gsd-pi 2.76.0 → 2.77.0-dev.1594b263e

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

@@ -7,7 +7,7 @@ import { fileURLToPath } from "node:url";
 
 const __dirname = dirname(fileURLToPath(import.meta.url));
 
-import { resolvePreferredModelConfig, resolveModelId } from "../auto-model-selection.js";
+import { resolvePreferredModelConfig, resolveModelId, selectAndApplyModel } from "../auto-model-selection.js";
 
 function makeTempDir(prefix: string): string {
   return mkdtempSync(join(tmpdir(), prefix));
@@ -46,6 +46,7 @@ test("resolvePreferredModelConfig synthesizes heavy routing ceiling when models
     assert.deepEqual(config, {
       primary: "claude-opus-4-6",
       fallbacks: [],
+      source: "synthesized",
     });
   } finally {
     process.chdir(originalCwd);
@@ -88,6 +89,7 @@ test("resolvePreferredModelConfig falls back to auto start model when heavy tier
     assert.deepEqual(config, {
       primary: "openai/gpt-5.4",
       fallbacks: [],
+      source: "synthesized",
     });
   } finally {
     process.chdir(originalCwd);
@@ -131,6 +133,7 @@ test("resolvePreferredModelConfig keeps explicit phase models as the ceiling", (
     assert.deepEqual(config, {
       primary: "claude-sonnet-4-6",
       fallbacks: [],
+      source: "explicit",
     });
   } finally {
     process.chdir(originalCwd);
@@ -141,6 +144,163 @@ test("resolvePreferredModelConfig keeps explicit phase models as the ceiling", (
   }
 });
 
