Add Discord-native hybrid front-end for Jarvis (bot + bridge)

Transform isair/jarvis into a Discord-controlled voice assistant running on
the Ubuntu VNC desktop, keeping the mature ~39k-line Python brain intact.

- bot/ (Node + bun, discord.js): /자비스 slash commands (ephemeral),
  voice channel join + voice receive/playback, pluggable VNC screen broadcast
  (selfbot live / noVNC / screenshot)
- bridge/ (Python, Flask): wraps jarvis STT + run_reply_engine + Piper TTS
  behind a thin localhost HTTP API
- .env.example, scripts/ (start_bridge/start_bot/dev), README rewrite,
  docs/language-comparison.md and docs/vnc-xfce-setup.md

Language decision: hybrid (Python brain + Node/bun Discord layer) because
Discord blocks bot video; native screen broadcast only works via a Node
selfbot library.

tests/conftest.py

@@ -0,0 +1,124 @@
+import sys
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import Any, Dict, List, Optional
+
+import pytest
+
+# Robustly locate repository root (directory containing src/jarvis)
+_this_file = Path(__file__).resolve()
+ROOT = None
+for parent in _this_file.parents:
+    if (parent / "src" / "jarvis").exists():
+        ROOT = parent
+        break
+if ROOT is None:
+    # Fallback to two levels up
+    ROOT = _this_file.parent.parent
+
+SRC = ROOT / "src"
+# Both ROOT and SRC are on sys.path so tests can write either
+#   ``from src.jarvis.x import ...``  (older style, ``src.`` prefix)
+# or
+#   ``from jarvis.x import ...``      (newer style, no prefix)
+# CAUTION: those two import paths resolve to *distinct module instances*.
+# A monkeypatch on ``src.jarvis.memory.conversation.X`` does NOT take
+# effect on ``jarvis.memory.conversation.X`` and vice versa. When a test
+# stubs out a symbol the production code calls, you MUST patch the same
+# module instance the production code resolves at runtime. Production code
+# in ``src/`` imports without the ``src.`` prefix (e.g. inside endpoint
+# handlers it's ``from jarvis.memory.conversation import ...``), so a test
+# that monkeypatches a symbol used by production should also import
+# without the prefix. This is the convention going forward; the older
+# ``from src.X`` style is left in place to avoid a churn-only sweep, but
+# do not adopt it for new tests that monkeypatch.
+# Add repository root so that 'src' is a package prefix.
+if str(ROOT) not in sys.path:
+    sys.path.insert(0, str(ROOT))
+# Also add the src directory (optional, for backwards compatibility with direct 'jarvis' imports)
+if str(SRC) not in sys.path:
+    sys.path.insert(0, str(SRC))
+
+
+@dataclass
+class MockConfig:
+    """Minimal config object for unit tests that need a config."""
+    ollama_base_url: str = "http://localhost:11434"
+    ollama_chat_model: str = "gemma4:e2b"
+    ollama_embed_model: str = "nomic-embed-text"
+    db_path: str = ":memory:"
+    sqlite_vss_path: Optional[str] = None
+    voice_debug: bool = True
+    tts_enabled: bool = False
+    tts_engine: str = "piper"
+    tts_voice: Optional[str] = None
+    tts_rate: int = 200
+    tts_piper_model_path: Optional[str] = None
+    tts_piper_speaker: Optional[int] = None
+    tts_piper_length_scale: float = 1.0
+    tts_piper_noise_scale: float = 0.667
+    tts_piper_noise_w: float = 0.8
+    tts_piper_sentence_silence: float = 0.2
+    tts_chatterbox_device: str = "cpu"
+    tts_chatterbox_audio_prompt: Optional[str] = None
+    tts_chatterbox_exaggeration: float = 0.5
+    tts_chatterbox_cfg_weight: float = 0.5
+    web_search_enabled: bool = True
+    brave_search_api_key: str = ""
+    wikipedia_fallback_enabled: bool = True
+    llm_tools_timeout_sec: float = 8.0
+    llm_embed_timeout_sec: float = 10.0
+    llm_chat_timeout_sec: float = 45.0
+    agentic_max_turns: int = 8
+    tool_selection_strategy: str = "embedding"
+    tool_router_model: str = ""
+    memory_enrichment_max_results: int = 5
+    memory_enrichment_source: str = "diary"
+    location_enabled: bool = True
+    location_ip_address: Optional[str] = None
+    location_auto_detect: bool = False
+    location_cgnat_resolve_public_ip: bool = False
+    dialogue_memory_timeout: int = 300
+    llm_thinking_enabled: bool = False
+    intent_judge_thinking_enabled: bool = False
+    dictation_thinking_enabled: bool = False
+    mcps: Dict[str, Any] = field(default_factory=dict)
+    use_stdin: bool = True
+
+
+@pytest.fixture
+def mock_config():
+    """Provide a mock configuration for unit tests."""
+    return MockConfig()
+
+
+@pytest.fixture
+def db():
+    """Provide an in-memory database for unit tests."""
+    from jarvis.memory.db import Database
+    database = Database(":memory:", sqlite_vss_path=None)
+    yield database
+    database.close()
+
+
+@pytest.fixture
+def dialogue_memory():
+    """Provide a dialogue memory instance for unit tests."""
+    from jarvis.memory.conversation import DialogueMemory
+    return DialogueMemory(inactivity_timeout=300, max_interactions=20)
+
+
+@pytest.fixture
+def qapp():
+    """Provide a shared QApplication for Qt-based UI tests.
+
+    Qt requires exactly one QApplication per process.  Re-uses an existing
+    instance when present so repeated test runs inside a single session
+    don't error.
+    """
+    from PyQt6.QtWidgets import QApplication
+    app = QApplication.instance()
+    if app is None:
+        app = QApplication([])
+    yield app
+