@vellumai/assistant 0.4.44 → 0.4.45

package/src/__tests__/recording-handler.test.ts

@@ -32,11 +32,8 @@ mock.module("../config/loader.js", () => ({
     contextWindow: {
       enabled: true,
       maxInputTokens: 180000,
-      targetInputTokens: 110000,
-      compactThreshold: 0.8,
-      preserveRecentUserTurns: 8,
-      summaryMaxTokens: 1200,
-      chunkTokens: 12000,
+      targetBudgetRatio: 0.30,
+      compactThreshold: 0.8,      summaryBudgetRatio: 0.05,
     },
   }),
   invalidateConfigCache: noop,
@@ -153,8 +150,8 @@ mock.module("node:fs", () => {
 import {
   __resetRecordingState,
   handleRecordingStart,
+  handleRecordingStatusCore,
   handleRecordingStop,
-  recordingHandlers,
 } from "../daemon/handlers/recording.js";
 import type { HandlerContext } from "../daemon/handlers/shared.js";
 import type { RecordingStatus } from "../daemon/message-types/computer-use.js";
@@ -206,15 +203,11 @@ describe("handleRecordingStart", () => {
     mockFileSize = 1024;
   });
 
-  test("sends recording_start IPC and returns a UUID", () => {
+  test("sends recording_start event and returns a UUID", () => {
     const { ctx, sent } = createCtx();
     const conversationId = "conv-1";
 
-    const recordingId = handleRecordingStart(
-      conversationId,
-      undefined,
-      ctx,
-    );
+    const recordingId = handleRecordingStart(conversationId, undefined, ctx);
 
     expect(recordingId).not.toBeNull();
     // UUID v4 format
@@ -256,19 +249,11 @@ describe("handleRecordingStart", () => {
   test("returns null when a different conversation already has an active recording (global guard)", () => {
     const { ctx, sent } = createCtx();
 
-    const id1 = handleRecordingStart(
-      "conv-global-a",
-      undefined,
-      ctx,
-    );
+    const id1 = handleRecordingStart("conv-global-a", undefined, ctx);
     expect(id1).toBeTruthy();
 
     // A second start from a different conversation should be rejected
-    const id2 = handleRecordingStart(
-      "conv-global-b",
-      undefined,
-      ctx,
-    );
+    const id2 = handleRecordingStart("conv-global-b", undefined, ctx);
     expect(id2).toBeNull();
 
     // Only the first call sends recording_start
@@ -294,11 +279,7 @@ describe("handleRecordingStop", () => {
     const conversationId = "conv-stop-1";
 
     // Start a recording first
-    const recordingId = handleRecordingStart(
-      conversationId,
-      undefined,
-      ctx,
-    );
+    const recordingId = handleRecordingStart(conversationId, undefined, ctx);
     expect(recordingId).not.toBeNull();
     sent.length = 0; // Clear the start message
 
@@ -353,7 +334,7 @@ describe("handleRecordingStop", () => {
   });
 });
 
-describe("recordingHandlers.recording_status", () => {
+describe("handleRecordingStatusCore", () => {
   beforeEach(() => {
     __resetRecordingState();
     mockMessages.length = 0;
@@ -368,11 +349,7 @@ describe("recordingHandlers.recording_status", () => {
     const { ctx } = createCtx();
     const conversationId = "conv-status-1";
 
-    const recordingId = handleRecordingStart(
-      conversationId,
-      undefined,
-      ctx,
-    );
+    const recordingId = handleRecordingStart(conversationId, undefined, ctx);
     expect(recordingId).not.toBeNull();
 
     const statusMsg: RecordingStatus = {
@@ -382,7 +359,7 @@ describe("recordingHandlers.recording_status", () => {
     };
 
     // Should not throw
-    await recordingHandlers.recording_status(statusMsg, ctx);
+    await handleRecordingStatusCore(statusMsg, ctx);
   });
 
   test("handles stopped status with file — creates attachment and notifies client", async () => {
@@ -391,11 +368,7 @@ describe("recordingHandlers.recording_status", () => {
 