+test("selectAndApplyModel honors explicit phase models without downgrading (#3617)", async () => {
+  const originalCwd = process.cwd();
+  const originalGsdHome = process.env.GSD_HOME;
+  const tempProject = makeTempDir("gsd-routing-project-");
+  const tempGsdHome = makeTempDir("gsd-routing-home-");
+  const setModelCalls: string[] = [];
+  let beforeModelSelectCalled = false;
+
+  try {
+    mkdirSync(join(tempProject, ".gsd"), { recursive: true });
+    writeFileSync(
+      join(tempProject, ".gsd", "PREFERENCES.md"),
+      [
+        "---",
+        "models:",
+        "  planning: claude-opus-4-6",
+        "dynamic_routing:",
+        "  enabled: true",
+        "  tier_models:",
+        "    light: gpt-4o-mini",
+        "    standard: claude-sonnet-4-6",
+        "    heavy: claude-opus-4-6",
+        "---",
+      ].join("\n"),
+      "utf-8",
+    );
+    process.env.GSD_HOME = tempGsdHome;
+    process.chdir(tempProject);
+
+    const availableModels = [
+      { id: "claude-opus-4-6", provider: "anthropic", api: "anthropic-messages" },
+      { id: "claude-sonnet-4-6", provider: "anthropic", api: "anthropic-messages" },
+      { id: "gpt-4o-mini", provider: "openai", api: "responses" },
+    ];
+
+    const result = await selectAndApplyModel(
+      {
+        modelRegistry: { getAvailable: () => availableModels },
+        sessionManager: { getSessionId: () => "test-session" },
+        ui: { notify: () => {} },
+        model: { provider: "anthropic", id: "claude-opus-4-6", api: "anthropic-messages" },
+      } as any,
+      {
+        setModel: async (model: { provider: string; id: string }) => {
+          setModelCalls.push(`${model.provider}/${model.id}`);
+          return true;
+        },
+        emitBeforeModelSelect: async () => {
+          beforeModelSelectCalled = true;
+          return undefined;
+        },
+        getActiveTools: () => [],
+        emitAdjustToolSet: async () => undefined,
+        setActiveTools: () => {},
+      } as any,
+      "plan-slice",
+      "slice-1",
+      tempProject,
+      undefined,
+      false,
+      { provider: "anthropic", id: "claude-opus-4-6" },
+      undefined,
+      true,
+    );
+
+    assert.equal(beforeModelSelectCalled, false, "explicit phase models should skip dynamic routing hooks");
+    assert.deepEqual(setModelCalls, ["anthropic/claude-opus-4-6"]);
+    assert.equal(result.routing, null, "explicit phase models should not record a routing downgrade");
+    assert.equal(result.appliedModel?.provider, "anthropic");
+    assert.equal(result.appliedModel?.id, "claude-opus-4-6");
+  } finally {
+    process.chdir(originalCwd);
+    if (originalGsdHome === undefined) delete process.env.GSD_HOME;
+    else process.env.GSD_HOME = originalGsdHome;
+    rmSync(tempProject, { recursive: true, force: true });
+    rmSync(tempGsdHome, { recursive: true, force: true });
+  }
+});
+
+test("selectAndApplyModel escalates dynamic routing tier when retry metadata is provided", async (t) => {
+  const originalCwd = process.cwd();
+  const originalGsdHome = process.env.GSD_HOME;
+  const tempProject = makeTempDir("gsd-routing-retry-project-");
+  const tempGsdHome = makeTempDir("gsd-routing-retry-home-");
+  const setModelCalls: string[] = [];
+  const notifications: Array<{ message: string; level: string }> = [];
+
+  t.after(() => {
+    process.chdir(originalCwd);
+    if (originalGsdHome === undefined) delete process.env.GSD_HOME;
+    else process.env.GSD_HOME = originalGsdHome;
+    rmSync(tempProject, { recursive: true, force: true });
+    rmSync(tempGsdHome, { recursive: true, force: true });
+  });
+
+  mkdirSync(join(tempProject, ".gsd"), { recursive: true });
+  writeFileSync(
+    join(tempProject, ".gsd", "PREFERENCES.md"),
+    [
+      "---",
+      "dynamic_routing:",
+      "  enabled: true",
+      "  hooks: false",
+      "  budget_pressure: false",
+      "  tier_models:",
+      "    light: claude-haiku-4-5",
+      "    standard: claude-sonnet-4-6",
+      "    heavy: claude-opus-4-6",
+      "---",
+    ].join("\n"),
+    "utf-8",
+  );
+  process.env.GSD_HOME = tempGsdHome;
+  process.chdir(tempProject);
+
+  const availableModels = [
+    { id: "claude-haiku-4-5", provider: "anthropic", api: "anthropic-messages" },
+    { id: "claude-sonnet-4-6", provider: "anthropic", api: "anthropic-messages" },
+    { id: "claude-opus-4-6", provider: "anthropic", api: "anthropic-messages" },
+  ];
+
+  const result = await selectAndApplyModel(
+    {
+      modelRegistry: { getAvailable: () => availableModels },
+      sessionManager: { getSessionId: () => "test-session" },
+      ui: { notify: (message: string, level: string) => notifications.push({ message, level }) },
+      model: { provider: "anthropic", id: "claude-opus-4-6", api: "anthropic-messages" },
+    } as any,
+    {
+      setModel: async (model: { provider: string; id: string }) => {
+        setModelCalls.push(`${model.provider}/${model.id}`);
+        return true;
+      },
+      emitBeforeModelSelect: async () => undefined,
+      getActiveTools: () => [],
+      emitAdjustToolSet: async () => undefined,
+      setActiveTools: () => {},
+    } as any,
+    "execute-task",
+    "M001/S01/T01",
+    tempProject,
+    undefined,
+    false,
+    { provider: "anthropic", id: "claude-opus-4-6" },
+    { isRetry: true, previousTier: "light" },
+    true,
+  );
+
+  assert.deepEqual(setModelCalls, ["anthropic/claude-sonnet-4-6"]);
+  assert.deepEqual(result.routing, { tier: "standard", modelDowngraded: true });
+  assert.equal(result.appliedModel?.id, "claude-sonnet-4-6");
+  assert.ok(
+    notifications.some(n => n.message.includes("Tier escalation: light") && n.message.includes("standard")),
+    "retry metadata should produce a visible tier escalation notification",
+  );
+});
+
 // ─── resolveModelId tests ─────────────────────────────────────────────────
 
 test("resolveModelId: bare ID resolves to claude-code when session is claude-code (#3772)", () => {
@@ -259,6 +419,18 @@ test("dynamic routing passes provider-qualified model keys to the router", () =>
   );
 });
 
