fix: align avatar-live model name and path with backend (#34)

The Python SDK used model name 'avatar-live' with path '/v1/avatar-live/stream',
but the backend registers the model as 'live_avatar' and serves all realtime
models on '/v1/stream' (model passed as query param).

This mismatch caused 404 errors for Python SDK users trying to use Avatar Live.

Changes:
- Rename model from 'avatar-live' to 'live_avatar' (matches backend + JS SDK)
- Change url_path from '/v1/avatar-live/stream' to '/v1/stream' (matches backend)
- Update RealTimeModels type, example, playground, and tests accordingly

Author: AdirAmsalem

decart/models.py

@@ -4,7 +4,7 @@
 from .types import FileInput, MotionTrajectoryInput
 
 
-RealTimeModels = Literal["mirage", "mirage_v2", "lucy_v2v_720p_rt", "lucy_2_rt", "avatar-live"]
+RealTimeModels = Literal["mirage", "mirage_v2", "lucy_v2v_720p_rt", "lucy_2_rt", "live_avatar"]
 VideoModels = Literal[
     "lucy-dev-i2v",
     "lucy-fast-v2v",
@@ -182,9 +182,9 @@ class ImageToImageInput(DecartBaseModel):
             height=720,
             input_schema=BaseModel,
         ),
-        "avatar-live": ModelDefinition(
-            name="avatar-live",
-            url_path="/v1/avatar-live/stream",
+        "live_avatar": ModelDefinition(
+            name="live_avatar",
+            url_path="/v1/stream",
             fps=25,
             width=1280,
             height=720,

decart/realtime/webrtc_connection.py

@@ -246,7 +246,7 @@ async def on_ice_connection_state_change():
             self._pc.addTransceiver("audio", direction="recvonly")
             logger.debug("Added video+audio transceivers (recvonly) for subscribe mode")
         else:
-            if model_name == "avatar-live":
+            if model_name == "live_avatar":
                 self._pc.addTransceiver("video", direction="recvonly")
                 logger.debug("Added video transceiver (recvonly) for avatar-live mode")
             self._pc.addTrack(local_track)

examples/avatar_live.py

@@ -85,7 +85,7 @@ async def main():
 
     print("\nCreating Decart client...")
     async with DecartClient(api_key=api_key) as client:
-        model = models.realtime("avatar-live")
+        model = models.realtime("live_avatar")
         print(f"Using model: {model.name}")
 
         frame_count = 0

playground/playground.py

@@ -91,9 +91,9 @@ def _check_deps() -> None:
 
 # ── Constants ────────────────────────────────────────────────────────────────
 
-REALTIME_MODELS = ["mirage", "mirage_v2", "lucy_v2v_720p_rt", "lucy_2_rt", "avatar-live"]
+REALTIME_MODELS = ["mirage", "mirage_v2", "lucy_v2v_720p_rt", "lucy_2_rt", "live_avatar"]
 CAMERA_MODELS = {"mirage", "mirage_v2", "lucy_v2v_720p_rt", "lucy_2_rt"}
-AVATAR_MODELS = {"avatar-live"}
+AVATAR_MODELS = {"live_avatar"}
 
 BANNER = """
 ╔══════════════════════════════════════╗

tests/test_models.py

@@ -18,12 +18,12 @@ def test_realtime_models() -> None:
     assert model.url_path == "/v1/stream"
 
     # avatar-live model
-    model = models.realtime("avatar-live")
-    assert model.name == "avatar-live"
+    model = models.realtime("live_avatar")
+    assert model.name == "live_avatar"
     assert model.fps == 25
     assert model.width == 1280
     assert model.height == 720
-    assert model.url_path == "/v1/avatar-live/stream"
+    assert model.url_path == "/v1/stream"
 
 
 def test_video_models() -> None:

tests/test_realtime_unit.py

@@ -357,12 +357,12 @@ async def set_event():
 
 def test_avatar_live_model_available():
     """Test that avatar-live model is available"""
-    model = models.realtime("avatar-live")
-    assert model.name == "avatar-live"
+    model = models.realtime("live_avatar")
+    assert model.name == "live_avatar"
     assert model.fps == 25
     assert model.width == 1280
     assert model.height == 720
-    assert model.url_path == "/v1/avatar-live/stream"
+    assert model.url_path == "/v1/stream"
 
 
 @pytest.mark.asyncio
@@ -400,14 +400,14 @@ async def test_avatar_live_connect_with_initial_image():
             api_key=client.api_key,
             local_track=mock_track,
             options=RealtimeConnectOptions(
-                model=models.realtime("avatar-live"),
+                model=models.realtime("live_avatar"),
                 on_remote_stream=lambda t: None,
                 initial_state=ModelState(image=b"fake image bytes"),
             ),
         )
 
         assert realtime_client is not None
-        assert realtime_client._model_name == "avatar-live"
+        assert realtime_client._model_name == "live_avatar"
         mock_image_to_b64.assert_called_once()
         # Verify initial_image was passed to connect
         mock_manager.connect.assert_called_once()
@@ -448,7 +448,7 @@ async def test_avatar_live_set_image():
             api_key=client.api_key,
             local_track=mock_track,
             options=RealtimeConnectOptions(
-                model=models.realtime("avatar-live"),
+                model=models.realtime("live_avatar"),
                 on_remote_stream=lambda t: None,
             ),
         )
@@ -614,7 +614,7 @@ async def test_avatar_live_set_image_timeout():
             api_key=client.api_key,
             local_track=mock_track,
             options=RealtimeConnectOptions(
-                model=models.realtime("avatar-live"),
+                model=models.realtime("live_avatar"),
                 on_remote_stream=lambda t: None,
             ),
         )
@@ -659,7 +659,7 @@ async def test_avatar_live_set_image_server_error():
             api_key=client.api_key,
             local_track=mock_track,
             options=RealtimeConnectOptions(
-                model=models.realtime("avatar-live"),
+                model=models.realtime("live_avatar"),
                 on_remote_stream=lambda t: None,
             ),
         )