@vellumai/assistant 0.11.2 → 0.11.3-dev.202608080427.92f668a

package/src/live-voice/__tests__/live-voice-integration.test.ts

@@ -254,6 +254,9 @@ function createMultiCycleHarness(startVoiceTurn: LiveVoiceTurnStarter) {
   const session = createLiveVoiceSession(context, {
     // Credential-free harness: every leg is injected, so skip the preflight.
     resolveCredentialReadiness: null,
+    // These cycle mechanics use one discrete mic chunk per utterance. Keep
+    // the adaptive playback classifier out of their timing model.
+    echoBargeInMargin: 1,
     resolveTranscriber,
     startVoiceTurn,
     streamTtsAudio,
@@ -546,6 +549,8 @@ describe("LiveVoiceSession integration smoke harness", () => {
     let turnCount = 0;
     const session = createLiveVoiceSession(context, {
       resolveCredentialReadiness: null,
+      // This cycle mechanic uses one discrete mic chunk per utterance.
+      echoBargeInMargin: 1,
       resolveTranscriber,
       startVoiceTurn,
       streamTtsAudio,
@@ -736,9 +741,9 @@ describe("live-voice audio archiving default (JARVIS-1283)", () => {
 
   test("liveVoice.archiveAudio=true wires the default archiver through the factory", async () => {
     // The real linkers write to the DB; these are role-less inputs, so stub
-    // them (synchronously — the linkers are sync) with a valid archived result
-    // per role. The point under test is the config→default-archiver wiring, not
-    // the DB write, which the injected-archiver tests above already cover.
+    // them with a valid archived result per role. The point under test is the
+    // config→default-archiver wiring, not the DB write, which the
+    // injected-archiver tests above already cover.
     const audioInput = {
       sessionId: "session-123",
       turnId: "live-turn-1",
@@ -751,11 +756,13 @@ describe("live-voice audio archiving default (JARVIS-1283)", () => {
     const userSpy = spyOn(
       liveVoiceArchive,
       "linkLiveVoiceUserUtteranceAudioToMessage",
-    ).mockReturnValue(makeArchiveResult({ ...audioInput, role: "user" }));
+    ).mockResolvedValue(makeArchiveResult({ ...audioInput, role: "user" }));
     const assistantSpy = spyOn(
       liveVoiceArchive,
       "linkLiveVoiceAssistantResponseAudioToMessage",
-    ).mockReturnValue(makeArchiveResult({ ...audioInput, role: "assistant" }));
+    ).mockResolvedValue(
+      makeArchiveResult({ ...audioInput, role: "assistant" }),
+    );
     const originalRaw = loadRawConfig();
     saveRawConfig({ ...originalRaw, liveVoice: { archiveAudio: true } });
 

package/src/live-voice/__tests__/live-voice-metrics.test.ts

@@ -274,7 +274,7 @@ describe("LiveVoiceMetricsCollector", () => {
     expect(turn.timestamps.utteranceEndAtMs).toBe(2_000);
   });
 
-  test("accumulates endpoint decisions and records the first ack kind", () => {
+  test("accumulates endpoint decisions", () => {
     const clock = makeClock(0);
     const collector = new LiveVoiceMetricsCollector({
       sessionId: "session-ep",
@@ -295,24 +295,18 @@ describe("LiveVoiceMetricsCollector", () => {
       action: "release",
       latencyMs: 210,
     });
-    collector.markAckSpoken("turn-ep", "tool_use");
-    // First ack kind wins — the per-turn budget allows one spoken ack, so a
-    // second mark must not overwrite the recorded kind.
-    collector.markAckSpoken("turn-ep", "first_delta");
     const completed = collector.completeTurn();
 
     expect(completed).toMatchObject({
       turnId: "turn-ep",
       endpointHoldCount: 2,
       endpointDecisionMaxLatencyMs: 210,
-      ackSpoken: "tool_use",
     });
     expect(
       getLiveVoiceMetricsAggregateFields(collector.getSnapshot(), "turn-ep"),
     ).toMatchObject({
       endpointHoldCount: 2,
       endpointDecisionMaxLatencyMs: 210,
-      ackSpoken: "tool_use",
     });
   });
 
@@ -338,6 +332,70 @@ describe("LiveVoiceMetricsCollector", () => {
     expect(completed).not.toHaveProperty("ackSpoken");
   });
 
