@vellumai/assistant 0.10.7 → 0.10.8-dev.202607102228.5945895

package/src/__tests__/contact-store-user-file.test.ts

@@ -1,15 +1,8 @@
 import { Database } from "bun:sqlite";
-import { beforeEach, describe, expect, mock, test } from "bun:test";
+import { beforeEach, describe, expect, test } from "bun:test";
 
 import { drizzle } from "drizzle-orm/bun-sqlite";
 
-mock.module("../util/logger.js", () => ({
-  getLogger: () =>
-    new Proxy({} as Record<string, unknown>, {
-      get: () => () => {},
-    }),
-}));
-
 import {
   generateUserFileSlug,
   upsertContact,
@@ -57,7 +50,6 @@ describe("upsertContact user_file selection", () => {
   test("assigns a fresh slug per contact; principalId is gateway-owned and no longer groups siblings locally", () => {
     const primary = upsertContact({
       displayName: "Chris",
-      role: "guardian",
       channels: [
         {
           type: "vellum",
@@ -72,7 +64,6 @@ describe("upsertContact user_file selection", () => {
     // (collision-incremented) slug. Sibling grouping is owned by the gateway.
     const slack = upsertContact({
       displayName: "chris",
-      role: "guardian",
       channels: [
         {
           type: "slack",
@@ -88,7 +79,6 @@ describe("upsertContact user_file selection", () => {
   test("generates a slug from the display name for a brand-new contact", () => {
     const contact = upsertContact({
       displayName: "Alice",
-      role: "contact",
       channels: [
         {
           type: "slack",
@@ -103,7 +93,6 @@ describe("upsertContact user_file selection", () => {
   test("still auto-increments when principalId is not set and displayName collides", () => {
     const first = upsertContact({
       displayName: "Bob",
-      role: "contact",
       channels: [
         {
           type: "slack",
@@ -114,7 +103,6 @@ describe("upsertContact user_file selection", () => {
     });
     const second = upsertContact({
       displayName: "Bob",
-      role: "contact",
       channels: [
         {
           type: "slack",
@@ -137,7 +125,6 @@ describe("upsertContact user_file selection", () => {
 
     const contact = upsertContact({
       displayName: "Legacy",
-      role: "guardian",
       channels: [
         {
           type: "phone",

package/src/__tests__/contacts-tools.test.ts

@@ -11,13 +11,6 @@ import {
 // Track the gateway URL; updated once the test server starts.
 let testGatewayUrl = "http://127.0.0.1:0";
 
-mock.module("../util/logger.js", () => ({
-  getLogger: () =>
-    new Proxy({} as Record<string, unknown>, {
-      get: () => () => {},
-    }),
-}));
-
 mock.module("../config/loader.js", () => ({
   getConfig: () => ({
     ui: {},

package/src/__tests__/contacts-write.test.ts

@@ -10,14 +10,7 @@
 
 import { existsSync } from "node:fs";
 import { join } from "node:path";
-import { beforeEach, describe, expect, mock, test } from "bun:test";
-
-mock.module("../util/logger.js", () => ({
-  getLogger: () =>
-    new Proxy({} as Record<string, unknown>, {
-      get: () => () => {},
-    }),
-}));
+import { beforeEach, describe, expect, test } from "bun:test";
 
 import { upsertContactChannel } from "../contacts/contacts-write.js";
 import { getSqlite } from "../persistence/db-connection.js";

package/src/__tests__/context-search-memory-source.test.ts

@@ -398,7 +398,6 @@ function makeNode(overrides: Partial<MemoryNode> = {}): MemoryNode {
     narrativeRole: null,
     partOfStory: null,
     imageRefs: null,
-    scopeId: "scope-default",
     ...overrides,
   };
 }

package/src/__tests__/context-search-memory-v2-source.test.ts

@@ -28,12 +28,6 @@ import type { AssistantConfig } from "../config/schema.js";
 import type { RecallSearchContext } from "../plugins/defaults/memory/context-search/types.js";
 import type { EdgeIndex } from "../plugins/defaults/memory/v2/edge-index.js";
 import type { ConceptPage } from "../plugins/defaults/memory/v2/types.js";
-import { makeMockLogger } from "./helpers/mock-logger.js";
-
-mock.module("../util/logger.js", () => ({
-  getLogger: () => makeMockLogger(),
-}));
-
 let denseEmbedReturn: number[] = [0.1, 0.2, 0.3];
 mock.module("../persistence/embeddings/embedding-backend.js", () => ({
   isEmbeddingDimensionAvailable: async () => true,

package/src/__tests__/context-window-manager-compact-retry.test.ts

@@ -6,27 +6,6 @@
  */
 import { beforeEach, describe, expect, mock, test } from "bun:test";
 
