gsd-pi 2.63.0 → 2.64.0-dev.9c14bd0

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

@@ -26,13 +26,20 @@ import { runUnit } from "./run-unit.js";
 import { debugLog } from "../debug-logger.js";
 import { PROJECT_FILES } from "../detection.js";
 import { MergeConflictError } from "../git-service.js";
-import { join, basename } from "node:path";
-import { existsSync, cpSync } from "node:fs";
+import { join, basename, dirname, parse as parsePath } from "node:path";
+import { existsSync, cpSync, readdirSync } from "node:fs";
 import { logWarning, logError } from "../workflow-logger.js";
 import { gsdRoot } from "../paths.js";
 import { atomicWriteSync } from "../atomic-write.js";
 import { verifyExpectedArtifact, diagnoseExpectedArtifact, buildLoopRemediationSteps } from "../auto-recovery.js";
 import { writeUnitRuntimeRecord } from "../unit-runtime.js";
+import { withTimeout, FINALIZE_POST_TIMEOUT_MS } from "./finalize-timeout.js";
+import { getEligibleSlices } from "../slice-parallel-eligibility.js";
+import { startSliceParallel } from "../slice-parallel-orchestrator.js";
+import { isDbAvailable, getMilestoneSlices } from "../gsd-db.js";
+import { resetEvidence } from "../safety/evidence-collector.js";
+import { createCheckpoint, cleanupCheckpoint, rollbackToCheckpoint } from "../safety/git-checkpoint.js";
+import { resolveSafetyHarnessConfig } from "../safety/safety-harness.js";
 
 // ─── generateMilestoneReport ──────────────────────────────────────────────────
 
@@ -218,6 +225,63 @@ export async function runPreDispatch(
     statePhase: state.phase,
   });
 
+  // ── Slice-level parallelism gate (#2340) ─────────────────────────────
+  // When slice_parallel is enabled, check if multiple slices are eligible
+  // for parallel execution. If so, dispatch them in parallel and stop the
+  // sequential loop. Workers are spawned via slice-parallel-orchestrator.ts.
+  if (
+    prefs?.slice_parallel?.enabled &&
+    mid &&
+    !process.env.GSD_PARALLEL_WORKER &&
+    isDbAvailable()
+  ) {
+    try {
+      const dbSlices = getMilestoneSlices(mid);
+      if (dbSlices.length > 0) {
+        const doneIds = new Set(dbSlices.filter(sl => sl.status === "complete" || sl.status === "done").map(sl => sl.id));
+        const sliceInputs = dbSlices.map(sl => ({
+          id: sl.id,
+          done: doneIds.has(sl.id),
+          depends: sl.depends ?? [],
+        }));
+        const eligible = getEligibleSlices(sliceInputs, doneIds);
+        if (eligible.length > 1) {
+          debugLog("autoLoop", {
+            phase: "slice-parallel-dispatch",
+            iteration: ic.iteration,
+            mid,
+            eligibleSlices: eligible.map(e => e.id),
+          });
+          ctx.ui.notify(
+            `Slice-parallel: dispatching ${eligible.length} eligible slices for ${mid}.`,
+            "info",
+          );
+          const result = await startSliceParallel(
+            s.basePath,
+            mid,
+            eligible,
+            { maxWorkers: prefs.slice_parallel.max_workers ?? 2 },
+          );
+          if (result.started.length > 0) {
+            ctx.ui.notify(
+              `Slice-parallel: started ${result.started.length} worker(s): ${result.started.join(", ")}.`,
+              "info",
+            );
+            await deps.stopAuto(ctx, pi, `Slice-parallel dispatched for ${mid}`);
+            return { action: "break", reason: "slice-parallel-dispatched" };
+          }
+          // Fall through to sequential if no workers started
+        }
+      }
+    } catch (err) {
+      debugLog("autoLoop", {
+        phase: "slice-parallel-check-error",
+        error: err instanceof Error ? err.message : String(err),
+      });
+      // Non-fatal — fall through to sequential dispatch
+    }
+  }
+
   // ── Milestone transition ────────────────────────────────────────────
   if (mid && s.currentMilestoneId && mid !== s.currentMilestoneId) {
     deps.emitJournalEvent({ ts: new Date().toISOString(), flowId: ic.flowId, seq: ic.nextSeq(), eventType: "milestone-transition", data: { from: s.currentMilestoneId, to: mid } });
@@ -948,11 +1012,38 @@ export async function runUnitPhase(
     }
     const hasProjectFile = PROJECT_FILES.some((f) => deps.existsSync(join(s.basePath, f)));
     const hasSrcDir = deps.existsSync(join(s.basePath, "src"));
-    if (!hasProjectFile && !hasSrcDir) {
+    // Xcode bundles have project-specific names (*.xcodeproj, *.xcworkspace)
+    // that cannot be matched by exact filename — scan the directory by suffix.
+    let hasXcodeBundle = false;
+    try {
+      const entries = deps.existsSync(s.basePath) ? readdirSync(s.basePath) : [];
+      hasXcodeBundle = entries.some((e: string) => e.endsWith(".xcodeproj") || e.endsWith(".xcworkspace"));
+    } catch (err) {
+      debugLog("runUnitPhase", { phase: "xcode-bundle-scan-failed", basePath: s.basePath, error: String(err) });
+    }
+    // Monorepo support (#2347): if no project files in the worktree directory,
+    // walk parent directories up to the filesystem root. In monorepos,
+    // package.json / Cargo.toml etc. live in a parent directory.
+    let hasProjectFileInParent = false;
+    if (!hasProjectFile && !hasSrcDir && !hasXcodeBundle) {
+      let checkDir = dirname(s.basePath);
+      const { root } = parsePath(checkDir);
+      while (checkDir !== root) {
+        // Stop at git repository boundary — ancestors above the repo root
+        // (e.g. ~ or /usr/local) may contain unrelated project files.
+        if (deps.existsSync(join(checkDir, ".git"))) break;
+        if (PROJECT_FILES.some((f) => deps.existsSync(join(checkDir, f)))) {
+          hasProjectFileInParent = true;
+          break;
+        }
+        checkDir = dirname(checkDir);
+      }
+    }
+    if (!hasProjectFile && !hasSrcDir && !hasXcodeBundle && !hasProjectFileInParent) {
       // Greenfield projects won't have project files yet — the first task creates them.
       // Log a warning but allow execution to proceed. The .git check above is sufficient
       // to ensure we're in a valid working directory.
-      debugLog("runUnitPhase", { phase: "worktree-health-warn-greenfield", basePath: s.basePath, hasProjectFile, hasSrcDir });
+      debugLog("runUnitPhase", { phase: "worktree-health-warn-greenfield", basePath: s.basePath, hasProjectFile, hasSrcDir, hasXcodeBundle });
       ctx.ui.notify(`Warning: ${s.basePath} has no recognized project files — proceeding as greenfield project`, "warning");
     }
   }
@@ -966,6 +1057,7 @@ export async function runUnitPhase(
   const previousTier = s.currentUnitRouting?.tier;
 
