refactor(voice): isolate gemini live provider core

src/voice/geminiIntegration.test.mjs

@@ -0,0 +1,287 @@
+import test from 'node:test';
+import assert from 'node:assert/strict';
+import {
+  nativeIntegrationFixture,
+  assertSuccessfulZoom,
+  turnOfEventLoop,
+} from '../testSupport/geminiIntegrationFixture.mjs';
+
+const options = { concurrency: false, timeout: 5000 };
+const resultFor = (h, id) =>
+  h.events.filter(
+    (event) => event.type === 'action-result' && event.callId === id,
+  );
+const callFor = (h, id) =>
+  h.events.filter(
+    (event) => event.type === 'action-call' && event.callId === id,
+  );
+
+test(
+  'native tool reaches actual common session and camera runner without a caption',
+  options,
+  async (t) => {
+    const h = nativeIntegrationFixture(t);
+    const ws = await h.start();
+    h.activity(ws); // Native activity is input framing, not transcript admission.
+    await h.tool('ordinary', ws);
+    assertSuccessfulZoom(h, ws, 'ordinary', 1000);
+    assert.equal(h.runnerCalls.length, 1);
+    assert.equal(h.runnerCalls[0].options.callId, 'ordinary');
+    assert.equal(h.runnerCalls[0].options.signal.aborted, false);
+    assert.equal(h.runnerCalls[0].options.isCurrent(), true);
+    h.complete(ws);
+    assert.equal(h.controller.narration.state().tools, 0);
+    assert.equal(h.controller.narration.state().pendingTimers, 0);
+  },
+);
+
+for (const cancellation of ['stop', 'native interruption']) {
+  test(
+    `${cancellation} discards a stale completed action result; successor changes camera once`,
+    options,
+    async (t) => {
+      const h = nativeIntegrationFixture(t);
+      const oldSocket = await h.start();
+      h.activity(oldSocket);
+      const gate = h.holdResult('old');
+      const oldDispatch = h.tool('old', oldSocket);
+      const committedResult = await gate.ready;
+      assert.equal(committedResult.ok, true);
+      assert.equal(
+        h.state.height,
+        750,
+        'the legitimate first action committed before cancellation',
+      );
+      assert.equal(callFor(h, 'old').length, 1);
+      assert.deepEqual(resultFor(h, 'old'), []);
+      assert.deepEqual(h.outputs(oldSocket), []);
+
+      if (cancellation === 'stop') h.session.stop();
+      else
+        oldSocket.message({
+          serverContent: { interrupted: true, turnComplete: true },
+        });
+      assert.equal(h.runnerCalls[0].options.signal.aborted, true);
+      assert.equal(h.runnerCalls[0].options.isCurrent(), false);
+      await oldDispatch; // Cancellation must settle before the held result returns.
+      gate.release();
+      await turnOfEventLoop();
+      assert.equal(
+        h.state.height,
+        750,
+        'settling a cancelled result adds no camera mutation',
+      );
+      assert.equal(h.state.zooms.length, 1);
+      assert.deepEqual(resultFor(h, 'old'), []);
+      assert.deepEqual(h.outputs(oldSocket), []);
+      if (cancellation === 'native interruption')
+        assert.ok(h.events.some((event) => event.type === 'interruption'));
+
+      const successorSocket =
+        cancellation === 'stop' ? await h.start() : oldSocket;
+      h.activity(successorSocket);
+      await h.tool('successor', successorSocket);
+      assertSuccessfulZoom(h, successorSocket, 'successor', 750);
+      assert.equal(h.state.zooms.length, 2);
+      assert.equal(h.runnerCalls.length, 2);
+      assert.equal(
+        h.outputs(successorSocket).some((output) => output.id === 'old'),
+        false,
+      );
+      h.complete(successorSocket);
+    },
+  );
+}
+
+for (const phase of ['common connecting', 'adapter connecting']) {
+  test(
+    `Stop from ${phase} observer prevents acquisition and permits fresh Start`,
+    options,
+    async (t) => {
+      const h = nativeIntegrationFixture(t);
+      let stopped = false;
+      const remove = h.session.subscribe((event) => {
+        const matchingDetail =
+          phase === 'common connecting'
+            ? event.detail === undefined
+            : typeof event.detail === 'string' && event.detail.length > 0;
+        if (
+          !stopped &&
+          event.type === 'state' &&
+          event.state === 'connecting' &&
+          matchingDetail
+        ) {
+          stopped = true;
+          h.session.stop();
+        }
+      });
+      await h.session.start();
+      remove();
+      assert.equal(
+        stopped,
+        true,
+        'the chosen production observer boundary was exercised',
+      );
+      assert.deepEqual(h.acquisitions, {
+        microphone: 0,
+        audio: 0,
+        token: 0,
+        socket: 0,
+      });
+      assert.equal(h.session.state, 'idle');
+      assert.equal(h.controller.narration.state().tools, 0);
+      assert.equal(h.controller.narration.state().pendingTimers, 0);
+      assert.equal(h.clock.timers.size, 0);
+      const ws = await h.start();
+      h.activity(ws);
+      await h.tool('fresh-after-start-cancel', ws);
+      assertSuccessfulZoom(h, ws, 'fresh-after-start-cancel', 1000);
+      h.complete(ws);
+    },
+  );
+}
+
+test(
+  'Stop from common action-call observer prevents runner and leaves no unhandled rejection',
+  options,
+  async (t) => {
+    const h = nativeIntegrationFixture(t);
+    const ws = await h.start();
+    const remove = h.session.subscribe((event) => {
+      if (event.type === 'action-call' && event.callId === 'cancel-at-call')
+        h.session.stop();
+    });
+    h.activity(ws);
+    await h.tool('cancel-at-call', ws);
+    // Do not install a swallowing process rejection listener. node:test records
+    // an unhandled rejection as a failure; these checkpoints let it surface here.
