@vellumai/assistant 0.12.2-staging.4 → 0.12.2-staging.6

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

@@ -800,6 +800,54 @@ describe("parseLiveVoiceClientTextFrame", () => {
     });
   });
 
+  test("parses the lookFrames capability on the start frame", () => {
+    const result = validateLiveVoiceClientFrame({
+      type: "start",
+      lookFrames: true,
+      audio: { mimeType: "audio/pcm", sampleRate: 24000, channels: 1 },
+    });
+
+    expect(result.ok).toBe(true);
+    if (!result.ok) {
+      return;
+    }
+    expect(result.frame).toMatchObject({ type: "start", lookFrames: true });
+  });
+
+  test("omits lookFrames from the start frame when false", () => {
+    // False and absent mean the same thing: no look frame is coming, so the
+    // session must not wait for one.
+    const result = validateLiveVoiceClientFrame({
+      type: "start",
+      lookFrames: false,
+      audio: { mimeType: "audio/pcm", sampleRate: 24000, channels: 1 },
+    });
+
+    expect(result.ok).toBe(true);
+    if (!result.ok) {
+      return;
+    }
+    expect("lookFrames" in result.frame).toBe(false);
+  });
+
+  test("returns a typed protocol error for a non-boolean lookFrames", () => {
+    const result = validateLiveVoiceClientFrame({
+      type: "start",
+      lookFrames: 1,
+      audio: { mimeType: "audio/pcm", sampleRate: 24000, channels: 1 },
+    });
+
+    expect(result.ok).toBe(false);
+    if (result.ok) {
+      return;
+    }
+    expect(result.error).toMatchObject({
+      code: "invalid_field",
+      field: "lookFrames",
+      frameType: "start",
+    });
+  });
+
   test("returns typed protocol errors for missing audio configuration fields", () => {
     const result = validateLiveVoiceClientFrame({
       type: "start",

package/src/live-voice/__tests__/session-controls.test.ts

@@ -53,6 +53,12 @@ describe("requestedSessionControl", () => {
     });
   });
 
+  test("task stop needs no client declaration", () => {
+    expect(requestedSessionControl("Okay. [TASK:STOP]", [])).toEqual({
+      action: "task_stop",
+    });
+  });
+
   test("a client control needs the client's declaration", () => {
     expect(requestedSessionControl("Bye. [END_CALL]", [])).toBeNull();
     expect(requestedSessionControl("Bye. [END_CALL]", ["end"])).toEqual({
@@ -68,9 +74,21 @@ describe("sessionControlTeaching", () => {
     expect(teaching).toContain("[UPDATES:FEWER]");
     expect(teaching).toContain("[MUTE]");
     expect(teaching).not.toContain("[END_CALL]");
+    expect(teaching).not.toContain("[TASK:STOP]");
     expect(teaching).toContain("Never emit any other bracketed marker.");
   });
 
+  test("teaches task stop only while an unfinished task is pending", () => {
+    const teaching = sessionControlTeaching(
+      [],
+      {},
+      { unfinishedTaskPending: true },
+    );
+
+    expect(teaching).toContain("[TASK:STOP]");
+    expect(teaching).toContain("not when they interrupt with a question");
+  });
+
   test("a device that can show both asks which one a bare look means", () => {
     const teaching = sessionControlTeaching(["look_screen", "look_camera"], {});
 
@@ -105,6 +123,31 @@ describe("sessionControlTeaching", () => {
     );
   });
 
+  // A client that sends a fresh frame for every look lets the session answer
+  // it, so the reply asking for the look only acknowledges, and asking again is
+  // how a share already running gets looked at as it is now.
+  test("a client that sends look frames is taught the look is answered for it", () => {
+    const teaching = sessionControlTeaching(
+      ["look_screen", "look_camera"],
+      {},
+      { lookFrames: true },
+    );
+
+    expect(teaching).toContain("[LOOK:SCREEN]");
+    expect(teaching).toContain(
+      "Use it even when their screen is already shared with you",
+    );
+    expect(teaching).toContain("Use it even when the camera is already on");
+    expect(teaching).not.toContain("take it from their next words");
+  });
+
+  test("a client that sends no look frames keeps the next-words look", () => {
+    const teaching = sessionControlTeaching(["look_screen"], {});
+
+    expect(teaching).toContain("take it from their next words");
+    expect(teaching).not.toContain("Use it even when");
+  });
+
   test("the front-door leg keeps its verdict tokens", () => {
     expect(sessionControlTeaching(["end"], { frontDoor: true })).not.toContain(
       "Never emit any other bracketed marker.",