-function makeLoggerStub(): Record<string, unknown> {
-  const stub: Record<string, unknown> = {};
-  for (const m of [
-    "info",
-    "warn",
-    "error",
-    "debug",
-    "trace",
-    "fatal",
-    "silent",
-    "child",
-  ]) {
-    stub[m] = m === "child" ? () => makeLoggerStub() : () => {};
-  }
-  return stub;
-}
-
-mock.module("../util/logger.js", () => ({
-  getLogger: () => makeLoggerStub(),
-}));
-
 mock.module("../daemon/conversation-registry.js", () => ({
   findConversationOrSubagent: () => ({
     systemPrompt: "you are a test assistant",
@@ -59,6 +38,9 @@ interface CompactionRunResult {
 let runCalls: Array<{
   messages: unknown;
   previousEstimatedInputTokens: number;
+  force: boolean | undefined;
+  targetTokens: number | undefined;
+  fixedTailStartIndex: number | undefined;
 }> = [];
 let runResults: CompactionRunResult[] = [];
 const estimateReturns: number[] = [];
@@ -87,13 +69,20 @@ mock.module("../config/loader.js", () => ({
 }));
 
 mock.module("../context/compactor.js", () => ({
+  isSyntheticCompactionMessage: () => false,
   runAssistantDrivenCompaction: async (args: {
     messages: unknown;
     previousEstimatedInputTokens: number;
+    force: boolean | undefined;
+    targetTokens: number | undefined;
+    fixedTailStartIndex: number | undefined;
   }): Promise<CompactionRunResult> => {
     runCalls.push({
       messages: args.messages,
       previousEstimatedInputTokens: args.previousEstimatedInputTokens,
+      force: args.force,
+      targetTokens: args.targetTokens,
+      fixedTailStartIndex: args.fixedTailStartIndex,
     });
     const idx = runCalls.length - 1;
     const result = runResults[idx];
@@ -377,6 +366,114 @@ describe("ContextWindowManager.maybeCompact retry budget", () => {
   });
 });
 
