Fix url generation (#631)

* fix url generation

* move import

* move these back

* fix tests

* make ruff happy

Author: PaulAsjes

src/elevenlabs/conversational_ai/conversation.py

@@ -6,6 +6,7 @@
 import asyncio
 from concurrent.futures import ThreadPoolExecutor
 from enum import Enum
+import urllib.parse
 
 from websockets.sync.client import connect, Connection
 import websockets
@@ -329,8 +330,11 @@ def __init__(
 
     def _get_wss_url(self):
         base_http_url = self.client._client_wrapper.get_base_url()
-        base_ws_url = base_http_url.replace("https://", "wss://").replace("http://", "ws://")
-        return f"{base_ws_url}/v1/convai/conversation?agent_id={self.agent_id}&source=python_sdk&version={__version__}"
+        base_ws_url = urllib.parse.urlparse(base_http_url)._replace(scheme="wss" if base_http_url.startswith("https") else "ws").geturl()
+        # Ensure base URL ends with '/' for proper joining
+        if not base_ws_url.endswith('/'):
+            base_ws_url += '/'
+        return f"{base_ws_url}v1/convai/conversation?agent_id={self.agent_id}&source=python_sdk&version={__version__}"
 
     def _get_signed_url(self):
         response = self.client.conversational_ai.conversations.get_signed_url(agent_id=self.agent_id)

tests/test_async_convai.py

@@ -1,11 +1,12 @@
 import asyncio
 import json
 from unittest.mock import AsyncMock, MagicMock, patch
+
 import pytest
 
 from elevenlabs.conversational_ai.conversation import (
-    AsyncConversation,
     AsyncAudioInterface,
+    AsyncConversation,
     ConversationInitiationData,
 )
 
@@ -45,7 +46,6 @@ def create_mock_async_websocket(messages=None):
 
     # Convert messages to JSON strings
     json_messages = [json.dumps(msg) for msg in messages]
-    json_messages.extend(['{"type": "keep_alive"}'] * 10)  # Add some keep-alive messages
 
     # Create an iterator
     message_iter = iter(json_messages)
@@ -54,8 +54,9 @@ async def mock_recv():
         try:
             return next(message_iter)
         except StopIteration:
-            # Simulate connection close after messages
-            raise asyncio.TimeoutError()
+            # After all messages, simulate timeout by sleeping forever
+            # This will be caught by asyncio.wait_for timeout in the conversation
+            await asyncio.sleep(float("inf"))
 
     mock_ws.recv = mock_recv
     return mock_ws
@@ -66,6 +67,7 @@ async def test_async_conversation_basic_flow():
     # Mock setup
     mock_ws = create_mock_async_websocket()
     mock_client = MagicMock()
+    mock_client._client_wrapper.get_base_url.return_value = "https://api.elevenlabs.io"
     agent_response_callback = AsyncMock()
     test_user_id = "test_user_123"
 
@@ -94,7 +96,7 @@ async def test_async_conversation_basic_flow():
 
     # Assertions - check the call was made with the right structure
     send_calls = [call[0][0] for call in mock_ws.send.call_args_list]
-    init_messages = [json.loads(call) for call in send_calls if 'conversation_initiation_client_data' in call]
+    init_messages = [json.loads(call) for call in send_calls if "conversation_initiation_client_data" in call]
     assert len(init_messages) == 1
     init_message = init_messages[0]
 
@@ -148,6 +150,7 @@ async def test_async_conversation_with_dynamic_variables():
     # Mock setup
     mock_ws = create_mock_async_websocket()
     mock_client = MagicMock()
+    mock_client._client_wrapper.get_base_url.return_value = "https://api.elevenlabs.io"
     agent_response_callback = AsyncMock()
 
     dynamic_variables = {"name": "angelo"}
@@ -177,7 +180,7 @@ async def test_async_conversation_with_dynamic_variables():
 
     # Assertions - check the call was made with the right structure
     send_calls = [call[0][0] for call in mock_ws.send.call_args_list]
-    init_messages = [json.loads(call) for call in send_calls if 'conversation_initiation_client_data' in call]
+    init_messages = [json.loads(call) for call in send_calls if "conversation_initiation_client_data" in call]
     assert len(init_messages) == 1
     init_message = init_messages[0]
 