     // Bind socket
 
-    const recordingId = handleRecordingStart(
-      conversationId,
-      undefined,
-      ctx,
-    );
+    const recordingId = handleRecordingStart(conversationId, undefined, ctx);
     expect(recordingId).not.toBeNull();
     sent.length = 0;
 
@@ -414,7 +387,7 @@ describe("recordingHandlers.recording_status", () => {
       durationMs: 5000,
     };
 
-    await recordingHandlers.recording_status(statusMsg, ctx);
+    await handleRecordingStatusCore(statusMsg, ctx);
 
     // Should have sent assistant_text_delta and message_complete
     const textDeltas = sent.filter((m) => m.type === "assistant_text_delta");
@@ -442,12 +415,7 @@ describe("recordingHandlers.recording_status", () => {
     const { ctx, sent } = createCtx();
     const conversationId = "conv-status-no-msg";
 
-
-    const recordingId = handleRecordingStart(
-      conversationId,
-      undefined,
-      ctx,
-    );
+    const recordingId = handleRecordingStart(conversationId, undefined, ctx);
     expect(recordingId).not.toBeNull();
     sent.length = 0;
 
@@ -461,7 +429,7 @@ describe("recordingHandlers.recording_status", () => {
       durationMs: 3000,
     };
 
-    await recordingHandlers.recording_status(statusMsg, ctx);
+    await handleRecordingStatusCore(statusMsg, ctx);
 
     // An assistant message should have been created via addMessage mock
     expect(mockMessages.length).toBeGreaterThanOrEqual(1);
@@ -475,11 +443,7 @@ describe("recordingHandlers.recording_status", () => {
 
     mockFileExists = false;
 
-    const recordingId = handleRecordingStart(
-      conversationId,
-      undefined,
-      ctx,
-    );
+    const recordingId = handleRecordingStart(conversationId, undefined, ctx);
     expect(recordingId).not.toBeNull();
     sent.length = 0;
 
@@ -492,7 +456,7 @@ describe("recordingHandlers.recording_status", () => {
     };
 
     // Should not throw — the handler logs the error and notifies the client
-    await recordingHandlers.recording_status(statusMsg, ctx);
+    await handleRecordingStatusCore(statusMsg, ctx);
 
     // No attachment should have been created
     expect(mockAttachments.length).toBe(0);
@@ -514,11 +478,7 @@ describe("recordingHandlers.recording_status", () => {
     mockFileExists = true;
     mockFileSize = 0;
 
-    const recordingId = handleRecordingStart(
-      conversationId,
-      undefined,
-      ctx,
-    );
+    const recordingId = handleRecordingStart(conversationId, undefined, ctx);
     expect(recordingId).not.toBeNull();
     sent.length = 0;
 
@@ -530,7 +490,7 @@ describe("recordingHandlers.recording_status", () => {
       durationMs: 2000,
     };
 
-    await recordingHandlers.recording_status(statusMsg, ctx);
+    await handleRecordingStatusCore(statusMsg, ctx);
 
     // No attachment should have been created for a zero-length file
     expect(mockAttachments.length).toBe(0);
@@ -559,11 +519,7 @@ describe("recordingHandlers.recording_status", () => {
     mockFileExists = true;
     mockFileSize = 4096;
 
-    const recordingId = handleRecordingStart(
-      conversationId,
-      undefined,
-      ctx,
-    );
+    const recordingId = handleRecordingStart(conversationId, undefined, ctx);
     expect(recordingId).not.toBeNull();
     sent.length = 0;
 
@@ -575,7 +531,7 @@ describe("recordingHandlers.recording_status", () => {
       durationMs: 5000,
     };
 
-    await recordingHandlers.recording_status(statusMsg, ctx);
+    await handleRecordingStatusCore(statusMsg, ctx);
 