+describe("ContextWindowManager.maybeCompact fixed boundary", () => {
+  beforeEach(() => {
+    runCalls = [];
+    runResults = [];
+    estimateReturns.length = 0;
+  });
+
+  test("skips threshold gating and omits the target budget", async () => {
+    /**
+     * A fixed boundary is a user request, not a fullness response: the run
+     * fires even when the estimate is far below the auto threshold and no
+     * `force` option is passed, and the compactor receives no `targetTokens`
+     * (the boundary is not a budget outcome) but does receive `force` so its
+     * own threshold gate never trips.
+     */
+    // GIVEN an estimate far below the 140k threshold, then a post-compaction
+    // recompute of 50k
+    estimateReturns.push(10_000, 50_000);
+    runResults = [
+      compactResult({
+        estimatedInputTokens: 50_000,
+        summaryText: "fixed summary",
+        compactedPersistedMessages: 5,
+      }),
+    ];
+
+    // WHEN maybeCompact runs with only a fixed boundary
+    const manager = buildManager();
+    const result = await manager.maybeCompact(makeMessages(10), undefined, {
+      fixedTailStartIndex: 4,
+    });
+
+    // THEN exactly one compactor call fired with the boundary forwarded
+    expect(runCalls.length).toBe(1);
+    expect(runCalls[0]!.fixedTailStartIndex).toBe(4);
+    expect(runCalls[0]!.targetTokens).toBeUndefined();
+    expect(runCalls[0]!.force).toBe(true);
+    expect(runCalls[0]!.previousEstimatedInputTokens).toBe(10_000);
+
+    // AND the result maps the compactor output with a recomputed estimate
+    expect(result.compacted).toBe(true);
+    expect(result.summaryText).toBe("fixed summary");
+    expect(result.compactedPersistedMessages).toBe(5);
+    expect(result.estimatedInputTokens).toBe(50_000);
+    expect(result.exhausted).toBeUndefined();
+  });
+
+  test("single attempt even when the result stays above threshold", async () => {
+    /**
+     * The retry ladder never runs on the fixed path — a retry would
+     * re-summarize past the boundary the user picked.
+     */
+    // GIVEN a pass that stays above the 140k threshold
+    estimateReturns.push(180_000, 165_000);
+    runResults = [
+      compactResult({ estimatedInputTokens: 165_000 }),
+      // Never consumed — proves no retry fired:
+      compactResult({ estimatedInputTokens: 50_000 }),
+    ];
+
+    // WHEN maybeCompact runs with a fixed boundary and retries available
+    const manager = buildManager(3);
+    const result = await manager.maybeCompact(makeMessages(10), undefined, {
+      force: true,
+      fixedTailStartIndex: 6,
+    });
+
+    // THEN exactly one attempt ran and the result is not marked exhausted
+    expect(runCalls.length).toBe(1);
+    expect(result.compacted).toBe(true);
+    expect(result.estimatedInputTokens).toBe(165_000);
+    expect(result.exhausted).toBeUndefined();
+  });
+
+  test("failed pass returns as-is with no retry", async () => {
+    /**
+     * A `summaryFailed` early return passes straight through so the caller
+     * can surface the failure — the manager never burns a second summary
+     * call against the same user-chosen boundary.
+     */
+    // GIVEN a compactor pass that failed its summary call
+    estimateReturns.push(180_000);
+    runResults = [
+      compactResult({
+        compacted: false,
+        summaryFailed: true,
+        estimatedInputTokens: 180_000,
+        compactedMessages: 0,
+        summaryCalls: 1,
+      }),
+      // Never consumed:
+      compactResult({ estimatedInputTokens: 50_000 }),
+    ];
+
+    // WHEN maybeCompact runs with a fixed boundary
+    const manager = buildManager(3);
+    const result = await manager.maybeCompact(makeMessages(10), undefined, {
+      force: true,
+      fixedTailStartIndex: 4,
+    });
+
+    // THEN the failed result passes through after a single attempt
+    expect(runCalls.length).toBe(1);
+    expect(result.compacted).toBe(false);
+    expect(result.summaryFailed).toBe(true);
+  });
+});
+
 describe("ContextWindowManager.emergencyCompact", () => {
   beforeEach(() => {
     emergencyCalls = [];

package/src/__tests__/context-window-manager-overflow-rung.test.ts

@@ -9,27 +9,6 @@
  */
 import { beforeEach, describe, expect, mock, test } from "bun:test";
 
-function makeLoggerStub(): Record<string, unknown> {
-  const stub: Record<string, unknown> = {};
-  for (const m of [
-    "info",
-    "warn",
-    "error",
-    "debug",
-    "trace",
-    "fatal",
-    "silent",
-    "child",
-  ]) {
-    stub[m] = m === "child" ? () => makeLoggerStub() : () => {};
-  }
-  return stub;
-}
-
-mock.module("../util/logger.js", () => ({
-  getLogger: () => makeLoggerStub(),
-}));
-
 mock.module("../daemon/conversation-registry.js", () => ({
   findConversationOrSubagent: () => ({
     systemPrompt: "you are a test assistant",

package/src/__tests__/conversation-abort-tool-results.test.ts

@@ -8,11 +8,6 @@ import type {
   ProviderResponse,
 } from "../providers/types.js";
 
-mock.module("../util/logger.js", () => ({
-  getLogger: () =>
-    new Proxy({} as Record<string, unknown>, { get: () => () => {} }),
-}));
-
 mock.module("../providers/registry.js", () => ({
   getProvider: () => ({ name: "mock-provider" }),
   initializeProviders: async () => {},

package/src/__tests__/conversation-agent-loop-disk-pressure.test.ts

@@ -8,13 +8,6 @@ import type { ServerMessage } from "../daemon/message-protocol.js";
 
 type Context = Conversation;
 