@@ -196,6 +199,7 @@ async def test_async_conversation_with_contextual_update():
     # Mock setup
     mock_ws = create_mock_async_websocket([])
     mock_client = MagicMock()
+    mock_client._client_wrapper.get_base_url.return_value = "https://api.elevenlabs.io"
 
     # Setup the conversation
     conversation = AsyncConversation(
@@ -228,6 +232,7 @@ async def test_async_conversation_send_user_message():
     # Mock setup
     mock_ws = create_mock_async_websocket([])
     mock_client = MagicMock()
+    mock_client._client_wrapper.get_base_url.return_value = "https://api.elevenlabs.io"
 
     # Setup the conversation
     conversation = AsyncConversation(
@@ -260,6 +265,7 @@ async def test_async_conversation_register_user_activity():
     # Mock setup
     mock_ws = create_mock_async_websocket([])
     mock_client = MagicMock()
+    mock_client._client_wrapper.get_base_url.return_value = "https://api.elevenlabs.io"
 
     # Setup the conversation
     conversation = AsyncConversation(
@@ -300,29 +306,21 @@ async def test_async_conversation_callback_flows():
             "type": "agent_response_correction",
             "agent_response_correction_event": {
                 "original_agent_response": "Hello ther!",
-                "corrected_agent_response": "Hello there!"
-            }
-        },
-        {
-            "type": "user_transcript",
-            "user_transcription_event": {"user_transcript": "Hi, how are you?"}
-        },
-        {
-            "type": "ping",
-            "ping_event": {"event_id": "123", "ping_ms": 50}
-        },
-        {
-            "type": "interruption",
-            "interruption_event": {"event_id": "456"}
+                "corrected_agent_response": "Hello there!",
+            },
         },
+        {"type": "user_transcript", "user_transcription_event": {"user_transcript": "Hi, how are you?"}},
+        {"type": "ping", "ping_event": {"event_id": "123", "ping_ms": 50}},
+        {"type": "interruption", "interruption_event": {"event_id": "456"}},
         {
             "type": "audio",
-            "audio_event": {"event_id": "789", "audio_base_64": "dGVzdA=="}  # "test" in base64
-        }
+            "audio_event": {"event_id": "789", "audio_base_64": "dGVzdA=="},  # "test" in base64
+        },
     ]
 
     mock_ws = create_mock_async_websocket(messages)
     mock_client = MagicMock()
+    mock_client._client_wrapper.get_base_url.return_value = "https://api.elevenlabs.io"
 
     # Setup callbacks
     agent_response_callback = AsyncMock()
@@ -368,7 +366,6 @@ async def test_async_conversation_callback_flows():
 
 @pytest.mark.asyncio
 async def test_async_conversation_wss_url_generation_without_get_environment():
-
     from elevenlabs.core.client_wrapper import SyncClientWrapper
 
     # Test with various base URL formats to ensure robustness
@@ -383,24 +380,20 @@ async def test_async_conversation_wss_url_generation_without_get_environment():
         # Create a real SyncClientWrapper to ensure it doesn't have get_environment method
         mock_client = MagicMock()
         mock_client._client_wrapper = SyncClientWrapper(
-            base_url=base_url,
-            api_key="test_key",
-            httpx_client=MagicMock(),
-            timeout=30.0
+            base_url=base_url, api_key="test_key", httpx_client=MagicMock(), timeout=30.0
         )
 
         conversation = AsyncConversation(
-            client=mock_client,
-            agent_id=TEST_AGENT_ID,
-            requires_auth=False,
-            audio_interface=MockAsyncAudioInterface()
+            client=mock_client, agent_id=TEST_AGENT_ID, requires_auth=False, audio_interface=MockAsyncAudioInterface()
         )
 
         try:
             wss_url = conversation._get_wss_url()
 