     // Attachment should have been created
     expect(mockAttachments.length).toBe(1);
@@ -600,11 +556,7 @@ describe("recordingHandlers.recording_status", () => {
     mockFileExists = true;
     mockFileSize = 4096;
 
-    const recordingId = handleRecordingStart(
-      conversationId,
-      undefined,
-      ctx,
-    );
+    const recordingId = handleRecordingStart(conversationId, undefined, ctx);
     expect(recordingId).not.toBeNull();
     sent.length = 0;
 
@@ -616,7 +568,7 @@ describe("recordingHandlers.recording_status", () => {
       durationMs: 5000,
     };
 
-    await recordingHandlers.recording_status(statusMsg, ctx);
+    await handleRecordingStatusCore(statusMsg, ctx);
 
     // No attachment should have been created — path is outside allowlist
     expect(mockAttachments.length).toBe(0);
@@ -637,12 +589,7 @@ describe("recordingHandlers.recording_status", () => {
     const { ctx, sent } = createCtx();
     const conversationId = "conv-status-fail-final";
 
-
-    const recordingId = handleRecordingStart(
-      conversationId,
-      undefined,
-      ctx,
-    );
+    const recordingId = handleRecordingStart(conversationId, undefined, ctx);
     expect(recordingId).not.toBeNull();
     sent.length = 0;
 
@@ -654,7 +601,7 @@ describe("recordingHandlers.recording_status", () => {
       error: "Video writer finished with non-completed status 3",
     };
 
-    await recordingHandlers.recording_status(statusMsg, ctx);
+    await handleRecordingStatusCore(statusMsg, ctx);
 
     // No attachment should have been created
     expect(mockAttachments.length).toBe(0);
@@ -676,12 +623,7 @@ describe("recordingHandlers.recording_status", () => {
     const { ctx, sent } = createCtx();
     const conversationId = "conv-status-failed";
 
-
-    const recordingId = handleRecordingStart(
-      conversationId,
-      undefined,
-      ctx,
-    );
+    const recordingId = handleRecordingStart(conversationId, undefined, ctx);
     expect(recordingId).not.toBeNull();
     sent.length = 0;
 
@@ -692,7 +634,7 @@ describe("recordingHandlers.recording_status", () => {
       error: "Permission denied",
     };
 
-    await recordingHandlers.recording_status(statusMsg, ctx);
+    await handleRecordingStatusCore(statusMsg, ctx);
 
     // Should send error notification
     const textDeltas = sent.filter((m) => m.type === "assistant_text_delta");
@@ -708,12 +650,7 @@ describe("recordingHandlers.recording_status", () => {
     const { ctx, sent } = createCtx();
     const conversationId = "conv-status-failed-no-err";
 
-
-    const recordingId = handleRecordingStart(
-      conversationId,
-      undefined,
-      ctx,
-    );
+    const recordingId = handleRecordingStart(conversationId, undefined, ctx);
     expect(recordingId).not.toBeNull();
     sent.length = 0;
 
@@ -723,7 +660,7 @@ describe("recordingHandlers.recording_status", () => {
       status: "failed",
     };
 
-    await recordingHandlers.recording_status(statusMsg, ctx);
+    await handleRecordingStatusCore(statusMsg, ctx);
 
     const textDeltas = sent.filter((m) => m.type === "assistant_text_delta");
     expect(textDeltas.length).toBeGreaterThanOrEqual(1);
@@ -734,7 +671,6 @@ describe("recordingHandlers.recording_status", () => {
     const { ctx, sent } = createCtx();
     const conversationId = "conv-fallback";
 
-
     // Send a recording_status directly with attachToConversationId
     // without having started a recording through handleRecordingStart
     const statusMsg: RecordingStatus = {
@@ -746,7 +682,7 @@ describe("recordingHandlers.recording_status", () => {
     };
 