-mock.module("../util/logger.js", () => ({
-  getLogger: () =>
-    new Proxy({} as Record<string, unknown>, {
-      get: () => () => {},
-    }),
-}));
-
 mock.module("../config/loader.js", () => ({
   getConfig: () => ({
     llm: {
@@ -139,7 +132,6 @@ function makeCtx(overrides: Partial<Context> = {}): Conversation {
     contextCompactedAt: null,
     conversationType: "background",
     source: "memory",
-    memoryPolicy: { scopeId: "default", includeDefaultFallback: true },
     currentActiveSurfaceId: undefined,
     currentPage: undefined,
     surfaceState: new Map(),
@@ -155,10 +147,6 @@ function makeCtx(overrides: Partial<Context> = {}): Conversation {
     preactivatedSkillIds: undefined,
     skillProjectionState: new Map(),
     skillProjectionCache: new Map() as Context["skillProjectionCache"],
-    profiler: {
-      startRequest: () => {},
-      emitSummary: () => {},
-    } as unknown as Context["profiler"],
     usageStats: {
       totalInputTokens: 0,
       totalOutputTokens: 0,

package/src/__tests__/conversation-agent-loop-inference-profile.test.ts

@@ -19,8 +19,6 @@ import type { AgentLoopRunOptions } from "../agent/loop.js";
 import type { LLMCallSite } from "../config/schemas/llm.js";
 import { resetPluginRegistryAndRegisterDefaults } from "../plugins/defaults/index.js";
 import type { Message, ToolDefinition } from "../providers/types.js";
-import { makeMockLogger } from "./helpers/mock-logger.js";
-
 // Snapshot the real `conversation-crud` exports before `mock.module()` below
 // replaces them. We use a synchronous CJS-style require (via `createRequire`)
 // because static `import` is hoisted above `mock.module()` calls — but bun's
@@ -45,10 +43,6 @@ const conversationDiskViewRealSnapshot = {
 
 // ── Module mocks (must precede imports of the module under test) ─────
 
-mock.module("../util/logger.js", () => ({
-  getLogger: () => makeMockLogger(),
-}));
-
 mock.module("../config/loader.js", () => ({
   getConfig: () => ({
     llm: {
@@ -76,7 +70,12 @@ mock.module("../config/loader.js", () => ({
         },
       },
       profiles: {
+        // Complete (materialized) shape: override-or-default semantics skip
+        // profiles that don't carry their own provider+model.
         "quality-optimized": {
+          source: "user",
+          provider: "anthropic",
+          model: "claude-opus-4-8",
           contextWindow: { maxInputTokens: 50000 },
         },
       },
@@ -447,8 +446,6 @@ function makeCtx(
     contextCompactedMessageCount: 0,
     contextCompactedAt: null,
 
-    memoryPolicy: { scopeId: "default", includeDefaultFallback: true },
-
     currentActiveSurfaceId: undefined,
     currentPage: undefined,
     surfaceState: new Map(),
@@ -468,10 +465,6 @@ function makeCtx(
     skillProjectionCache:
       new Map() as unknown as Conversation["skillProjectionCache"],
 
-    profiler: {
-      startRequest: () => {},
-      emitSummary: () => {},
-    } as unknown as Conversation["profiler"],
     usageStats: {
       totalInputTokens: 0,
       totalOutputTokens: 0,

package/src/__tests__/conversation-agent-loop-overflow.test.ts

@@ -12,7 +12,24 @@
  * Tests 2, 8, 9, and 10 are now active and passing against current code.
  */
 import { createRequire } from "node:module";
-import { afterAll, beforeEach, describe, expect, mock, test } from "bun:test";
+import {
+  afterAll,
+  beforeAll,
+  beforeEach,
+  describe,
+  expect,
+  mock,
+  test,
+} from "bun:test";
+
+import { setOverridesForTesting } from "./feature-flag-test-helpers.js";
+
+// Legacy-shaped fixtures (llm.default-centric resolution): pinned to the
+// flag-off cascade. Override-or-default (flag-on) semantics are pinned by
+// llm-resolver-override-or-default.test.ts and its companion suites.
+beforeAll(() => {
+  setOverridesForTesting({ "override-or-default-resolution": false });
+});
 
 import type { LoopToolExecutor } from "../agent/loop.js";
 import type { LLMConfig } from "../config/schemas/llm.js";
@@ -54,11 +71,6 @@ mock.module("../plugins/defaults/compaction/manager-store.js", () => ({
   },
 }));
 