             # Verify the URL is correctly generated
-            expected_url = f"{expected_ws_base}/v1/convai/conversation?agent_id={TEST_AGENT_ID}&source=python_sdk&version="
+            expected_url = (
+                f"{expected_ws_base}/v1/convai/conversation?agent_id={TEST_AGENT_ID}&source=python_sdk&version="
+            )
             assert wss_url.startswith(expected_url), f"URL should start with {expected_url}, got {wss_url}"
 
             # Verify the URL contains version parameter
@@ -414,3 +407,44 @@ async def test_async_conversation_wss_url_generation_without_get_environment():
 
         except Exception as e:
             assert False, f"Unexpected error generating WebSocket URL: {e}"
+
+
+@pytest.mark.asyncio
+async def test_async_websocket_url_construction_edge_cases():
+    """Test WebSocket URL construction edge cases for async conversation, specifically for trailing slash handling."""
+    from elevenlabs.conversational_ai.conversation import AsyncConversation
+    from elevenlabs.core.client_wrapper import SyncClientWrapper
+
+    # Test cases with various base URL formats
+    test_cases = [
+        # Base URLs without trailing slashes (the main edge case)
+        ("https://api.eu.residency.elevenlabs.io", "wss://api.eu.residency.elevenlabs.io"),
+        ("https://api.elevenlabs.io", "wss://api.elevenlabs.io"),
+        ("http://localhost:8000", "ws://localhost:8000"),
+        # Base URLs with trailing slashes (should still work)
+        ("https://api.eu.residency.elevenlabs.io/", "wss://api.eu.residency.elevenlabs.io"),
+        ("https://api.elevenlabs.io/", "wss://api.elevenlabs.io"),
+        ("http://localhost:8000/", "ws://localhost:8000"),
+    ]
+
+    for base_url, expected_ws_base in test_cases:
+        # Test async conversation WebSocket URL construction
+        mock_client = MagicMock()
+        mock_client._client_wrapper = SyncClientWrapper(
+            base_url=base_url, api_key="test_key", httpx_client=MagicMock(), timeout=30.0
+        )
+
+        conversation = AsyncConversation(
+            client=mock_client, agent_id=TEST_AGENT_ID, requires_auth=False, audio_interface=MockAsyncAudioInterface()
+        )
+
+        # Test conversation URL generation
+        conv_url = conversation._get_wss_url()
+        expected_conv_url = f"{expected_ws_base}/v1/convai/conversation"
+        assert (
+            expected_conv_url in conv_url
+        ), f"Async conversation URL should contain {expected_conv_url}, got {conv_url}"
+
+        # Ensure no double slashes in the path (except after the protocol)
+        url_path = conv_url.split("://", 1)[1]  # Remove protocol
+        assert "//" not in url_path, f"Async conversation URL should not contain double slashes in path: {conv_url}"

tests/test_convai.py

@@ -3,7 +3,6 @@
 import json
 import time
 
-
 class MockAudioInterface(AudioInterface):
     def start(self, input_callback):
         print("Audio interface started")
@@ -51,6 +50,7 @@ def test_conversation_basic_flow():
     # Mock setup
     mock_ws = create_mock_websocket()
     mock_client = MagicMock()
+    mock_client._client_wrapper.get_base_url.return_value = "https://api.elevenlabs.io"
     agent_response_callback = MagicMock()
     test_user_id = "test_user_123"
 
@@ -132,6 +132,7 @@ def test_conversation_with_dynamic_variables():
     # Mock setup
     mock_ws = create_mock_websocket()
     mock_client = MagicMock()
+    mock_client._client_wrapper.get_base_url.return_value = "https://api.elevenlabs.io"
     agent_response_callback = MagicMock()
 
     dynamic_variables = {"name": "angelo"}
@@ -181,6 +182,7 @@ def test_conversation_with_contextual_update():
     # Mock setup
     mock_ws = create_mock_websocket([])
     mock_client = MagicMock()
+    mock_client._client_wrapper.get_base_url.return_value = "https://api.elevenlabs.io"
 
