What does this PR do?

Marks Hermes's auto voice reply (the TTS-generated final response for users with voice.auto_tts enabled) as notify-worthy so Telegram delivers the final voice note as a normal push instead of a silent message. Mirrors the final-text precedent in gateway/platforms/base.py:3214, which already sets metadata["notify"] = True for the final text response. In Telegram's default notifications=important mode, TelegramPlatformAdapter._notification_kwargs() returns disable_notification=True for every send unless metadata["notify"] is truthy; the auto-TTS path through GatewayRunner._send_voice_reply was forwarding thread metadata unmodified, so the final voice note arrived silently. The fix clones the thread metadata and sets notify=True before passing it to adapter.send_voice (clone-then-mark so we don't leak notify=True back into the typing-indicator state that shares this metadata).

Related Issue

Fixes #27970 (Bug 2 of 2 — the silent-final-voice-reply piece; Bug 1, MP3 vs. native OGG fallback, is already covered by open PRs #20182 and #20878 per alt-glitch's pointer on the issue thread, and the reporter agreed to split).

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 🔒 Security fix
• 📝 Documentation update
• ✅ Tests (adding or improving test coverage)
• ♻️ Refactor (no behavior change)
• 🎯 New skill (bundled or hub)

Changes Made

• gateway/run.py — in _send_voice_reply, clone thread_meta and set notify=True before passing it to adapter.send_voice so Telegram's _notification_kwargs does not strip the push notification. Clone-then-mark to avoid leaking notify=True into the shared typing-indicator metadata.
• tests/gateway/test_send_voice_reply_notify.py — new regression file with two cases (DM with no thread metadata gets a fresh {"notify": True}; DM topic with existing thread metadata gets notify=True added without mutating the source dict).
• tests/gateway/test_voice_command.py — updated TestSendVoiceReply::test_auto_voice_reply_uses_thread_metadata_helper to assert notify: True in the dispatched metadata.

How to Test

1. uv run --with pytest --with pytest-xdist --with pytest-asyncio python3 -m pytest tests/gateway/test_voice_command.py tests/gateway/test_send_voice_reply_notify.py -v — 8 passed.
2. Adjacent suites also pass: tests/gateway/test_tts_media_routing.py, tests/gateway/test_voice_mode_platform_isolation.py, tests/gateway/test_matrix_voice.py.
3. Regression guard: the two new tests fail against origin/main with AssertionError: notify is None, and pass with the fix applied.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run focused tests for the touched code and all pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS 15.x

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

Sibling Audit

Sibling code paths that may need the same fix: GatewayRunner._deliver_media_from_response (gateway/run.py:10630) also delivers final-turn voice/audio media via adapter.send_voice without notify=True. Intentionally left out of this PR's scope to keep the diff small — happy to widen if preferred.

Other platforms (Discord, Slack, Matrix) ignore metadata["notify"], so they are unaffected.