+  test("decisions without a source are attributed to the front door", () => {
+    const clock = makeClock(0);
+    const collector = new LiveVoiceMetricsCollector({
+      sessionId: "session-src-default",
+      clock: clock.now,
+    });
+
+    collector.startTurn("turn-src-default");
+    collector.markEndpointDecision("turn-src-default", {
+      action: "hold",
+      latencyMs: 900,
+    });
+    const completed = collector.completeTurn();
+
+    expect(completed).toMatchObject({
+      endpointHoldCount: 1,
+      endpointDecisionMaxLatencyMs: 900,
+      endpointDecisionSource: "front-door",
+    });
+    expect(
+      getLiveVoiceMetricsAggregateFields(
+        collector.getSnapshot(),
+        "turn-src-default",
+      ),
+    ).toMatchObject({ endpointDecisionSource: "front-door" });
+  });
+
+  test("flux decisions report their source with unchanged latency accounting", () => {
+    const clock = makeClock(0);
+    const collector = new LiveVoiceMetricsCollector({
+      sessionId: "session-src-flux",
+      clock: clock.now,
+    });
+
+    collector.startTurn("turn-src-flux");
+    collector.markEndpointDecision("turn-src-flux", {
+      action: "hold",
+      latencyMs: 40,
+      source: "provider",
+    });
+    collector.markEndpointDecision("turn-src-flux", {
+      action: "release",
+      latencyMs: 120,
+      source: "provider",
+    });
+    const completed = collector.completeTurn();
+
+    expect(completed).toMatchObject({
+      endpointHoldCount: 1,
+      endpointDecisionMaxLatencyMs: 120,
+      endpointDecisionSource: "provider",
+    });
+    expect(
+      getLiveVoiceMetricsAggregateFields(
+        collector.getSnapshot(),
+        "turn-src-flux",
+      ),
+    ).toMatchObject({
+      endpointHoldCount: 1,
+      endpointDecisionMaxLatencyMs: 120,
+      endpointDecisionSource: "provider",
+    });
+  });
+
   test("accumulates spoken progress updates on the turn and aggregate fields", () => {
     const clock = makeClock(0);
     const collector = new LiveVoiceMetricsCollector({
@@ -361,7 +419,7 @@ describe("LiveVoiceMetricsCollector", () => {
     ).toMatchObject({ progressUpdatesSpoken: 2 });
   });
 
-  test("omits endpoint, ack, and progress fields for turns that never touch the features", () => {
+  test("omits endpoint and progress fields for turns that never touch the features", () => {
     const clock = makeClock(0);
     const collector = new LiveVoiceMetricsCollector({
       sessionId: "session-off",
@@ -373,8 +431,10 @@ describe("LiveVoiceMetricsCollector", () => {
     collector.markFirstAudio();
     const completed = collector.completeTurn();
 
+    expect(completed).not.toHaveProperty("endpointCommitLatencyMs");
     expect(completed).not.toHaveProperty("endpointHoldCount");
     expect(completed).not.toHaveProperty("endpointDecisionMaxLatencyMs");
+    expect(completed).not.toHaveProperty("endpointDecisionSource");
     expect(completed).not.toHaveProperty("ackSpoken");
     expect(completed).not.toHaveProperty("progressUpdatesSpoken");
 
@@ -382,12 +442,59 @@ describe("LiveVoiceMetricsCollector", () => {
       collector.getSnapshot(),
       "turn-off",
     );
+    expect(aggregateFields).not.toHaveProperty("endpointCommitLatencyMs");
     expect(aggregateFields).not.toHaveProperty("endpointHoldCount");
     expect(aggregateFields).not.toHaveProperty("endpointDecisionMaxLatencyMs");
+    expect(aggregateFields).not.toHaveProperty("endpointDecisionSource");
     expect(aggregateFields).not.toHaveProperty("ackSpoken");
     expect(aggregateFields).not.toHaveProperty("progressUpdatesSpoken");
   });
 