+test("selectAndApplyModel re-applies captured thinking level after setModel success", () => {
+  const src = readFileSync(join(__dirname, "..", "auto-model-selection.ts"), "utf-8");
+  assert.ok(
+    src.includes("autoModeStartThinkingLevel?: ReturnType<ExtensionAPI[\"getThinkingLevel\"]> | null"),
+    "selectAndApplyModel should accept an autoModeStartThinkingLevel parameter",
+  );
+  assert.ok(
+    src.includes("reapplyThinkingLevel(pi, autoModeStartThinkingLevel)"),
+    "selectAndApplyModel should re-apply captured thinking level after model changes",
+  );
+});
+
 test("resolveModelId: anthropic wins over claude-code when session provider is not claude-code", () => {
   const availableModels = [
     { id: "claude-sonnet-4-6", provider: "claude-code" },

package/src/resources/extensions/gsd/tests/auto-paused-session-validation.test.ts

@@ -39,6 +39,18 @@ test("auto.ts validates milestone before restoring paused session (#1664)", () =
     source.includes('resolveMilestoneFile(base, meta.milestoneId, "SUMMARY")'),
     "auto.ts must check for SUMMARY file to detect completed milestones",
   );
+
+  // Resume path must sanitize paused session file metadata before unlink/recovery.
+  assert.ok(
+    source.includes("normalizeSessionFilePath(meta.sessionFile ?? null)"),
+    "auto.ts must sanitize paused-session metadata sessionFile before using it",
+  );
+
+  // Pause path must sanitize live session file path before persisting metadata.
+  assert.ok(
+    source.includes("normalizeSessionFilePath(ctx?.sessionManager?.getSessionFile() ?? null)"),
+    "auto.ts must sanitize sessionManager getSessionFile output before persisting",
+  );
 });
 
 // ─── Filesystem validation unit tests ───────────────────────────────────────

package/src/resources/extensions/gsd/tests/auto-phases-lifecycle.test.ts

@@ -0,0 +1,61 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { runFinalize } from "../auto/phases.ts";
+import { AutoSession } from "../auto/session.ts";
+
+test("runFinalize clears currentUnit after successful finalize", async () => {
+  const s = new AutoSession();
+  s.basePath = "/tmp/gsd-finalize-current-unit";
+  s.currentUnit = {
+    type: "execute-task",
+    id: "M001/S01/T01",
+    startedAt: Date.now(),
+  };
+
+  const deps = {
+    clearUnitTimeout() {},
+    buildSnapshotOpts() {
+      return {};
+    },
+    stopAuto: async () => {},
+    pauseAuto: async () => {},
+    updateProgressWidget() {},
+    postUnitPreVerification: async () => "continue",
+    runPostUnitVerification: async () => "continue",
+    postUnitPostVerification: async () => "continue",
+  };
+
+  const result = await runFinalize(
+    {
+      ctx: { ui: { notify() {} } },
+      pi: {},
+      s,
+      deps,
+      prefs: undefined,
+      iteration: 1,
+      flowId: "flow-1",
+      nextSeq: () => 1,
+    } as any,
+    {
+      unitType: "execute-task",
+      unitId: "M001/S01/T01",
+      prompt: "",
+      finalPrompt: "",
+      pauseAfterUatDispatch: false,
+      state: {} as any,
+      mid: "M001",
+      midTitle: "Milestone",
+      isRetry: false,
+      previousTier: undefined,
+    },
+    {
+      recentUnits: [],
+      stuckRecoveryAttempts: 0,
+      consecutiveFinalizeTimeouts: 0,
+    },
+  );
+
+  assert.equal(result.action, "next");
+  assert.equal(s.currentUnit, null);
+});

package/src/resources/extensions/gsd/tests/auto-recovery.test.ts

@@ -706,6 +706,79 @@ test("verifyExpectedArtifact complete-milestone passes with impl files (#1703)",
   }
 });
 
