gsd-pi 2.75.0 → 2.76.0

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

@@ -50,6 +50,7 @@ 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 type { MinimalModelRegistry } from "../context-budget.js";
 import { ensurePlanV2Graph } from "../uok/plan-v2.js";
 import { resolveUokFlags } from "../uok/flags.js";
 import { UokGateRunner } from "../uok/gate-runner.js";
@@ -59,6 +60,7 @@ import { resolveSafetyHarnessConfig } from "../safety/safety-harness.js";
 import {
   getWorkflowTransportSupportError,
   getRequiredWorkflowToolsForAutoUnit,
+  supportsStructuredQuestions,
 } from "../workflow-mcp.js";
 
 // ─── Session timeout auto-resume state ────────────────────────────────────────
@@ -216,6 +218,62 @@ async function emitCancelledUnitEnd(
   });
 }
 
+async function failClosedOnFinalizeTimeout(
+  ic: IterationContext,
+  iterData: IterationData,
+  loopState: LoopState,
+  stage: "pre" | "post",
+  startedAt: number,
+): Promise<PhaseResult> {
+  const { ctx, pi, s, deps } = ic;
+  const now = Date.now();
+  const unitType = iterData.unitType;
+  const unitId = iterData.unitId;
+  const timeoutMs = stage === "pre" ? FINALIZE_PRE_TIMEOUT_MS : FINALIZE_POST_TIMEOUT_MS;
+  const progressKind = stage === "pre" ? "finalize-pre-timeout" : "finalize-post-timeout";
+
+  writeUnitRuntimeRecord(s.basePath, unitType, unitId, startedAt, {
+    phase: "finalize-timeout",
+    timeoutAt: now,
+    lastProgressAt: now,
+    lastProgressKind: progressKind,
+  });
+
+  deps.emitJournalEvent({
+    ts: new Date(now).toISOString(),
+    flowId: ic.flowId,
+    seq: ic.nextSeq(),
+    eventType: "unit-end",
+    data: {
+      unitType,
+      unitId,
+      status: "timed-out-finalize",
+      artifactVerified: false,
+      finalizeStage: stage,
+    },
+  });
+
+  loopState.consecutiveFinalizeTimeouts++;
+  debugLog("autoLoop", {
+    phase: progressKind,
+    iteration: ic.iteration,
+    unitType,
+    unitId,
+    consecutiveTimeouts: loopState.consecutiveFinalizeTimeouts,
+  });
+
+  ctx.ui.notify(
+    `${stage === "pre" ? "postUnitPreVerification" : "postUnitPostVerification"} timed out after ${timeoutMs / 1000}s for ${unitType} ${unitId} (${loopState.consecutiveFinalizeTimeouts}/${MAX_FINALIZE_TIMEOUTS}) — pausing auto-mode for recovery.`,
+    "warning",
+  );
+
+  await deps.pauseAuto(ctx, pi);
+  s.currentUnit = null;
+  clearCurrentPhase();
+  drainLogs();
+  return { action: "break", reason: progressKind };
+}
+
 // ─── runPreDispatch ───────────────────────────────────────────────────────────
 
 /**
@@ -781,6 +839,15 @@ export async function runDispatch(
   const { ctx, pi, s, deps, prefs } = ic;
   const { state, mid, midTitle } = preData;
   const STUCK_WINDOW_SIZE = 6;
+  const provider = ctx.model?.provider;
+  const authMode = provider && typeof ctx.modelRegistry?.getProviderAuthMode === "function"
+    ? ctx.modelRegistry.getProviderAuthMode(provider)
+    : undefined;
+  const activeTools = typeof pi.getActiveTools === "function" ? pi.getActiveTools() : [];
+  const structuredQuestionsAvailable = supportsStructuredQuestions(activeTools, {
+    authMode,
+    baseUrl: ctx.model?.baseUrl,
+  }) ? "true" : "false";
 
   debugLog("autoLoop", { phase: "dispatch-resolve", iteration: ic.iteration });
   const dispatchResult = await deps.resolveDispatch({
@@ -790,6 +857,9 @@ export async function runDispatch(
     state,
     prefs,
     session: s,
+    structuredQuestionsAvailable,
+    sessionContextWindow: ctx.model?.contextWindow,
+    modelRegistry: ctx.modelRegistry as MinimalModelRegistry | undefined,
   });
 