+  test("markEndpointCommit records the commit latency independently of the decision fields", () => {
+    const clock = makeClock(0);
+    const collector = new LiveVoiceMetricsCollector({
+      sessionId: "session-commit",
+      conversationId: "conversation-commit",
+      clock: clock.now,
+    });
+
+    collector.startTurn("turn-commit");
+    collector.markEndpointCommit("turn-commit", 1_340);
+    const completed = collector.completeTurn();
+
+    expect(completed).toMatchObject({ endpointCommitLatencyMs: 1_340 });
+    // No decision was ever recorded, so the diagnostic trio stays absent: the
+    // comparable number does not depend on the decider having been consulted.
+    expect(completed).not.toHaveProperty("endpointDecisionMaxLatencyMs");
+    expect(
+      getLiveVoiceMetricsAggregateFields(
+        collector.getSnapshot(),
+        "turn-commit",
+      ),
+    ).toMatchObject({ endpointCommitLatencyMs: 1_340 });
+  });
+
+  test("the first commit latency wins and a non-finite one records zero", () => {
+    const clock = makeClock(0);
+    const collector = new LiveVoiceMetricsCollector({
+      sessionId: "session-commit-first",
+      clock: clock.now,
+    });
+
+    collector.startTurn("turn-first");
+    collector.markEndpointCommit("turn-first", 900);
+    collector.markEndpointCommit("turn-first", 4_000);
+    expect(collector.completeTurn()).toMatchObject({
+      endpointCommitLatencyMs: 900,
+    });
+
+    collector.startTurn("turn-nan");
+    collector.markEndpointCommit("turn-nan", Number.NaN);
+    expect(collector.completeTurn()).toMatchObject({
+      endpointCommitLatencyMs: 0,
+    });
+  });
+
   test("markBargeIn records a first-wins timestamp on the active turn", () => {
     const clock = makeClock(3_000);
     const collector = new LiveVoiceMetricsCollector({

package/src/live-voice/__tests__/live-voice-photo.test.ts

@@ -0,0 +1,100 @@
+import { describe, expect, test } from "bun:test";
+
+import {
+  createMockProvider,
+  textResponse,
+} from "../../__tests__/helpers/mock-provider.js";
+import { setConfig } from "../../__tests__/helpers/set-config.js";
+import { waitFor } from "../../__tests__/helpers/wait-for.js";
+
+setConfig("memory", { enabled: false });
+
+import type { AssistantEvent } from "../../api/index.js";
+import { Conversation } from "../../daemon/conversation.js";
+import {
+  deleteConversation,
+  setConversation,
+} from "../../daemon/conversation-registry.js";
+import {
+  getAttachmentsForMessage,
+  uploadAttachment,
+} from "../../persistence/attachments-store.js";
+import {
+  createConversation,
+  getMessages,
+} from "../../persistence/conversation-crud.js";
+import { initializeDb } from "../../persistence/db-init.js";
+import { assistantEventHub } from "../../runtime/assistant-event-hub.js";
+import { persistLiveVoicePhoto } from "../live-voice-photo.js";
+
+await initializeDb();
+
+const IMAGE_BASE64 =
+  "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mNk";
+
+describe("persistLiveVoicePhoto", () => {
+  test("persists and echoes the image with human-readable context", async () => {
+    const conversation = createConversation("Live voice photo");
+    const { provider } = createMockProvider([textResponse("")]);
+    const activeConversation = new Conversation(
+      conversation.id,
+      provider,
+      "system prompt",
+      () => {},
+      "/tmp",
+      { maxTokens: 4096 },
+    );
+    activeConversation.setTrustContext({
+      trustClass: "guardian",
+      sourceChannel: "vellum",
+    });
+    setConversation(conversation.id, activeConversation);
+    const attachment = await uploadAttachment(
+      "photo.png",
+      "image/png",
+      IMAGE_BASE64,
+    );
+    const published: AssistantEvent[] = [];
+    const subscription = assistantEventHub.subscribe({
+      type: "process",
+      filter: { conversationId: conversation.id },
+      callback: (event) => {
+        published.push(event.message);
+      },
+    });
+
+    try {
+      const result = await persistLiveVoicePhoto(
+        conversation.id,
+        attachment.id,
+      );
+      expect(result.ok).toBe(true);
+
+      await waitFor(
+        () => published.some((event) => event.type === "user_message_echo"),
+        { message: "Timed out waiting for live-voice photo echo" },
+      );
+
+      const [message] = getMessages(conversation.id);
+      if (!message) {
+        throw new Error("Live-voice photo message was not persisted");
+      }
+      expect(message.content.filter((block) => block.type === "text")).toEqual([
+        { type: "text", text: "here's a photo:" },
+      ]);
+      expect(getAttachmentsForMessage(message.id)).toHaveLength(1);
+      expect(
+        published.find((event) => event.type === "user_message_echo"),
+      ).toMatchObject({
+        type: "user_message_echo",
+        text: "here's a photo:",
+        conversationId: conversation.id,
+        messageId: message.id,
+      });
+    } finally {
+      subscription.dispose();
+      deleteConversation(conversation.id);
+      activeConversation.dispose();
+    }
+  });
+});