fix(tests): neutralize inherited voice TTS state

scripts/run_tests_parallel.py

@@ -34,6 +34,10 @@
     HERMES_TEST_WORKERS  Override worker count (default: os.cpu_count())
     HERMES_TEST_PATHS    Override discovery roots (colon-sep, default: 'tests')
 
+The runner intentionally neutralizes runtime voice/TTS environment flags in
+pytest subprocesses. Local interactive sessions can toggle those flags on;
+tests must never inherit them and play real audio.
+
 Exit code: 0 if every file's pytest exited 0; 1 otherwise.
 """
 
@@ -91,6 +95,22 @@
 _DURATIONS_FILE = "test_durations.json"
 
 
+# Runtime-only interactive voice flags must not leak from a developer's live
+# Hermes process into test subprocesses. If ``HERMES_VOICE_TTS=1`` is inherited,
+# gateway tests that complete fake assistant turns can call the real TTS stack
+# and play audio (for example the fixture string "partial answer complete").
+_PYTEST_CHILD_ENV_OVERRIDES = {
+    "HERMES_VOICE": "0",
+    "HERMES_VOICE_TTS": "0",
+    "HERMES_VOICE_DEBUG": "0",
+}
+
+
+def _pytest_child_env() -> dict[str, str]:
+    """Return a copy of the environment safe for pytest subprocesses."""
+    return {**os.environ, **_PYTEST_CHILD_ENV_OVERRIDES}
+
+
 def _approximately_count_tests(
     files: List[Path], repo_root: Path
 ) -> dict[Path, int]:
@@ -260,7 +280,7 @@ def _run_one_file(
         stdout=subprocess.PIPE,
         stderr=subprocess.STDOUT,
         text=True,
-        env=os.environ,
+        env=_pytest_child_env(),
         # POSIX: place the child at the head of its own process group so
         # _kill_tree can SIGKILL the group atomically.
         # Windows: this maps to CREATE_NEW_PROCESS_GROUP in CPython 3.12+;

tests/conftest.py

@@ -11,8 +11,9 @@
    CI. Code using ``Path.home() / ".hermes"`` instead of the canonical
    ``get_hermes_home()`` is a bug to fix at the callsite.)
 3. **Deterministic runtime.** TZ=UTC, LANG=C.UTF-8, PYTHONHASHSEED=0.
-4. **No HERMES_SESSION_* inheritance** — the agent's current gateway
-   session must not leak into tests.
+4. **No runtime Hermes state inheritance** — the agent's current gateway
+   session, voice/TTS toggles, and other live-session flags must not leak into
+   tests.
 
 These invariants make the local test run match CI closely. Gaps that
 remain (CPU count, xdist worker count) are addressed by the canonical
@@ -26,6 +27,17 @@
 
 import pytest
 
+# These process-wide runtime flags must be neutralized while this root
+# conftest is imported, before pytest imports any test module for collection.
+# Function-scoped fixtures are too late for collection-time module code.
+_HERMES_RUNTIME_ZERO_VARS = frozenset({
+    "HERMES_VOICE",
+    "HERMES_VOICE_TTS",
+    "HERMES_VOICE_DEBUG",
+})
+for _runtime_var in _HERMES_RUNTIME_ZERO_VARS:
+    os.environ[_runtime_var] = "0"
+
 # Ensure project root is importable
 PROJECT_ROOT = Path(__file__).parent.parent
 if str(PROJECT_ROOT) not in sys.path:
@@ -170,6 +182,12 @@ def _looks_like_credential(name: str) -> bool:
 # unconditionally — individual tests that need them set do so explicitly.
 _HERMES_BEHAVIORAL_VARS = frozenset({
     "HERMES_YOLO_MODE",
+    # Runtime voice toggles are process-wide env flags in the TUI gateway.
+    # A live developer session can leave them enabled; tests must opt in
+    # explicitly or fake assistant completions may invoke real audio playback.
+    "HERMES_VOICE",
+    "HERMES_VOICE_TTS",
+    "HERMES_VOICE_DEBUG",
     "HERMES_INTERACTIVE",
     "HERMES_QUIET",
     "HERMES_TOOL_PROGRESS",
@@ -324,6 +342,11 @@ def _looks_like_credential(name: str) -> bool:
     "MATRIX_REQUIRE_MENTION",
 })
 
+# Voice flags are special: they are checked by background gateway threads after
+# a test emits ``message.complete``. They are neutralized at root-conftest import
+# for collection safety and kept at zero across fixture teardown so background
+# work cannot observe an inherited live interactive value between tests.
+
 
 @pytest.fixture(autouse=True)
 def _hermetic_environment(tmp_path, monkeypatch):
@@ -340,7 +363,10 @@ def _hermetic_environment(tmp_path, monkeypatch):
 
     # 2. Blank behavioral HERMES_* vars that could change test semantics.
     for name in _HERMES_BEHAVIORAL_VARS:
-        monkeypatch.delenv(name, raising=False)
+        if name in _HERMES_RUNTIME_ZERO_VARS:
+            os.environ[name] = "0"
+        else:
+            monkeypatch.delenv(name, raising=False)
 
     # 3. Redirect HERMES_HOME to a per-test tempdir. Code that reads
     #    ``~/.hermes/*`` via ``get_hermes_home()`` now gets the tempdir.

