null;
 
         // Log graph metrics for observability

package/src/resources/extensions/gsd/auto-model-selection.ts

@@ -10,12 +10,14 @@ import type { ExtensionAPI, ExtensionContext } from "@gsd/pi-coding-agent";
 import type { GSDPreferences } from "./preferences.js";
 import { resolveModelWithFallbacksForUnit, resolveDynamicRoutingConfig } from "./preferences.js";
 import type { ComplexityTier } from "./complexity-classifier.js";
-import { classifyUnitComplexity, tierLabel } from "./complexity-classifier.js";
+import { classifyUnitComplexity, extractTaskMetadata, tierLabel } from "./complexity-classifier.js";
 import { resolveModelForComplexity, escalateTier, getEligibleModels, loadCapabilityOverrides, adjustToolSet, filterToolsForProvider } from "./model-router.js";
 import { getLedger, getProjectTotals } from "./metrics.js";
 import { unitPhaseLabel } from "./auto-dashboard.js";
 import { getSessionModelOverride } from "./session-model-override.js";
 import { logWarning } from "./workflow-logger.js";
+import { resolveUokFlags } from "./uok/flags.js";
+import { applyModelPolicyFilter } from "./uok/model-policy.js";
 
 export interface ModelSelectionResult {
   /** Routing metadata for metrics recording */
@@ -75,6 +77,7 @@ export async function selectAndApplyModel(
   /** Explicit /gsd model pin captured at bootstrap for long-running auto loops. */
   sessionModelOverride?: { provider: string; id: string } | null,
 ): Promise<ModelSelectionResult> {
+  const uokFlags = resolveUokFlags(prefs);
   const effectiveSessionModelOverride = sessionModelOverride === undefined
     ? getSessionModelOverride(ctx.sessionManager.getSessionId())
     : (sessionModelOverride ?? undefined);
@@ -97,6 +100,9 @@ export async function selectAndApplyModel(
 
   if (modelConfig) {
     const availableModels = ctx.modelRegistry.getAvailable();
+    const modelPolicyTraceId = `model:${ctx.sessionManager.getSessionId()}:${Date.now()}`;
+    const modelPolicyTurnId = `${unitType}:${unitId}`;
+    let policyAllowedModelKeys: Set<string> | null = null;
 
     // ─── Dynamic Model Routing ─────────────────────────────────────────
     // Dynamic routing (complexity-based downgrading) only applies in auto-mode.
@@ -106,8 +112,40 @@ export async function selectAndApplyModel(
     if (!isAutoMode) {
       routingConfig.enabled = false;
     }
+    // burn-max defaults to quality-first dispatch (no downgrade routing).
+    if (prefs?.token_profile === "burn-max") {
+      routingConfig.enabled = false;
+    }
     let effectiveModelConfig = modelConfig;
     let routingTierLabel = "";
+    let routingEligibleModels = availableModels;
+
+    const taskMetadataForPolicy = unitType === "execute-task"
+      ? extractTaskMetadata(unitId, basePath)
+      : undefined;
+
+    if (uokFlags.modelPolicy) {
+      const policy = applyModelPolicyFilter(
+        availableModels,
+        {
+          basePath,
+          traceId: modelPolicyTraceId,
+          turnId: modelPolicyTurnId,
+          unitType,
+          taskMetadata: taskMetadataForPolicy,
+          currentProvider: ctx.model?.provider,
+          allowCrossProvider: routingConfig.cross_provider !== false,
+          requiredTools: pi.getActiveTools(),
+        },
+      );
+      routingEligibleModels = policy.eligible;
+      policyAllowedModelKeys = new Set(
+        policy.eligible.map((m) => `${m.provider.toLowerCase()}/${m.id.toLowerCase()}`),
+      );
+      if (routingEligibleModels.length === 0) {
+        throw new Error(`Model policy denied all candidate models for ${unitType}/${unitId}`);
+      }
+    }
 