+test("verifyExpectedArtifact complete-milestone fails when DB milestone is not complete (#4658)", () => {
+  const base = makeGitBase();
+  try {
+    execFileSync("git", ["checkout", "-b", "feat/ms-db-active"], { cwd: base, stdio: "ignore" });
+    mkdirSync(join(base, ".gsd", "milestones", "M001"), { recursive: true });
+    writeFileSync(join(base, ".gsd", "milestones", "M001", "M001-SUMMARY.md"), "# Milestone Summary\nverification FAILED — not complete.");
+    mkdirSync(join(base, "src"), { recursive: true });
+    writeFileSync(join(base, "src", "app.ts"), "console.log('hello');");
+    execFileSync("git", ["add", "."], { cwd: base, stdio: "ignore" });
+    execFileSync("git", ["commit", "-m", "feat: implementation with failed summary"], { cwd: base, stdio: "ignore" });
+
+    openDatabase(join(base, ".gsd", "gsd.db"));
+    insertMilestone({ id: "M001", title: "Milestone One", status: "active" });
+
+    const result = verifyExpectedArtifact("complete-milestone", "M001", base);
+    assert.equal(result, false, "complete-milestone must fail when DB status is not complete");
+  } finally {
+    cleanup(base);
+  }
+});
+
+test("verifyExpectedArtifact complete-milestone passes when DB milestone is complete (#4658)", () => {
+  const base = makeGitBase();
+  try {
+    execFileSync("git", ["checkout", "-b", "feat/ms-db-complete"], { cwd: base, stdio: "ignore" });
+    mkdirSync(join(base, ".gsd", "milestones", "M001"), { recursive: true });
+    writeFileSync(join(base, ".gsd", "milestones", "M001", "M001-SUMMARY.md"), "# Milestone Summary\nDone.");
+    mkdirSync(join(base, "src"), { recursive: true });
+    writeFileSync(join(base, "src", "app.ts"), "console.log('hello');");
+    execFileSync("git", ["add", "."], { cwd: base, stdio: "ignore" });
+    execFileSync("git", ["commit", "-m", "feat: implementation complete"], { cwd: base, stdio: "ignore" });
+
+    openDatabase(join(base, ".gsd", "gsd.db"));
+    insertMilestone({ id: "M001", title: "Milestone One", status: "complete" });
+
+    const result = verifyExpectedArtifact("complete-milestone", "M001", base);
+    assert.equal(result, true, "complete-milestone should pass when DB status is complete");
+  } finally {
+    cleanup(base);
+  }
+});
+
+test("verifyExpectedArtifact complete-milestone tolerates transient DB lag when SUMMARY is canonical success (#4658)", () => {
+  const base = makeGitBase();
+  try {
+    execFileSync("git", ["checkout", "-b", "feat/ms-db-lag-success"], { cwd: base, stdio: "ignore" });
+    mkdirSync(join(base, ".gsd", "milestones", "M001"), { recursive: true });
+    writeFileSync(
+      join(base, ".gsd", "milestones", "M001", "M001-SUMMARY.md"),
+      [
+        "---",
+        "id: M001",
+        "status: complete",
+        "---",
+        "",
+        "# M001: Success",
+      ].join("\n"),
+    );
+    mkdirSync(join(base, "src"), { recursive: true });
+    writeFileSync(join(base, "src", "app.ts"), "console.log('hello');");
+    execFileSync("git", ["add", "."], { cwd: base, stdio: "ignore" });
+    execFileSync("git", ["commit", "-m", "feat: implementation with stale db"], { cwd: base, stdio: "ignore" });
+
+    openDatabase(join(base, ".gsd", "gsd.db"));
+    insertMilestone({ id: "M001", title: "Milestone One", status: "active" });
+
+    const result = verifyExpectedArtifact("complete-milestone", "M001", base);
+    assert.equal(result, true, "canonical success SUMMARY should pass verification during transient DB lag");
+  } finally {
+    cleanup(base);
+  }
+});
+
 test("verifyExpectedArtifact checks pending gate-evaluate artifacts without ESM require failures", () => {
   const base = makeTmpProject();
 