   if (dispatchResult.action === "stop") {
@@ -848,6 +918,17 @@ export async function runDispatch(
           s.basePath,
         );
         if (artifactExists) {
+          if (unitType === "complete-milestone") {
+            const stuckDiag = diagnoseExpectedArtifact(unitType, unitId, s.basePath);
+            const stuckParts = [
+              `Detected ${unitType} ${unitId} output on disk, but the same unit is still being derived.`,
+              "This usually means the milestone summary exists while the DB row still does not mark the milestone complete.",
+            ];
+            if (stuckDiag) stuckParts.push(`Expected: ${stuckDiag}`);
+            ctx.ui.notify(stuckParts.join(" "), "warning");
+            await deps.pauseAuto(ctx, pi);
+            return { action: "break", reason: "complete-milestone-artifact-db-mismatch" };
+          }
           debugLog("autoLoop", {
             phase: "stuck-recovery",
             level: 1,
@@ -1241,6 +1322,8 @@ export async function runUnitPhase(
   // per-unit. Without this, the module-level _buffer accumulates across every
   // unit in the same Node process (see workflow-logger.ts module header).
   _resetLogs();
+  const dispatchKey = `${unitType}/${unitId}`;
+  s.unitDispatchCount.set(dispatchKey, (s.unitDispatchCount.get(dispatchKey) ?? 0) + 1);
   s.currentUnit = { type: unitType, id: unitId, startedAt: Date.now() };
   s.lastGitActionFailure = null;
   s.lastGitActionStatus = null;
@@ -1307,7 +1390,7 @@ export async function runUnitPhase(
         : s.pendingCrashRecovery;
     finalPrompt = `${capped}\n\n---\n\n${finalPrompt}`;
     s.pendingCrashRecovery = null;
-  } else if ((s.unitDispatchCount.get(`${unitType}/${unitId}`) ?? 0) > 1) {
+  } else if ((s.unitDispatchCount.get(dispatchKey) ?? 0) > 1) {
     const diagnostic = deps.getDeepDiagnostic(s.basePath);
     if (diagnostic) {
       const cappedDiag =
@@ -1666,7 +1749,7 @@ export async function runUnitPhase(
     skipArtifactVerification ||
     verifyExpectedArtifact(unitType, unitId, s.basePath);
   if (artifactVerified) {
-    s.unitDispatchCount.delete(`${unitType}/${unitId}`);
+    s.unitDispatchCount.delete(dispatchKey);
     s.unitRecoveryCount.delete(`${unitType}/${unitId}`);
   }
 
@@ -1771,36 +1854,13 @@ export async function runFinalize(
   );
 
   if (preResultGuard.timedOut) {
-    // Detach session from the timed-out unit so late async completions
-    // cannot mutate state for the next unit (#3757).
-    s.currentUnit = null;
-    clearCurrentPhase();
-    // Drop any logger entries from the timed-out unit so they don't bleed
-    // into the next iteration's drain.
-    drainLogs();
-    loopState.consecutiveFinalizeTimeouts++;
-    debugLog("autoLoop", {
-      phase: "pre-verification-timeout",
-      iteration: ic.iteration,
-      unitType: iterData.unitType,
-      unitId: iterData.unitId,
-      consecutiveTimeouts: loopState.consecutiveFinalizeTimeouts,
-    });
-
-    if (loopState.consecutiveFinalizeTimeouts >= MAX_FINALIZE_TIMEOUTS) {
-      ctx.ui.notify(
-        `postUnitPreVerification timed out ${loopState.consecutiveFinalizeTimeouts} consecutive times — stopping auto-mode to prevent budget waste`,
-        "error",
-      );
-      await deps.stopAuto(ctx, pi, `${loopState.consecutiveFinalizeTimeouts} consecutive finalize timeouts`);
-      return { action: "break", reason: "finalize-timeout-escalation" };
-    }
-
-    ctx.ui.notify(
-      `postUnitPreVerification timed out after ${FINALIZE_PRE_TIMEOUT_MS / 1000}s for ${iterData.unitType} ${iterData.unitId} (${loopState.consecutiveFinalizeTimeouts}/${MAX_FINALIZE_TIMEOUTS}) — continuing to next iteration`,
-      "warning",
+    return failClosedOnFinalizeTimeout(
+      ic,
+      iterData,
+      loopState,
+      "pre",
+      preUnitSnapshot?.startedAt ?? Date.now(),
     );
-    return { action: "next", data: undefined as void };
   }
 