     // Should not throw — uses attachToConversationId as fallback
-    await recordingHandlers.recording_status(statusMsg, ctx);
+    await handleRecordingStatusCore(statusMsg, ctx);
 
     const textDeltas = sent.filter((m) => m.type === "assistant_text_delta");
     expect(textDeltas.length).toBeGreaterThanOrEqual(1);

package/src/__tests__/registry.test.ts

@@ -128,33 +128,24 @@ describe("tool registry host tools", () => {
 });
 
 describe("tool registry dynamic-tools tools", () => {
-  test("registers scaffold, delete, and skill_load tools", async () => {
+  test("registers skill_load tool", async () => {
     await initializeTools();
 
-    const dynamicToolNames = [
-      "scaffold_managed_skill",
-      "delete_managed_skill",
-      "skill_load",
-    ] as const;
-
-    for (const toolName of dynamicToolNames) {
-      const tool = getTool(toolName);
-      expect(tool).toBeDefined();
-    }
+    const tool = getTool("skill_load");
+    expect(tool).toBeDefined();
 
     const definitionNames = getAllToolDefinitions().map((def) => def.name);
-    for (const toolName of dynamicToolNames) {
-      expect(definitionNames).toContain(toolName);
-    }
+    expect(definitionNames).toContain("skill_load");
   });
 
-  test("scaffold and delete are registered as High risk", async () => {
+  test("scaffold and delete are NOT in the core tool registry (moved to bundled skill)", async () => {
     await initializeTools();
-    for (const name of ["scaffold_managed_skill", "delete_managed_skill"]) {
-      const tool = getTool(name);
-      expect(tool).toBeDefined();
-      expect(tool?.defaultRiskLevel).toBe(RiskLevel.High);
-    }
+    // scaffold_managed_skill and delete_managed_skill moved to the
+    // skill-management bundled skill — they are no longer registered as core
+    // tools. Their High risk classification is handled by classifyRisk() in
+    // checker.ts so security behavior is preserved.
+    expect(getTool("scaffold_managed_skill")).toBeUndefined();
+    expect(getTool("delete_managed_skill")).toBeUndefined();
   });
 
   test("skill_load is registered as Low risk", async () => {
@@ -176,29 +167,30 @@ describe("tool manifest", () => {
   });
 
   test("manifest declares expected core lazy tools", () => {
-    // bash and swarm_delegate moved from lazy to eager registration
+    // bash moved from lazy to eager registration
+    // swarm_delegate moved to the orchestration bundled skill
     const lazyNames = new Set(lazyTools.map((t) => t.name));
     expect(lazyNames.has("bash")).toBe(false);
     expect(lazyNames.has("evaluate_typescript_code")).toBe(false);
     expect(lazyNames.has("claude_code")).toBe(false);
     expect(lazyNames.has("swarm_delegate")).toBe(false);
-    // Verify they are in eager tools instead
+    // bash is in eager tools; swarm_delegate is now a bundled skill tool
     expect(eagerModuleToolNames).toContain("bash");
-    expect(eagerModuleToolNames).toContain("swarm_delegate");
+    expect(eagerModuleToolNames).not.toContain("swarm_delegate");
+    expect(eagerModuleToolNames).not.toContain("version");
   });
 
   test("eager module tool names list contains expected count", () => {
-    expect(eagerModuleToolNames.length).toBe(15);
+    expect(eagerModuleToolNames.length).toBe(11);
   });
 
-  test("explicit tools list includes memory, credential, and watch tools", () => {
+  test("explicit tools list includes memory and credential tools", () => {
     const names = explicitTools.map((t) => t.name);
     expect(names).toContain("memory_recall");
-    expect(names).toContain("memory_save");
-    expect(names).toContain("memory_update");
+    expect(names).toContain("memory_manage");
     expect(names).toContain("credential_store");
-    expect(names).toContain("account_manage");
-    expect(names).toContain("start_screen_watch");
+    // start_screen_watch moved to the screen-watch bundled skill
+    expect(names).not.toContain("start_screen_watch");
   });
 