   s.currentUnit = { type: unitType, id: unitId, startedAt: Date.now() };
+  s.lastToolInvocationError = null; // #2883: clear stale error from previous unit
   const unitStartSeq = ic.nextSeq();
   deps.emitJournalEvent({ ts: new Date().toISOString(), flowId: ic.flowId, seq: unitStartSeq, eventType: "unit-start", data: { unitType, unitId } });
   deps.captureAvailableSkills();
@@ -985,30 +1077,25 @@ export async function runUnitPhase(
     },
   );
 
-  // Select and apply model (with tier escalation on retry — normal units only)
-  const modelResult = await deps.selectAndApplyModel(
-    ctx,
-    pi,
-    unitType,
-    unitId,
-    s.basePath,
-    prefs,
-    s.verbose,
-    s.autoModeStartModel,
-    sidecarItem ? undefined : { isRetry, previousTier },
-  );
-  s.currentUnitRouting =
-    modelResult.routing as AutoSession["currentUnitRouting"];
-  s.currentUnitModel =
-    modelResult.appliedModel as AutoSession["currentUnitModel"];
-
-  // Status bar + progress widget
+  // Status bar (widget + preconditions deferred until after model selection — see #2899)
   ctx.ui.setStatus("gsd-auto", "auto");
   if (mid)
     deps.updateSliceProgressCache(s.basePath, mid, state.activeSlice?.id);
-  deps.updateProgressWidget(ctx, unitType, unitId, state);
 
-  deps.ensurePreconditions(unitType, unitId, s.basePath, state);
+  // ── Safety harness: reset evidence + create checkpoint ──
+  const safetyConfig = resolveSafetyHarnessConfig(
+    prefs?.safety_harness as Record<string, unknown> | undefined,
+  );
+  if (safetyConfig.enabled && safetyConfig.evidence_collection) {
+    resetEvidence();
+  }
+  // Only checkpoint code-executing units (not lifecycle/planning units)
+  if (safetyConfig.enabled && safetyConfig.checkpoints && unitType === "execute-task") {
+    s.checkpointSha = createCheckpoint(s.basePath, unitId);
+    if (s.checkpointSha) {
+      debugLog("runUnitPhase", { phase: "checkpoint-created", unitId, sha: s.checkpointSha.slice(0, 8) });
+    }
+  }
 