   const preResult = preResultGuard.value;
@@ -1876,36 +1936,13 @@ export async function runFinalize(
   );
 
   if (postResultGuard.timedOut) {
-    // Detach session from the timed-out unit so late async completions
-    // cannot mutate state for the next unit (#3757).
-    s.currentUnit = null;
-    clearCurrentPhase();
-    // Drop any logger entries from the timed-out unit so they don't bleed
-    // into the next iteration's drain.
-    drainLogs();
-    loopState.consecutiveFinalizeTimeouts++;
-    debugLog("autoLoop", {
-      phase: "post-verification-timeout",
-      iteration: ic.iteration,
-      unitType: iterData.unitType,
-      unitId: iterData.unitId,
-      consecutiveTimeouts: loopState.consecutiveFinalizeTimeouts,
-    });
-
-    if (loopState.consecutiveFinalizeTimeouts >= MAX_FINALIZE_TIMEOUTS) {
-      ctx.ui.notify(
-        `postUnitPostVerification timed out ${loopState.consecutiveFinalizeTimeouts} consecutive times — stopping auto-mode to prevent budget waste`,
-        "error",
-      );
-      await deps.stopAuto(ctx, pi, `${loopState.consecutiveFinalizeTimeouts} consecutive finalize timeouts`);
-      return { action: "break", reason: "finalize-timeout-escalation" };
-    }
-
-    ctx.ui.notify(
-      `postUnitPostVerification timed out after ${FINALIZE_POST_TIMEOUT_MS / 1000}s for ${iterData.unitType} ${iterData.unitId} (${loopState.consecutiveFinalizeTimeouts}/${MAX_FINALIZE_TIMEOUTS}) — continuing to next iteration`,
-      "warning",
+    return failClosedOnFinalizeTimeout(
+      ic,
+      iterData,
+      loopState,
+      "post",
+      preUnitSnapshot?.startedAt ?? Date.now(),
     );
-    return { action: "next", data: undefined as void };
   }
 
   const postResult = postResultGuard.value;

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

@@ -73,7 +73,7 @@ export function isSessionSwitchInFlight(): boolean {
 /**
  * Force-resolve the pending unit promise with { status: "cancelled" }.
  *
- * Used by pauseAuto, handleAgentEnd early-return, and supervision catch
+ * Used by pauseAuto and supervision catch
  * blocks to ensure the autoLoop is never stuck awaiting a promise that
  * will never resolve. Safe to call when no resolver is pending (no-op).
  */

package/src/resources/extensions/gsd/auto/run-unit.ts

@@ -82,10 +82,20 @@ export async function runUnit(
   if (s.currentUnitModel && typeof pi.setModel === "function") {
     const restored = await pi.setModel(s.currentUnitModel, { persist: false });
     if (!restored) {
+      const message =
+        `Failed to restore configured model ${s.currentUnitModel.provider}/${s.currentUnitModel.id} after session creation`;
       ctx.ui.notify(
-        `Failed to restore ${s.currentUnitModel.provider}/${s.currentUnitModel.id} after session creation. Using session default.`,
+        `${message}. Cancelling unit before dispatch.`,
         "warning",
       );
+      return {
+        status: "cancelled",
+        errorContext: {
+          message,
+          category: "session-failed",
+          isTransient: false,
+        },
+      };
     }
   }
 

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

@@ -172,6 +172,10 @@ export class AutoSession {
   // ── Signal handler ───────────────────────────────────────────────────────
   sigtermHandler: (() => void) | null = null;
 
+  // ── Remote command polling ───────────────────────────────────────────────
+  /** Cleanup function returned by startCommandPolling(); null when not running. */
+  commandPollingCleanup: (() => void) | null = null;
+
   // ── Loop promise state ──────────────────────────────────────────────────
   // Per-unit resolve function and session-switch guard live at module level
   // in auto-loop.ts (_currentResolve, _sessionSwitchInFlight).
@@ -267,6 +271,9 @@ export class AutoSession {
     // Signal handler
     this.sigtermHandler = null;
 