   test("registered tool count is at least eager + lazy + host", async () => {
@@ -211,6 +203,11 @@ describe("tool manifest", () => {
 });
 
 describe("baseline characterization: hardcoded tool loading", () => {
+  test("version is NOT registered in the global registry after initializeTools()", async () => {
+    await initializeTools();
+    expect(getTool("version")).toBeUndefined();
+  });
+
   test("gmail tools are NOT registered in the global registry after initializeTools()", async () => {
     await initializeTools();
     const allTools = getAllTools();
@@ -223,9 +220,7 @@ describe("baseline characterization: hardcoded tool loading", () => {
       "gmail_mark_read",
       "gmail_draft",
       "gmail_archive",
-      "gmail_batch_archive",
       "gmail_label",
-      "gmail_batch_label",
       "gmail_trash",
       "gmail_send",
       "gmail_unsubscribe",
@@ -243,9 +238,7 @@ describe("baseline characterization: hardcoded tool loading", () => {
       "gmail_mark_read",
       "gmail_draft",
       "gmail_archive",
-      "gmail_batch_archive",
       "gmail_label",
-      "gmail_batch_label",
       "gmail_trash",
       "gmail_send",
       "gmail_unsubscribe",
@@ -517,7 +510,7 @@ describe("computer-use registration split", () => {
   // Start each test from a completely empty registry so assertions are
   // non-vacuous — the split functions must actually register tools.
 
-  test("registerComputerUseActionTools registers all 12 CU action tools and nothing else", async () => {
+  test("registerComputerUseActionTools registers all 10 CU action tools and nothing else", async () => {
     const { registerComputerUseActionTools } =
       await import("../tools/computer-use/registry.js");
 
@@ -527,7 +520,7 @@ describe("computer-use registration split", () => {
     registerComputerUseActionTools();
 
     const registered = getAllTools();
-    expect(registered).toHaveLength(12);
+    expect(registered).toHaveLength(10);
     expect(registered.every((t) => t.name.startsWith("computer_use_"))).toBe(
       true,
     );

package/src/__tests__/relay-server.test.ts

@@ -40,7 +40,6 @@ mock.module("../util/platform.js", () => ({
   isMacOS: () => process.platform === "darwin",
   isLinux: () => process.platform === "linux",
   isWindows: () => process.platform === "win32",
-  getSocketPath: () => join(testDir, "test.sock"),
   getPidPath: () => join(testDir, "test.pid"),
   getDbPath: () => join(testDir, "test.db"),
   getLogPath: () => join(testDir, "test.log"),

package/src/__tests__/ride-shotgun-handler.test.ts

@@ -73,10 +73,6 @@ mock.module("../tools/browser/auto-navigate.js", () => ({
   autoNavigate: async () => [],
 }));
 
-mock.module("../tools/browser/x-auto-navigate.js", () => ({
-  navigateXPages: async () => [],
-}));
-
 mock.module("../util/logger.js", () => ({
   getLogger: () => ({
     info: () => {},
@@ -248,10 +244,7 @@ describe("ride-shotgun-handler", () => {
 
     // Find the session and stop it
     const watchId = [...watchSessions.keys()][0]!;
-    await handleRideShotgunStop(
-      { type: "ride_shotgun_stop", watchId },
-      ctx,
-    );
+    await handleRideShotgunStop({ type: "ride_shotgun_stop", watchId }, ctx);
 
     expect(mockMinimizeCalled).toBe(true);
     expect(mockMinimizeBaseUrl).toBe("http://localhost:9222");
@@ -282,10 +275,7 @@ describe("ride-shotgun-handler", () => {
 
     // Find the session and stop it
     const watchId = [...watchSessions.keys()][0]!;
-    await handleRideShotgunStop(
-      { type: "ride_shotgun_stop", watchId },
-      ctx,
-    );
+    await handleRideShotgunStop({ type: "ride_shotgun_stop", watchId }, ctx);
 
     expect(mockMinimizeCalled).toBe(false);
   });
@@ -311,10 +301,7 @@ describe("ride-shotgun-handler", () => {
 
