fix: valida mensagens e sequência de eventos WebSocket

src/voice_assistant/audio_gateway/app/main.py

@@ -554,14 +554,24 @@ async def audio_socket(websocket: WebSocket) -> None:
                 )
                 continue
 
-            event_type = payload.get("type")
-
-            if not isinstance(event_type, str):
+            if not isinstance(payload, dict):
                 await send_json(
                     websocket,
                     error_event(
                         "INVALID_MESSAGE",
-                        "Field 'type' is required.",
+                        "Message must be a JSON object.",
+                    ),
+                )
+                continue
+
+            event_type = payload.get("type")
+
+            if not isinstance(event_type, str) or not event_type.strip():
+                await send_json(
+                    websocket,
+                    error_event(
+                        "INVALID_MESSAGE",
+                        "Field 'type' must be a non-empty string.",
                     ),
                 )
                 continue
@@ -654,6 +664,16 @@ async def audio_socket(websocket: WebSocket) -> None:
             # ---------------------------------------------------------
             # Todos os eventos abaixo exigem sessão ativa
             # ---------------------------------------------------------
+            if session is None:
+                await send_json(
+                    websocket,
+                    error_event(
+                        "SESSION_NOT_FOUND",
+                        "session.start is required before this event.",
+                    ),
+                )
+                continue
+
             if session.closed:
                 await send_json(
                     websocket,
@@ -669,6 +689,18 @@ async def audio_socket(websocket: WebSocket) -> None:
             # audio.start
             # ---------------------------------------------------------
             if event_type == "audio.start":
+                if session.active_turn_id is not None:
+                    await send_json(
+                        websocket,
+                        error_event(
+                            "INVALID_TURN",
+                            "A turn is already active.",
+                            session_id=session.session_id,
+                            turn_id=session.active_turn_id,
+                        ),
+                    )
+                    continue
+
                 turn_id: int | None = None
 
                 try:

tests/test_gateway.py

@@ -1044,7 +1044,7 @@ def test_websocket_interrupt_timeout_preserves_active_turn():
             next_response = websocket.receive_json()
 
             assert next_response["type"] == "error"
-            assert next_response["code"] == "STT_START_FAILED"
+            assert next_response["code"] == "INVALID_TURN"
 
 
 def test_websocket_interrupt_during_agent():
@@ -1625,7 +1625,7 @@ def test_websocket_audio_start_abort_failure_is_not_retryable():
             next_response = websocket.receive_json()
 
             assert next_response["type"] == "error"
-            assert next_response["code"] == "STT_START_FAILED"
+            assert next_response["code"] == "INVALID_TURN"
 
 
 
@@ -2410,3 +2410,121 @@ def test_websocket_interrupt_wins_pending_tts_completion():
 
                 assert response["type"] == "interrupted"
                 assert response["turn_id"] == next_turn["turn_id"]
+
+
+@pytest.mark.parametrize("session_started", [False, True])
+@pytest.mark.parametrize(
+    "payload",
+    [
+        None,
+        [],
+        "text",
+        42,
+        True,
+        {},
+        {"type": None},
+        {"type": 42},
+        {"type": ""},
+        {"type": " "},
+    ],
+)
+def test_websocket_rejects_invalid_message_and_remains_usable(
+    payload,
+    session_started,
+):
+    with client.websocket_connect("/ws/audio") as websocket:
+        if session_started:
+            websocket.send_json({"type": "session.start"})
+            assert websocket.receive_json()["type"] == "session.ready"
+
+        websocket.send_json(payload)
+
+        error = websocket.receive_json()
+        assert error["type"] == "error"
+        assert error["code"] == "INVALID_MESSAGE"
+
+        if session_started:
+            websocket.send_json({"type": "session.end"})
+            assert websocket.receive_json()["type"] == "session.ended"
+        else:
+            websocket.send_json({"type": "session.start"})
+            assert websocket.receive_json()["type"] == "session.ready"
+
+
+@pytest.mark.parametrize(
+    "event_type",
+    [
+        "audio.start",
+        "audio.chunk",
+        "audio.stop",
+        "interrupt",
+        "session.end",
+    ],
+)
+def test_websocket_requires_session_before_control_events(event_type):
+    with client.websocket_connect("/ws/audio") as websocket:
+        websocket.send_json(
+            {
+                "type": event_type,
+                "turn_id": 1,
+                "sequence_number": 0,
+            }
+        )
+
+        error = websocket.receive_json()
+        assert error["type"] == "error"
+        assert error["code"] == "SESSION_NOT_FOUND"
+
+        websocket.send_json({"type": "session.start"})
+        assert websocket.receive_json()["type"] == "session.ready"
+
+
+@pytest.mark.parametrize(
+    "text",
+    ["{", "", '{type: "session.start"}'],
+)
+def test_websocket_rejects_malformed_json_and_remains_usable(text):
+    with client.websocket_connect("/ws/audio") as websocket:
+        websocket.send_text(text)
+
+        error = websocket.receive_json()
+        assert error["type"] == "error"
+        assert error["code"] == "INVALID_MESSAGE"
+
+        websocket.send_json({"type": "session.start"})
+        assert websocket.receive_json()["type"] == "session.ready"
+
+
+def test_websocket_duplicate_audio_start_preserves_active_turn():
+    created_stt.clear()
+    created_turn_managers.clear()
+
+    with client.websocket_connect("/ws/audio") as websocket:
+        websocket.send_json({"type": "session.start"})
+        assert websocket.receive_json()["type"] == "session.ready"
+
+        websocket.send_json({"type": "audio.start"})
+        started = websocket.receive_json()
+        assert started["type"] == "turn.started"
+
+        turn_id = started["turn_id"]
+        stt = created_stt[0]
+        turn = created_turn_managers[0].active_turn
+
+        websocket.send_json({"type": "audio.start"})
+
+        error = websocket.receive_json()
+        assert error["type"] == "error"
+        assert error["code"] == "INVALID_TURN"
+        assert error["turn_id"] == turn_id
+        assert len(created_stt) == 1
+        assert stt._started is True
+        assert created_turn_managers[0].active_turn is turn
+        assert turn.state == TurnState.LISTENING
+
+        websocket.send_json(
+            {"type": "interrupt", "turn_id": turn_id}
+        )
+        assert websocket.receive_json()["type"] == "interrupted"
+        assert stt._started is False
+        assert created_turn_managers[0].active_turn is None