   // Prompt injection
   let finalPrompt = prompt;
@@ -1074,6 +1161,23 @@ export async function runUnitPhase(
     logWarning("engine", "Prompt reorder failed", { error: msg });
   }
 
+  // Select and apply model (with tier escalation on retry — normal units only)
+  const modelResult = await deps.selectAndApplyModel(
+    ctx,
+    pi,
+    unitType,
+    unitId,
+    s.basePath,
+    prefs,
+    s.verbose,
+    s.autoModeStartModel,
+    sidecarItem ? undefined : { isRetry, previousTier },
+  );
+  s.currentUnitRouting =
+    modelResult.routing as AutoSession["currentUnitRouting"];
+  s.currentUnitModel =
+    modelResult.appliedModel as AutoSession["currentUnitModel"];
+
   // Apply sidecar/pre-dispatch hook model override (takes priority over standard model selection)
   const hookModelOverride = sidecarItem?.model ?? iterData.hookModelOverride;
   if (hookModelOverride) {
@@ -1099,6 +1203,17 @@ export async function runUnitPhase(
     }
   }
 
+  // Store the final dispatched model ID so the dashboard can read it (#2899).
+  // This accounts for hook model overrides applied after selectAndApplyModel.
+  s.currentDispatchedModelId = s.currentUnitModel
+    ? `${(s.currentUnitModel as any).provider ?? ""}/${(s.currentUnitModel as any).id ?? ""}`
+    : null;
+
+  // Progress widget + preconditions — deferred to after model selection so the
+  // widget's first render tick shows the correct model (#2899).
+  deps.updateProgressWidget(ctx, unitType, unitId, state);
+  deps.ensurePreconditions(unitType, unitId, s.basePath, state);
+
   // Start unit supervision
   deps.clearUnitTimeout();
   deps.startUnitSupervision({
@@ -1211,11 +1326,13 @@ export async function runUnitPhase(
     );
   }
 
-  // ── Zero tool-call guard (#1833) ──────────────────────────────────
-  // An execute-task agent that completes with 0 tool calls made no
-  // real changes — its summary is hallucinated. Treat as failed so
-  // the task is retried instead of silently marked complete.
-  if (unitType === "execute-task") {
+  // ── Zero tool-call guard (#1833, #2653) ──────────────────────────
+  // Any unit that completes with 0 tool calls made no real progress —
+  // likely context exhaustion where all tool calls errored out. Treat
+  // as failed so the unit is retried in a fresh context instead of
+  // silently passing through to artifact verification (which loops
+  // forever when the unit never produced its artifact).
+  {
     const currentLedger = deps.getLedger() as { units: Array<{ type: string; id: string; startedAt: number; toolCalls: number }> } | null;
     if (currentLedger?.units) {
       const lastUnit = [...currentLedger.units].reverse().find(
@@ -1226,14 +1343,14 @@ export async function runUnitPhase(
           phase: "zero-tool-calls",
           unitType,
           unitId,
-          warning: "Task completed with 0 tool calls — likely hallucinated, marking as failed",
+          warning: "Unit completed with 0 tool calls — likely context exhaustion, marking as failed",
         });
         ctx.ui.notify(
-          `${unitType} ${unitId} completed with 0 tool calls — hallucinated summary, will retry`,
+          `${unitType} ${unitId} completed with 0 tool calls — context exhaustion, will retry`,
           "warning",
         );
         // Fall through to next iteration where dispatch will re-derive
-        // and re-dispatch this task.
+        // and re-dispatch this unit.
         return { action: "next", data: { unitStartedAt: s.currentUnit?.startedAt } };
       }
     }
@@ -1277,6 +1394,27 @@ export async function runUnitPhase(
 
   deps.emitJournalEvent({ ts: new Date().toISOString(), flowId: ic.flowId, seq: ic.nextSeq(), eventType: "unit-end", data: { unitType, unitId, status: unitResult.status, artifactVerified, ...(unitResult.errorContext ? { errorContext: unitResult.errorContext } : {}) }, causedBy: { flowId: ic.flowId, seq: unitStartSeq } });
 
