Enhance session management and logging configuration

- Updated .env.example to clarify audio frame size validation and default codec settings.
- Refactored logging setup in main.py to support JSON serialization based on log format configuration.
- Improved session.py to dynamically compute audio frame bytes and include protocol version in session events.
- Added tests to validate session start events and audio frame handling based on chunk size settings.

engine/tests/test_ws_protocol_session_start.py

@@ -290,6 +290,8 @@ async def test_handle_session_start_applies_whitelisted_overrides_and_ignores_wo
 @pytest.mark.asyncio
 async def test_handle_session_start_emits_config_resolved_when_enabled(monkeypatch):
     monkeypatch.setattr("core.session.settings.ws_emit_config_resolved", True)
+    monkeypatch.setattr("core.session.settings.ws_protocol_version", "v1-custom")
+    monkeypatch.setattr("core.session.settings.default_codec", "pcmu")
 
     session = Session.__new__(Session)
     session.id = "sess_start_emit_config"
@@ -368,10 +370,46 @@ async def test_handle_session_start_emits_config_resolved_when_enabled(monkeypat
     )
 
     config_event = next(item for item in events if item.get("type") == "config.resolved")
+    session_started_event = next(item for item in events if item.get("type") == "session.started")
+    assert session_started_event["protocolVersion"] == "v1-custom"
     assert "appId" not in config_event["config"]
     assert "configVersionId" not in config_event["config"]
     assert "services" not in config_event["config"]
+    assert config_event["config"]["protocolVersion"] == "v1-custom"
     assert config_event["config"]["channel"] == "web_debug"
     assert config_event["config"]["output"]["mode"] == "text"
+    assert config_event["config"]["output"]["codec"] == "pcmu"
     assert config_event["config"]["tools"]["enabled"] is True
     assert config_event["config"]["tools"]["count"] == 1
+
+
+@pytest.mark.asyncio
+async def test_handle_audio_uses_chunk_size_for_frame_validation(monkeypatch):
+    monkeypatch.setattr("core.session.settings.sample_rate", 16000)
+    monkeypatch.setattr("core.session.settings.chunk_size_ms", 10)
+
+    session = Session.__new__(Session)
+    session.id = "sess_chunk_frame"
+    session.ws_state = WsSessionState.ACTIVE
+
+    class _Pipeline:
+        def __init__(self):
+            self.frames = []
+
+        async def process_audio(self, frame: bytes):
+            self.frames.append(frame)
+
+    session.pipeline = _Pipeline()
+    errors = []
+
+    async def _send_error(sender, message, code, **kwargs):
+        _ = (sender, kwargs)
+        errors.append((code, message))
+
+    session._send_error = _send_error
+    payload = b"\x00\x01" * 320  # 640 bytes = 2 frames when chunk_size_ms=10
+    await session.handle_audio(payload)
+
+    assert errors == []
+    assert len(session.pipeline.frames) == 2
+    assert all(len(frame) == 320 for frame in session.pipeline.frames)