tests/test_run_tests_parallel.py

@@ -277,3 +277,183 @@ def test_positional_path_not_treated_as_flag(tmp_path: Path) -> None:
     # Discovery found the probe file (2 tests), proving the positional path
     # was consumed as a root, not forwarded to pytest as a bad flag.
     assert "test_flagprobe.py" in proc.stdout, proc.stdout
+
+
+def test_runner_scrubs_inherited_voice_tts_env(tmp_path: Path) -> None:
+    """Runtime voice flags from an interactive parent must not reach pytest.
+
+    The probe is intentionally outside ``tests/`` so the repo-level conftest
+    does not mask the runner behavior.
+    """
+    repo_root = Path(__file__).resolve().parent.parent
+    runner = repo_root / "scripts" / "run_tests_parallel.py"
+
+    probe_dir = tmp_path / "probe"
+    probe_dir.mkdir()
+    (probe_dir / "test_voice_env_probe.py").write_text(
+        textwrap.dedent(
+            """
+            import os
+
+            def test_voice_env_flags_are_neutralized():
+                assert os.environ.get("HERMES_VOICE") == "0"
+                assert os.environ.get("HERMES_VOICE_TTS") == "0"
+                assert os.environ.get("HERMES_VOICE_DEBUG") == "0"
+            """
+        ).strip()
+        + "\n"
+    )
+
+    env = os.environ.copy()
+    env.update(
+        {
+            "HERMES_VOICE": "1",
+            "HERMES_VOICE_TTS": "1",
+            "HERMES_VOICE_DEBUG": "1",
+        }
+    )
+    proc = subprocess.run(
+        [
+            sys.executable,
+            str(runner),
+            "--paths",
+            str(probe_dir),
+            "-j",
+            "1",
+            "--file-timeout",
+            "30",
+        ],
+        cwd=repo_root,
+        env=env,
+        stdout=subprocess.PIPE,
+        stderr=subprocess.STDOUT,
+        text=True,
+        timeout=60,
+    )
+
+    assert proc.returncode == 0, proc.stdout
+
+
+def test_pytest_conftest_scrubs_voice_env_before_test_module_collection(
+    tmp_path: Path,
+) -> None:
+    """Direct pytest runs must neutralize voice flags before module import."""
+    repo_root = Path(__file__).resolve().parent.parent
+    nonce = f"{os.getpid()}-{int(time.time() * 1000)}"
+    probe = repo_root / "tests" / f"_tmp_voice_collection_probe_{nonce}.py"
+    try:
+        probe.write_text(
+            textwrap.dedent(
+                """
+                import os
+
+                assert os.environ.get("HERMES_VOICE") == "0"
+                assert os.environ.get("HERMES_VOICE_TTS") == "0"
+                assert os.environ.get("HERMES_VOICE_DEBUG") == "0"
+
+                def test_collection_observed_neutral_voice_env():
+                    pass
+                """
+            ).strip()
+            + "\n"
+        )
+
+        env = os.environ.copy()
+        env.update(
+            {
+                "HERMES_VOICE": "1",
+                "HERMES_VOICE_TTS": "1",
+                "HERMES_VOICE_DEBUG": "1",
+            }
+        )
+        proc = subprocess.run(
+            [sys.executable, "-m", "pytest", "-q", str(probe)],
+            cwd=repo_root,
+            env=env,
+            stdout=subprocess.PIPE,
+            stderr=subprocess.STDOUT,
+            text=True,
+            timeout=60,
+        )
+
+        assert proc.returncode == 0, proc.stdout
+    finally:
+        try:
+            probe.unlink()
+        except FileNotFoundError:
+            pass
+
+
+def test_pytest_conftest_keeps_voice_tts_neutralized_after_test_teardown(
+    tmp_path: Path,
+) -> None:
+    """Background work after fixture teardown must still see voice TTS off."""
+    repo_root = Path(__file__).resolve().parent.parent
+    nonce = f"{os.getpid()}-{int(time.time() * 1000)}"
+    probe = repo_root / "tests" / f"_tmp_voice_teardown_probe_{nonce}.py"
+    handoff = tmp_path / "voice-env-after-teardown.txt"
+    try:
+        probe.write_text(
+            textwrap.dedent(
+                f"""
+                import json, os, threading, time
+                from pathlib import Path
+
+                HANDOFF = Path({str(handoff)!r})
+                VOICE_ENV_VARS = (
+                    "HERMES_VOICE",
+                    "HERMES_VOICE_TTS",
+                    "HERMES_VOICE_DEBUG",
+                )
+
+                def test_background_thread_observes_voice_env_after_return(monkeypatch):
+                    for name in VOICE_ENV_VARS:
+                        monkeypatch.setenv(name, "1")
+                        assert os.environ[name] == "1"
+
+                    def worker():
+                        deadline = time.monotonic() + 5
+                        while time.monotonic() < deadline and any(
+                            os.environ.get(name) == "1" for name in VOICE_ENV_VARS
+                        ):
+                            time.sleep(0.01)
+                        HANDOFF.write_text(json.dumps({{
+                            name: os.environ.get(name, "<missing>")
+                            for name in VOICE_ENV_VARS
+                        }}))
+
+                    threading.Thread(target=worker, daemon=False).start()
+                """
+            ).strip()
+            + "\n"
+        )
+
+        env = os.environ.copy()
+        env.update(
+            {
+                "HERMES_VOICE": "1",
+                "HERMES_VOICE_TTS": "1",
+                "HERMES_VOICE_DEBUG": "1",
+            }
+        )
+        proc = subprocess.run(
+            [sys.executable, "-m", "pytest", "-q", str(probe)],
+            cwd=repo_root,
+            env=env,
+            stdout=subprocess.PIPE,
+            stderr=subprocess.STDOUT,
+            text=True,
+            timeout=60,
+        )
+
+        assert proc.returncode == 0, proc.stdout
+        assert json.loads(handoff.read_text()) == {
+            "HERMES_VOICE": "0",
+            "HERMES_VOICE_TTS": "0",
+            "HERMES_VOICE_DEBUG": "0",
+        }
+    finally:
+        try:
+            probe.unlink()
+        except FileNotFoundError:
+            pass