+  // ── Safety harness: checkpoint cleanup or rollback ──
+  if (s.checkpointSha) {
+    if (unitResult.status === "error" && safetyConfig.auto_rollback) {
+      const rolled = rollbackToCheckpoint(s.basePath, unitId, s.checkpointSha);
+      if (rolled) {
+        ctx.ui.notify(`Rolled back to pre-unit checkpoint for ${unitId}`, "info");
+        debugLog("runUnitPhase", { phase: "checkpoint-rollback", unitId });
+      }
+    } else if (unitResult.status === "error") {
+      ctx.ui.notify(
+        `Unit ${unitId} failed. Pre-unit checkpoint available at ${s.checkpointSha.slice(0, 8)}`,
+        "warning",
+      );
+    } else {
+      // Success — clean up checkpoint ref
+      cleanupCheckpoint(s.basePath, unitId);
+      debugLog("runUnitPhase", { phase: "checkpoint-cleaned", unitId });
+    }
+    s.checkpointSha = null;
+  }
+
   return { action: "next", data: { unitStartedAt: s.currentUnit?.startedAt } };
 }
 
@@ -1377,7 +1515,30 @@ export async function runFinalize(
   }
 
   // Post-verification processing (DB dual-write, hooks, triage, quick-tasks)
-  const postResult = await deps.postUnitPostVerification(postUnitCtx);
+  // Timeout guard: if postUnitPostVerification hangs (e.g., module import
+  // deadlock, SQLite transaction hang), force-continue after timeout so the
+  // auto-loop is not permanently frozen (#2344).
+  const postResultGuard = await withTimeout(
+    deps.postUnitPostVerification(postUnitCtx),
+    FINALIZE_POST_TIMEOUT_MS,
+    "postUnitPostVerification",
+  );
+
+  if (postResultGuard.timedOut) {
+    debugLog("autoLoop", {
+      phase: "post-verification-timeout",
+      iteration: ic.iteration,
+      unitType: iterData.unitType,
+      unitId: iterData.unitId,
+    });
+    ctx.ui.notify(
+      `postUnitPostVerification timed out after ${FINALIZE_POST_TIMEOUT_MS / 1000}s for ${iterData.unitType} ${iterData.unitId} — continuing to next iteration`,
+      "warning",
+    );
+    return { action: "next", data: undefined as void };
+  }
+
+  const postResult = postResultGuard.value;
 
   if (postResult === "stopped") {
     debugLog("autoLoop", {

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

@@ -105,6 +105,8 @@ export class AutoSession {
   // ── Model state ──────────────────────────────────────────────────────────
   autoModeStartModel: StartModel | null = null;
   currentUnitModel: Model<Api> | null = null;
+  /** Fully-qualified model ID (provider/id) set after selectAndApplyModel + hook overrides (#2899). */
+  currentDispatchedModelId: string | null = null;
   originalModelId: string | null = null;
   originalModelProvider: string | null = null;
   lastBudgetAlertLevel: BudgetAlertLevel = 0;
@@ -120,6 +122,11 @@ export class AutoSession {
   // ── Sidecar queue ─────────────────────────────────────────────────────
   sidecarQueue: SidecarItem[] = [];
 
+  // ── Tool invocation errors (#2883) ──────────────────────────────────
+  /** 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;
+
   // ── Isolation degradation ────────────────────────────────────────────
   /** Set to true when worktree creation fails; prevents merge of nonexistent branch. */
   isolationDegraded = false;
@@ -138,6 +145,10 @@ export class AutoSession {
   lastBaselineCharCount: number | undefined;
   pendingQuickTasks: CaptureEntry[] = [];
 
+  // ── Safety harness ───────────────────────────────────────────────────────
+  /** SHA of the pre-unit git checkpoint ref. Cleared on success or rollback. */
+  checkpointSha: string | null = null;
+
   // ── Signal handler ───────────────────────────────────────────────────────
   sigtermHandler: (() => void) | null = null;
 
@@ -193,6 +204,7 @@ export class AutoSession {
     // Model
     this.autoModeStartModel = null;
     this.currentUnitModel = null;
+    this.currentDispatchedModelId = null;
     this.originalModelId = null;
     this.originalModelProvider = null;
     this.lastBudgetAlertLevel = 0;
@@ -212,8 +224,10 @@ export class AutoSession {
     this.pendingQuickTasks = [];
     this.sidecarQueue = [];
     this.rewriteAttemptCount = 0;
+    this.lastToolInvocationError = null;
     this.isolationDegraded = false;
     this.milestoneMergedInPhases = false;
+    this.checkpointSha = null;
 
     // Signal handler
     this.sigtermHandler = null;

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

@@ -438,6 +438,8 @@ export interface WidgetStateAccessors {
   isVerbose(): boolean;
   /** True while newSession() is in-flight — render must not access session state. */
   isSessionSwitching(): boolean;
+  /** Fully-qualified dispatched model ID (provider/id) set after model selection + hook overrides (#2899). */
+  getCurrentDispatchedModelId(): string | null;
 }
 
 export function updateProgressWidget(
@@ -583,10 +585,11 @@ export function updateProgressWidget(
         lines.push(rightAlign(headerLeft, headerRight, width));
 