+    await turnOfEventLoop();
+    await turnOfEventLoop();
+    remove();
+    assert.equal(callFor(h, 'cancel-at-call').length, 1);
+    assert.deepEqual(resultFor(h, 'cancel-at-call'), []);
+    assert.equal(h.runnerCalls.length, 0);
+    assert.equal(h.state.height, 1000);
+    assert.deepEqual(h.outputs(ws), []);
+    assert.equal(h.controller.narration.state().tools, 0);
+    assert.equal(h.controller.narration.state().pendingTimers, 0);
+    assert.equal(h.clock.timers.size, 0);
+    const successor = await h.start();
+    h.activity(successor);
+    await h.tool('after-call-cancel', successor);
+    assertSuccessfulZoom(h, successor, 'after-call-cancel', 1000);
+    h.complete(successor);
+  },
+);
+
+test(
+  'Stop from executing observer cannot recreate narration or dispatch the runner',
+  options,
+  async (t) => {
+    const h = nativeIntegrationFixture(t);
+    const ws = await h.start();
+    let stopped = false;
+    const remove = h.session.subscribe((event) => {
+      if (!stopped && event.type === 'state' && event.state === 'executing') {
+        stopped = true;
+        h.session.stop();
+      }
+    });
+    h.activity(ws);
+    await h.tool('cancel-at-executing', ws);
+    remove();
+    assert.equal(stopped, true);
+    assert.equal(h.runnerCalls.length, 0);
+    assert.deepEqual(callFor(h, 'cancel-at-executing'), []);
+    assert.deepEqual(resultFor(h, 'cancel-at-executing'), []);
+    assert.equal(h.state.height, 1000);
+    assert.deepEqual(h.outputs(ws), []);
+    assert.equal(h.controller.narration.state().tools, 0);
+    assert.equal(h.controller.narration.state().pendingTimers, 0);
+    assert.equal(h.clock.timers.size, 0);
+    assert.equal(ws.closed, true);
+    assert.ok(h.audios.every((audio) => audio.closed));
+    assert.ok(h.tracks.every((track) => track.readyState === 'ended'));
+    const before = {
+      events: h.events.length,
+      wire: ws.sent.length,
+      earcon: h.state.earconStarts,
+    };
+    h.clock.advance(60_000);
+    assert.deepEqual(
+      {
+        events: h.events.length,
+        wire: ws.sent.length,
+        earcon: h.state.earconStarts,
+      },
+      before,
+    );
+    const successor = await h.start();
+    h.activity(successor);
+    await h.tool('after-executing-cancel', successor);
+    assertSuccessfulZoom(h, successor, 'after-executing-cancel', 1000);
+    h.complete(successor);
+  },
+);
+
+test(
+  'native idle observer restart remains connecting and public Stop cancels its pending permission',
+  options,
+  async (t) => {
+    const h = nativeIntegrationFixture(t);
+    await h.start();
+    let grant;
+    const permission = new Promise((resolve) => {
+      grant = resolve;
+    });
+    h.controller.getUserMedia = () => permission;
+    const states = [];
+    let restarted = false;
+    let startup;
+    const remove = h.session.subscribe((event) => {
+      if (event.type === 'state') states.push(event.state);
+      if (!restarted && event.type === 'state' && event.state === 'idle') {
+        restarted = true;
+        startup = h.session.start();
+      }
+    });
+    h.session.stop();
+    remove();
+    assert.equal(restarted, true);
+    assert.equal(states.filter((state) => state === 'idle').length, 1);
+    assert.equal(h.session.state, 'connecting');
+    assert.equal(h.controller.status, 'connecting');
+    assert.equal(
+      h.session.isActive(),
+      true,
+      'the common button/facade must offer Stop',
+    );
+    h.session.stop();
+    assert.equal(h.session.state, 'idle');
+    assert.equal(h.controller.status, 'idle');
+    const lateTrack = {
+      readyState: 'live',
+      stopCalls: 0,
+      stop() {
+        this.stopCalls++;
+        this.readyState = 'ended';
+      },
+    };
+    h.tracks.push(lateTrack);
+    grant({ getTracks: () => [lateTrack], getAudioTracks: () => [lateTrack] });
+    await startup;
+    await turnOfEventLoop();
+    assert.equal(lateTrack.stopCalls, 1);
+    assert.equal(
+      h.requests.length,
+      1,
+      'cancelled restart must not mint another token',
+    );
+    assert.equal(
+      h.sockets.length,
+      1,
+      'cancelled restart must not open another socket',
+    );
+    assert.equal(h.session.state, 'idle');
+  },
+);