-mock.module("../util/logger.js", () => ({
-  getLogger: () =>
-    new Proxy({} as Record<string, unknown>, { get: () => () => {} }),
-}));
-
 const defaultLlmConfig: LLMConfig = {
   default: {
     provider: "anthropic",
@@ -163,7 +175,9 @@ const runMockReducer = async (
   cfg: unknown,
   state: unknown,
 ) => {
-  if (mockReducerStepFn) return mockReducerStepFn(msgs, cfg, state);
+  if (mockReducerStepFn) {
+    return mockReducerStepFn(msgs, cfg, state);
+  }
   return {
     messages: msgs,
     tier: "forced_compaction",
@@ -210,7 +224,9 @@ function makeOverflowLadderStub(): {
 } {
   let state: unknown;
   const reduceOverflowOneRung = async (msgs: Message[], opts: unknown) => {
-    if (!state) state = makeInitialReducerState();
+    if (!state) {
+      state = makeInitialReducerState();
+    }
     const step = (await runMockReducer(msgs, opts, state)) as {
       state: unknown;
     };
@@ -455,9 +471,15 @@ mock.module("../daemon/conversation-error.js", () => ({
     };
   },
   isUserCancellation: (err: unknown, ctx: { aborted?: boolean }) => {
-    if (!ctx.aborted) return false;
-    if (err instanceof DOMException && err.name === "AbortError") return true;
-    if (err instanceof Error && err.name === "AbortError") return true;
+    if (!ctx.aborted) {
+      return false;
+    }
+    if (err instanceof DOMException && err.name === "AbortError") {
+      return true;
+    }
+    if (err instanceof Error && err.name === "AbortError") {
+      return true;
+    }
     return false;
   },
   buildConversationErrorMessage: (
@@ -593,8 +615,6 @@ function makeCtx(
     contextCompactedMessageCount: 0,
     contextCompactedAt: null,
 
-    memoryPolicy: { scopeId: "default", includeDefaultFallback: true },
-
     currentActiveSurfaceId: undefined,
     currentPage: undefined,
     surfaceState: new Map(),
@@ -614,10 +634,6 @@ function makeCtx(
     skillProjectionCache:
       new Map() as unknown as Conversation["skillProjectionCache"],
 
-    profiler: {
-      startRequest: () => {},
-      emitSummary: () => {},
-    } as unknown as Conversation["profiler"],
     usageStats: {
       totalInputTokens: 0,
       totalOutputTokens: 0,
@@ -1262,7 +1278,9 @@ describe("session-agent-loop overflow recovery (JARVIS-110)", () => {
       mockEstimateTokens = () => {
         estimateCallCount++;
         // First call: preflight check — below budget
-        if (estimateCallCount === 1) return 100_000;
+        if (estimateCallCount === 1) {
+          return 100_000;
+        }
         // Subsequent calls: mid-loop check — above 85% threshold
         return 170_000;
       };
@@ -1360,10 +1378,16 @@ describe("session-agent-loop overflow recovery (JARVIS-110)", () => {
       mockEstimateTokens = () => {
         estimateCallCount++;
         // First call: preflight — well below budget
-        if (estimateCallCount === 1) return 50_000;
+        if (estimateCallCount === 1) {
+          return 50_000;
+        }
         // Checkpoint calls grow with each tool round
-        if (estimateCallCount === 2) return 100_000; // tool 1
-        if (estimateCallCount === 3) return 140_000; // tool 2
+        if (estimateCallCount === 2) {
+          return 100_000;
+        } // tool 1
+        if (estimateCallCount === 3) {
+          return 140_000;
+        } // tool 2
         // Tool 3: exceeds 161_500 threshold
         return 175_000;
       };