         // Worktree/branch right-aligned below header
-        if (worktreeName && cachedBranch) {
-          lines.push(rightAlign("", theme.fg("dim", `${worktreeName} (${cachedBranch})`), width));
-        } else if (cachedBranch) {
-          lines.push(rightAlign("", theme.fg("dim", cachedBranch), width));
+        const branchLabel = worktreeName && cachedBranch
+          ? `${worktreeName} (${cachedBranch})`
+          : cachedBranch ?? "";
+        if (branchLabel) {
+          lines.push(rightAlign("", theme.fg("dim", branchLabel), width));
         }
 
         // Show health signal details when degraded (yellow/red)
@@ -629,9 +632,15 @@ export function updateProgressWidget(
         const cxPctVal = cxUsage?.percent ?? 0;
         const cxPct = cxUsage?.percent !== null ? cxPctVal.toFixed(1) : "?";
 
-        // Model display — shown in context section, not stats
-        const modelId = cmdCtx?.model?.id ?? "";
-        const modelProvider = cmdCtx?.model?.provider ?? "";
+        // Model display — prefer dispatched model ID (set after selectAndApplyModel
+        // + hook overrides) over cmdCtx?.model which can be stale (#2899).
+        const dispatchedModelId = accessors.getCurrentDispatchedModelId();
+        const modelId = dispatchedModelId
+          ? dispatchedModelId.split("/").slice(1).join("/") || dispatchedModelId
+          : (cmdCtx?.model?.id ?? "");
+        const modelProvider = dispatchedModelId
+          ? dispatchedModelId.split("/")[0] || ""
+          : (cmdCtx?.model?.provider ?? "");
         const tierIcon = resolveServiceTierIcon(effectiveServiceTier, modelId);
         const modelDisplay = (modelProvider && modelId
           ? `${modelProvider}/${modelId}`

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

@@ -31,6 +31,9 @@ export function resolvePreferredModelConfig(
   const routingConfig = resolveDynamicRoutingConfig();
   if (!routingConfig.enabled || !routingConfig.tier_models) return undefined;
 
+  // Don't synthesize a routing config for flat-rate providers (#3453).
+  if (autoModeStartModel && isFlatRateProvider(autoModeStartModel.provider)) return undefined;
+
   const ceilingModel = routingConfig.tier_models.heavy
     ?? (autoModeStartModel ? `${autoModeStartModel.provider}/${autoModeStartModel.id}` : undefined);
   if (!ceilingModel) return undefined;
@@ -71,6 +74,27 @@ export async function selectAndApplyModel(
     let effectiveModelConfig = modelConfig;
     let routingTierLabel = "";
 
+    // Disable routing for flat-rate providers like GitHub Copilot (#3453).
+    // All models cost the same per request, so downgrading to a cheaper
+    // model provides no cost benefit — it only degrades quality.
+    // 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);
+      if (primaryModel) {
+        if (isFlatRateProvider(primaryModel.provider)) {
+          routingConfig.enabled = false;
+        }
+      } else if (
+        (autoModeStartModel && isFlatRateProvider(autoModeStartModel.provider))
+        || (ctx.model?.provider && isFlatRateProvider(ctx.model.provider))
+      ) {
+        // Primary model unresolvable but provider signals indicate flat-rate —
+        // disable routing to prevent quality degradation.
+        routingConfig.enabled = false;
+      }
+    }
+
     if (routingConfig.enabled) {
       let budgetPct: number | undefined;
       if (routingConfig.budget_pressure !== false) {
@@ -320,3 +344,15 @@ export function resolveModelId<T extends { id: string; provider: string }>(
   // Fall back to first non-extension candidate, or any candidate
   return candidates.find(m => !EXTENSION_PROVIDERS.has(m.provider)) ?? candidates[0];
 }
+
+/**
+ * Flat-rate providers charge the same per request regardless of model.
+ * Dynamic routing provides no cost benefit — it only degrades quality (#3453).
+ * Uses case-insensitive matching with alias support to prevent fail-open on
+ * provider naming variations (e.g. "copilot" vs "github-copilot").
+ */
+const FLAT_RATE_PROVIDERS = new Set(["github-copilot", "copilot"]);
+
+export function isFlatRateProvider(provider: string): boolean {
+  return FLAT_RATE_PROVIDERS.has(provider.toLowerCase());
+}