     # Setup the conversation
     conversation = Conversation(
@@ -256,3 +258,71 @@ def test_conversation_wss_url_generation_without_get_environment():
 
         except Exception as e:
             assert False, f"Unexpected error generating WebSocket URL: {e}"
+
+
+def test_websocket_url_construction_edge_cases():
+    """Test WebSocket URL construction edge cases, specifically for trailing slash handling."""
+    from elevenlabs.core.client_wrapper import SyncClientWrapper
+    from elevenlabs.conversational_ai.conversation import Conversation
+    from elevenlabs.realtime_tts import RealtimeTextToSpeechClient
+
+    # Test cases with various base URL formats
+    test_cases = [
+        # Base URLs without trailing slashes (the main edge case)
+        ("https://api.eu.residency.elevenlabs.io", "wss://api.eu.residency.elevenlabs.io", "wss://api.eu.residency.elevenlabs.io"),
+        ("https://api.elevenlabs.io", "wss://api.elevenlabs.io", "wss://api.elevenlabs.io"),
+        ("http://localhost:8000", "ws://localhost:8000", "wss://localhost:8000"),
+        # Base URLs with trailing slashes (should still work)
+        ("https://api.eu.residency.elevenlabs.io/", "wss://api.eu.residency.elevenlabs.io", "wss://api.eu.residency.elevenlabs.io/"),
+        ("https://api.elevenlabs.io/", "wss://api.elevenlabs.io", "wss://api.elevenlabs.io/"),
+        ("http://localhost:8000/", "ws://localhost:8000", "wss://localhost:8000/"),
+    ]
+
+    for base_url, expected_ws_base, expected_tts_ws_base in test_cases:
+        # Test conversation WebSocket URL construction
+        mock_client = MagicMock()
+        mock_client._client_wrapper = SyncClientWrapper(
+            base_url=base_url,
+            api_key="test_key",
+            httpx_client=MagicMock(),
+            timeout=30.0
+        )
+
+        conversation = Conversation(
+            client=mock_client,
+            agent_id=TEST_AGENT_ID,
+            requires_auth=False,
+            audio_interface=MockAudioInterface()
+        )
+
+        # Test conversation URL generation
+        conv_url = conversation._get_wss_url()
+        expected_conv_url = f"{expected_ws_base}/v1/convai/conversation"
+        assert expected_conv_url in conv_url, f"Conversation URL should contain {expected_conv_url}, got {conv_url}"
+
+        # Ensure no double slashes in the path (except after the protocol)
+        url_path = conv_url.split("://", 1)[1]  # Remove protocol
+        assert "//" not in url_path, f"URL should not contain double slashes in path: {conv_url}"
+
+        # Test realtime TTS WebSocket URL construction
+        realtime_client = RealtimeTextToSpeechClient(client_wrapper=mock_client._client_wrapper)
+
+        # Test the WebSocket base URL construction
+        # Note: realtime TTS always uses wss scheme, not ws
+        assert realtime_client._ws_base_url == expected_tts_ws_base, f"TTS WebSocket base URL should be {expected_tts_ws_base}, got {realtime_client._ws_base_url}"
+
+        # Test full URL construction using urljoin (simulating the actual method)
+        import urllib.parse
+        test_voice_id = "test_voice_123"
+        test_model = "eleven_turbo_v2_5"
+        test_format = "mp3_44100_128"
+        relative_path = f"v1/text-to-speech/{test_voice_id}/stream-input?model_id={test_model}&output_format={test_format}"
+
+        full_tts_url = urllib.parse.urljoin(realtime_client._ws_base_url, relative_path)
+        # For URLs with trailing slash, expect it to be preserved in the joined URL
+        expected_tts_url_base = expected_tts_ws_base.rstrip('/') + "/v1/text-to-speech/" + test_voice_id + "/stream-input"
+        assert expected_tts_url_base in full_tts_url, f"TTS URL should contain {expected_tts_url_base}, got {full_tts_url}"
+
+        # Ensure no double slashes in the path
+        tts_url_path = full_tts_url.split("://", 1)[1]
+        assert "//" not in tts_url_path, f"TTS URL should not contain double slashes in path: {full_tts_url}"