vocero-client/frontend-s2s/src/realtime/controller.test.ts

import { afterEach, beforeEach, describe, expect, it, vi, type Mock } from "vitest";

import type { ClientOptions } from "./client";
import { VoiceController } from "./controller";

const { MockClient, closeContext } = vi.hoisted(() => {
  const closeContext = vi.fn().mockResolvedValue(undefined);
  class MockClient {
    static instances: MockClient[] = [];
    readonly options: ClientOptions;
    private listeners = new Map<string, Set<(value: unknown) => void>>();
    constructor(options: ClientOptions) {
      this.options = options;
      MockClient.instances.push(this);
    }
    on(name: string, listener: (value: unknown) => void) {
      const listeners = this.listeners.get(name) ?? new Set();
      listeners.add(listener);
      this.listeners.set(name, listeners);
      return () => {
        listeners.delete(listener);
      };
    }
    emit(name: string, value: unknown) {
      this.listeners.get(name)?.forEach((listener) => {
        listener(value);
      });
    }
    connect = vi.fn(() => {
      this.emit("status", "listening");
      return Promise.resolve();
    });
    close = vi.fn(() => {
      this.emit("status", "closed");
      return Promise.resolve();
    });
    updateSession = vi.fn();
    setGate = vi.fn();
    setMuted = vi.fn();
    renderExternalAnswer = vi.fn();
    sendToolOutput = vi.fn();
    join = vi.fn();
  }
  return { MockClient, closeContext };
});
vi.mock("./client", async (importOriginal) => ({ ...(await importOriginal()), RealtimeClient: MockClient }));
vi.mock("../audio/engine", () => ({ createAudioContext: () => ({ state: "running", close: closeContext }) }));
let pendingReasoning: ((value: Response) => void) | undefined;
let controller: VoiceController;
let fetchMock: Mock<(path: string, init?: RequestInit) => Promise<Response>>;
beforeEach(() => {
  vi.useFakeTimers();
  vi.clearAllMocks();
  MockClient.instances = [];
  pendingReasoning = undefined;
  vi.stubGlobal("AudioContext", vi.fn());
  vi.stubGlobal("navigator", { mediaDevices: { getUserMedia: vi.fn() } });
  fetchMock = vi.fn((path: string) => {
    if (path === "/api/reasoning") {
      return new Promise<Response>((resolve) => {
        pendingReasoning = resolve;
      });
    }
    const pathname = new URL(path, "http://localhost").pathname;
    const responses: Record<string, unknown> = {
      "/api/config": { reasoningType: "NODE_RED", injectProfile: false, s2sUrl: "ws://localhost/v1/realtime" },
      "/api/me": { enabled: false },
      "/api/user/phone": { result: { name: "Ana" } },
      "/api/user/reports": { count: 0 },
    };
    return Promise.resolve(new Response(JSON.stringify(responses[pathname] ?? {})));
  });
  vi.stubGlobal("fetch", fetchMock);
  controller = new VoiceController({ phoneNumber: "1234", episodeId: "episode1" });
});
afterEach(async () => {
  await controller.stop();
  vi.useRealTimers();
  vi.unstubAllGlobals();
});
async function start() {
  await controller.initialize(new AbortController().signal);
  await controller.start();
  const client = MockClient.instances[0];
  if (!client) {
    throw new Error("No client created");
  }
  return client;
}
describe("voice application lifecycle", () => {
  it("keeps LOG_ONLY reasoning on the native voice path and preserves separate profile context", async () => {
    const client = await start();
    client.emit("transcript", { id: "u1", role: "user", text: "Hola", partial: false });
    pendingReasoning?.(new Response(JSON.stringify({ log_only: true, answer: "" })));
    await vi.advanceTimersByTimeAsync(1);
    const request = fetchMock.mock.calls.find(([path]) => path === "/api/reasoning");
    expect(request?.[1]?.body).toContain('"phoneNumber":"1234"');
    expect(request?.[1]?.body).toContain("Caller profile resolved");
    expect(client.options.instructions).not.toContain("Caller profile resolved");
    expect(client.renderExternalAnswer).not.toHaveBeenCalled();
    expect(controller.getSnapshot().messages).toHaveLength(1);
  });
  it("discards external reasoning that finishes after the session was stopped", async () => {
    const client = await start();
    client.emit("transcript", { id: "u1", role: "user", text: "Consulta", partial: false });
    await controller.stop();
    pendingReasoning?.(new Response(JSON.stringify({ answer: "Respuesta tardía" })));
    await vi.advanceTimersByTimeAsync(1);
    expect(client.renderExternalAnswer).not.toHaveBeenCalled();
    expect(controller.getSnapshot().messages.map((message) => message.text)).toEqual(["Consulta"]);
  });
  it("applies and removes RAG context in the live session", async () => {
    const client = await start();
    controller.setRag("Datos importantes");
    expect(client.updateSession).toHaveBeenLastCalledWith(expect.stringContaining("Datos importantes"), "Aiden", []);
    controller.setRag("");
    expect(client.updateSession).toHaveBeenLastCalledWith(expect.not.stringContaining("CUSTOM_RAG_BEGIN"), "Aiden", []);
  });
  it("closes after silence but does not cut off the assistant while speaking", async () => {
    const client = await start();
    client.emit("status", "ai-speaking");
    await vi.advanceTimersByTimeAsync(60000);
    expect(client.close).not.toHaveBeenCalled();
    client.emit("status", "listening");
    await vi.advanceTimersByTimeAsync(30000);
    expect(client.close).toHaveBeenCalledOnce();
    expect(controller.getSnapshot().status).toBe("closed");
  });
});