+    // Remote command polling — cleanup must be called before reset (auto.ts stopAuto)
+    this.commandPollingCleanup = null;
+
     // Loop promise state lives in auto-loop.ts module scope
   }
 

package/src/resources/extensions/gsd/auto-artifact-paths.ts

@@ -43,6 +43,16 @@ export function resolveExpectedArtifactPath(
       return dir ? join(dir, buildMilestoneFileName(mid, "ROADMAP")) : null;
     }
     case "research-slice": {
+      // #4414: Sentinel unitId "{mid}/parallel-research" fans out across
+      // multiple slices. Resolve to a milestone-level placeholder path so
+      // blocker escalation has somewhere to write. Verification for this
+      // sentinel is handled directly in verifyExpectedArtifact.
+      if (sid === "parallel-research") {
+        const mdir = resolveMilestonePath(base, mid);
+        return mdir
+          ? join(mdir, buildMilestoneFileName(mid, "PARALLEL-BLOCKER"))
+          : null;
+      }
       const dir = resolveSlicePath(base, mid, sid!);
       return dir ? join(dir, buildSliceFileName(sid!, "RESEARCH")) : null;
     }
@@ -50,6 +60,11 @@ export function resolveExpectedArtifactPath(
       const dir = resolveSlicePath(base, mid, sid!);
       return dir ? join(dir, buildSliceFileName(sid!, "PLAN")) : null;
     }
+    case "refine-slice": {
+      // ADR-011: refine-slice expands a sketch and writes the same PLAN.md as plan-slice.
+      const dir = resolveSlicePath(base, mid, sid!);
+      return dir ? join(dir, buildSliceFileName(sid!, "PLAN")) : null;
+    }
     case "reassess-roadmap": {
       const dir = resolveSlicePath(base, mid, sid!);
       return dir ? join(dir, buildSliceFileName(sid!, "ASSESSMENT")) : null;
@@ -109,9 +124,14 @@ export function diagnoseExpectedArtifact(
     case "plan-milestone":
       return `${relMilestoneFile(base, mid, "ROADMAP")} (milestone roadmap)`;
     case "research-slice":
+      if (sid === "parallel-research") {
+        return `${relMilestoneFile(base, mid, "PARALLEL-BLOCKER")} (parallel slice research sentinel)`;
+      }
       return `${relSliceFile(base, mid, sid!, "RESEARCH")} (slice research)`;
     case "plan-slice":
       return `${relSliceFile(base, mid, sid!, "PLAN")} (slice plan)`;
+    case "refine-slice":
+      return `${relSliceFile(base, mid, sid!, "PLAN")} (refined slice plan from sketch)`;
     case "execute-task": {
       return `Task ${tid} marked [x] in ${relSliceFile(base, mid, sid!, "PLAN")} + summary written`;
     }

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

@@ -18,6 +18,7 @@ import { getCurrentBranch } from "./worktree.js";
 import { getActiveHook } from "./post-unit-hooks.js";
 import { getLedger, getProjectTotals } from "./metrics.js";
 import { getErrorMessage } from "./error-utils.js";
+import { nativeIsRepo } from "./native-git-bridge.js";
 import {
   resolveMilestoneFile,
   resolveSliceFile,
@@ -96,6 +97,7 @@ export function unitVerb(unitType: string): string {
     case "research-slice": return "researching";
     case "plan-milestone":
     case "plan-slice": return "planning";
+    case "refine-slice": return "refining";
     case "execute-task": return "executing";
     case "complete-slice": return "completing";
     case "replan-slice": return "replanning";
@@ -116,6 +118,7 @@ export function unitPhaseLabel(unitType: string): string {
     case "research-slice": return "RESEARCH";
     case "plan-milestone": return "PLAN";
     case "plan-slice": return "PLAN";
+    case "refine-slice": return "REFINE";
     case "execute-task": return "EXECUTE";
     case "complete-slice": return "COMPLETE";
     case "replan-slice": return "REPLAN";
@@ -143,6 +146,7 @@ function peekNext(unitType: string, state: GSDState): string {
     case "plan-milestone": return "plan or execute first slice";
     case "research-slice": return `plan ${sid}`;
     case "plan-slice": return "execute first task";
+    case "refine-slice": return "execute first task";
     case "execute-task": return `continue ${sid}`;
     case "complete-slice": return "reassess roadmap";
     case "replan-slice": return `re-execute ${sid}`;
@@ -333,6 +337,10 @@ let lastCommitFetchedAt = 0;
 
 function refreshLastCommit(basePath: string): void {
   try {
+    if (!nativeIsRepo(basePath)) {
+      cachedLastCommit = null;
+      return;
+    }
     const raw = execFileSync("git", ["log", "-1", "--format=%cr|%s"], {
       cwd: basePath,
       encoding: "utf-8",
@@ -346,10 +354,12 @@ function refreshLastCommit(basePath: string): void {
         message: raw.slice(sep + 1),
       };
     }
-    lastCommitFetchedAt = Date.now();
   } catch (err) {
     // Non-fatal — just skip last commit display
+    cachedLastCommit = null;
     logWarning("dashboard", `operation failed: ${err instanceof Error ? err.message : String(err)}`);
+  } finally {
+    lastCommitFetchedAt = Date.now();
   }
 }
 