@@ -775,6 +848,55 @@ test("#4414: parallel-research sentinel path does not collide with RESEARCH suff
   }
 });
 
+test("#4068: verifyExpectedArtifact parallel-research treats PARALLEL-BLOCKER as terminal completion", () => {
+  // Regression: when a parallel-research unit times out and the timeout-recovery
+  // machinery writes a PARALLEL-BLOCKER placeholder, verifyExpectedArtifact must
+  // return true so the dispatch loop can advance.  Previously it only returned
+  // true when every slice had a RESEARCH file — meaning a timeout always left
+  // verifyExpectedArtifact returning false, the unit was never cleared from
+  // unitDispatchCount, and the dispatch rule re-fired on the next iteration
+  // (infinite loop, issue #4068 / #4355).
+  const base = makeTmpBase();
+  try {
+    // Write a minimal roadmap
+    writeFileSync(
+      join(base, ".gsd", "milestones", "M001", "M001-ROADMAP.md"),
+      [
+        "# M001: Timeout Test",
+        "",
+        "## Slices",
+        "",
+        "- [ ] **S01: Alpha** `risk:low` `depends:[]`",
+        "- [ ] **S02: Beta** `risk:low` `depends:[]`",
+        "",
+      ].join("\n"),
+      "utf-8",
+    );
+
+    // No RESEARCH files written — subagents timed out
+    clearParseCache();
+    invalidateAllCaches();
+
+    // Simulate timeout-recovery writing the PARALLEL-BLOCKER placeholder
+    const blockerPath = resolveExpectedArtifactPath("research-slice", "M001/parallel-research", base);
+    assert.ok(blockerPath, "PARALLEL-BLOCKER path must resolve for parallel-research unit");
+    writeFileSync(blockerPath!, "# BLOCKER — timeout recovery\n\n**Reason**: hard timeout.\n", "utf-8");
+
+    clearParseCache();
+    invalidateAllCaches();
+
+    // After blocker is written, verifyExpectedArtifact must return true
+    // so the dispatch loop treats this unit as complete and moves on.
+    assert.equal(
+      verifyExpectedArtifact("research-slice", "M001/parallel-research", base),
+      true,
+      "#4068: PARALLEL-BLOCKER on disk must satisfy verifyExpectedArtifact so the loop does not re-dispatch",
+    );
+  } finally {
+    cleanup(base);
+  }
+});
+
 test("#4414: verifyExpectedArtifact parallel-research succeeds when all research-ready slices have RESEARCH", () => {
   const base = makeTmpBase();
   try {

package/src/resources/extensions/gsd/tests/auto-remediate-slice-status.test.ts

@@ -15,6 +15,7 @@ import assert from 'node:assert/strict';
 import { readFileSync } from 'node:fs';
 import { fileURLToPath } from 'node:url';
 import { dirname, join } from 'node:path';
+import { extractSourceRegion } from "./test-helpers.ts";
 
 const __filename = fileURLToPath(import.meta.url);
 const __dirname = dirname(__filename);
@@ -43,7 +44,9 @@ describe('auto-remediate stale slice status (#3673)', () => {
     assert.match(before, /try\s*\{/,
       'updateSliceStatus should be inside a try block');
 
-    const after = source.slice(updateIdx, updateIdx + 300);
+    // Bound the region to stop before the rogue fallback so /catch/ only
+    // matches this try block's catch, not an unrelated later one.
+    const after = extractSourceRegion(source, 'updateSliceStatus(mid, sid', 'rogues.push({');
     assert.match(after, /catch/,
       'try block should have a catch for fallback');
   });