     // Disable routing for flat-rate providers like GitHub Copilot (#3453).
     // All models cost the same per request, so downgrading to a cheaper
@@ -115,7 +153,7 @@ export async function selectAndApplyModel(
     // Fail-closed: if primary model can't be resolved, fall back to
     // provider-level signals rather than allowing unwanted downgrades.
     if (routingConfig.enabled) {
-      const primaryModel = resolveModelId(modelConfig.primary, availableModels, ctx.model?.provider);
+      const primaryModel = resolveModelId(modelConfig.primary, routingEligibleModels, ctx.model?.provider);
       if (primaryModel) {
         const primaryFlatRateCtx = buildFlatRateContext(primaryModel.provider, ctx, prefs);
         if (isFlatRateProvider(primaryModel.provider, primaryFlatRateCtx)) {
@@ -149,8 +187,14 @@ export async function selectAndApplyModel(
       const shouldClassify = !isHook || routingConfig.hooks !== false;
 
       if (shouldClassify) {
-        let classification = classifyUnitComplexity(unitType, unitId, basePath, budgetPct);
-        const availableModelIds = availableModels.map(m => m.id);
+        let classification = classifyUnitComplexity(
+          unitType,
+          unitId,
+          basePath,
+          budgetPct,
+          taskMetadataForPolicy,
+        );
+        const availableModelIds = routingEligibleModels.map(m => m.id);
 
         // Escalate tier on retry when escalate_on_failure is enabled (default: true)
         if (
@@ -257,15 +301,28 @@ export async function selectAndApplyModel(
     }
 
     const modelsToTry = [effectiveModelConfig.primary, ...effectiveModelConfig.fallbacks];
+    let attemptedPolicyEligible = false;
 
     for (const modelId of modelsToTry) {
-      const model = resolveModelId(modelId, availableModels, ctx.model?.provider);
+      const resolutionPool = uokFlags.modelPolicy ? routingEligibleModels : availableModels;
+      const model = resolveModelId(modelId, resolutionPool, ctx.model?.provider);
 
       if (!model) {
         if (verbose) ctx.ui.notify(`Model ${modelId} not found, trying fallback.`, "info");
         continue;
       }
 
+      if (policyAllowedModelKeys) {
+        const key = `${model.provider.toLowerCase()}/${model.id.toLowerCase()}`;
+        if (!policyAllowedModelKeys.has(key)) {
+          if (verbose) {
+            ctx.ui.notify(`Model policy denied ${model.provider}/${model.id}; trying fallback.`, "warning");
+          }
+          continue;
+        }
+        attemptedPolicyEligible = true;
+      }
+
       // Warn if the ID is ambiguous across providers
       if (!modelId.includes("/")) {
         const providers = availableModels.filter(m => m.id === modelId).map(m => m.provider);
@@ -331,6 +388,10 @@ export async function selectAndApplyModel(
         }
       }
     }
+
+    if (uokFlags.modelPolicy && policyAllowedModelKeys && !attemptedPolicyEligible) {
+      throw new Error(`Model policy denied dispatch for ${unitType}/${unitId} before prompt send`);
+    }
   } else if (autoModeStartModel) {
     // No model preference for this unit type — re-apply the model captured
     // at auto-mode start to prevent bleed from shared global settings.json (#650).
@@ -474,10 +535,10 @@ export function buildFlatRateContext(
   prefs?: { flat_rate_providers?: readonly string[] },
 ): FlatRateContext {
   let authMode: FlatRateContext["authMode"];
-  const getAuthMode = ctx?.modelRegistry?.getProviderAuthMode;
-  if (typeof getAuthMode === "function") {
+  const registry = ctx?.modelRegistry;
+  if (registry && typeof registry.getProviderAuthMode === "function") {
     try {
-      const mode = getAuthMode(provider);
+      const mode = registry.getProviderAuthMode(provider);
       if (mode === "apiKey" || mode === "oauth" || mode === "externalCli" || mode === "none") {
         authMode = mode;
       }