@@ -361,6 +371,23 @@ function getLastCommit(basePath: string): { timeAgo: string; message: string } |
   return cachedLastCommit;
 }
 
+export function _resetLastCommitCacheForTests(): void {
+  cachedLastCommit = null;
+  lastCommitFetchedAt = 0;
+}
+
+export function _refreshLastCommitForTests(basePath: string): void {
+  refreshLastCommit(basePath);
+}
+
+export function _getLastCommitForTests(basePath: string): { timeAgo: string; message: string } | null {
+  return getLastCommit(basePath);
+}
+
+export function _getLastCommitFetchedAtForTests(): number {
+  return lastCommitFetchedAt;
+}
+
 // ─── Footer Factory ───────────────────────────────────────────────────────────
 
 /**
@@ -901,10 +928,24 @@ export function updateProgressWidget(
           }
           if (cumulativeCost) sp.push(theme.fg("warning", `$${cumulativeCost.toFixed(2)}`));
 
-          const cxDisplay = `${cxPct}%/${formatWidgetTokens(cxWindow)}`;
-          if (cxPctVal > 90) sp.push(theme.fg("error", cxDisplay));
-          else if (cxPctVal > 70) sp.push(theme.fg("warning", cxDisplay));
-          else sp.push(cxDisplay);
+          const CX_BAR_WIDTH = 8;
+          const cxBarFilled = Math.min(
+            CX_BAR_WIDTH,
+            Math.max(0, Math.round((cxPctVal / 100) * CX_BAR_WIDTH)),
+          );
+          const cxBarColor: "error" | "warning" | "success" =
+            cxPctVal > 90 ? "error" : cxPctVal > 70 ? "warning" : "success";
+          const cxBar =
+            theme.fg(cxBarColor, "━".repeat(cxBarFilled)) +
+            theme.fg("dim", "─".repeat(CX_BAR_WIDTH - cxBarFilled));
+          const cxPctText = `${cxPct}%/${formatWidgetTokens(cxWindow)}`;
+          const cxColorized =
+            cxPctVal > 90
+              ? theme.fg("error", cxPctText)
+              : cxPctVal > 70
+                ? theme.fg("warning", cxPctText)
+                : cxPctText;
+          sp.push(`${cxBar} ${cxColorized}`);
 
           const statsLine = sp.map(p => p.includes("\x1b[") ? p : theme.fg("dim", p))
             .join(theme.fg("dim", "  "));

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

@@ -28,6 +28,7 @@ import {
   buildReplanSlicePrompt,
 } from "./auto-prompts.js";
 import { loadEffectiveGSDPreferences } from "./preferences.js";
+import type { MinimalModelRegistry } from "./context-budget.js";
 import { pauseAuto } from "./auto.js";
 import {
   getWorkflowTransportSupportError,
@@ -104,7 +105,13 @@ export async function dispatchDirectPhase(
         }
         unitType = "plan-slice";
         unitId = `${mid}/${sid}`;
-        prompt = await buildPlanSlicePrompt(mid, midTitle, sid, sTitle, base);
+        prompt = await buildPlanSlicePrompt(
+          mid, midTitle, sid, sTitle, base, undefined,
+          {
+            sessionContextWindow: ctx.model?.contextWindow,
+            modelRegistry: ctx.modelRegistry as MinimalModelRegistry | undefined,
+          },
+        );
       } else {
         unitType = "plan-milestone";
         unitId = mid;
@@ -129,7 +136,13 @@ export async function dispatchDirectPhase(
       }
       unitType = "execute-task";
       unitId = `${mid}/${sid}/${tid}`;
-      prompt = await buildExecuteTaskPrompt(mid, sid, sTitle, tid, tTitle, base);
+      prompt = await buildExecuteTaskPrompt(
+        mid, sid, sTitle, tid, tTitle, base,
+        {
+          sessionContextWindow: ctx.model?.contextWindow,
+          modelRegistry: ctx.modelRegistry as MinimalModelRegistry | undefined,
+        },
+      );
       break;
     }