     // Clean up
     const watchId = [...watchSessions.keys()][0]!;
-    await handleRideShotgunStop(
-      { type: "ride_shotgun_stop", watchId },
-      ctx,
-    );
+    await handleRideShotgunStop({ type: "ride_shotgun_stop", watchId }, ctx);
   });
 
   test("sends watch_started message with session IDs", async () => {
@@ -342,10 +329,7 @@ describe("ride-shotgun-handler", () => {
 
     // Clean up
     const watchId = startMsg.watchId;
-    await handleRideShotgunStop(
-      { type: "ride_shotgun_stop", watchId },
-      ctx,
-    );
+    await handleRideShotgunStop({ type: "ride_shotgun_stop", watchId }, ctx);
   });
 
   test("sends ride_shotgun_error when ensureChromeWithCdp fails", async () => {

package/src/__tests__/runtime-attachment-metadata.test.ts

@@ -22,7 +22,6 @@ mock.module("../util/platform.js", () => ({
   isMacOS: () => process.platform === "darwin",
   isLinux: () => process.platform === "linux",
   isWindows: () => process.platform === "win32",
-  getSocketPath: () => join(testDir, "test.sock"),
   getPidPath: () => join(testDir, "test.pid"),
   getDbPath: () => join(testDir, "test.db"),
   getLogPath: () => join(testDir, "test.log"),

package/src/__tests__/runtime-events-sse-parity.test.ts

@@ -1,7 +1,7 @@
 /**
- * IPC parity tests for the SSE assistant-events endpoint.
+ * HTTP parity tests for the SSE assistant-events endpoint.
  *
- * Asserts that every streaming/delta IPC ServerMessage type is preserved
+ * Asserts that every streaming/delta ServerMessage type is preserved
  * exactly — field-for-field — when delivered through the SSE route.
  *
  * Message types covered:
@@ -30,7 +30,6 @@ mock.module("../util/platform.js", () => ({
   isMacOS: () => process.platform === "darwin",
   isLinux: () => process.platform === "linux",
   isWindows: () => process.platform === "win32",
-  getSocketPath: () => join(testDir, "test.sock"),
   getPidPath: () => join(testDir, "test.pid"),
   getDbPath: () => join(testDir, "test.db"),
   getLogPath: () => join(testDir, "test.log"),
@@ -125,7 +124,7 @@ async function publishAndReadFrame(
 // Tests
 // ---------------------------------------------------------------------------
 
-describe("SSE IPC parity — streaming/delta message types", () => {
+describe("SSE HTTP parity — streaming/delta message types", () => {
   beforeEach(() => {
     const db = getDb();
     db.run("DELETE FROM conversation_keys");

package/src/__tests__/runtime-events-sse.test.ts

@@ -21,7 +21,6 @@ mock.module("../util/platform.js", () => ({
   isMacOS: () => process.platform === "darwin",
   isLinux: () => process.platform === "linux",
   isWindows: () => process.platform === "win32",
-  getSocketPath: () => join(testDir, "test.sock"),
   getPidPath: () => join(testDir, "test.pid"),
   getDbPath: () => join(testDir, "test.db"),
   getLogPath: () => join(testDir, "test.log"),

package/src/__tests__/sandbox-diagnostics.test.ts

@@ -19,7 +19,6 @@ mock.module("../util/platform.js", () => ({
   isLinux: () => mockIsLinux,
   getRootDir: () => "/tmp/vellum-test",
   getDataDir: () => "/tmp/vellum-test/data",
-  getSocketPath: () => "/tmp/vellum-test/vellum.sock",
   getDbPath: () => "/tmp/vellum-test/data/db/assistant.db",
   getLogPath: () => "/tmp/vellum-test/data/logs/daemon.log",
   getSandboxRootDir: () => "/tmp/vellum-test/sandbox",