Enhance RimeTTS configuration with credential mode support

- Updated RimeTTS parameters to include `credential_mode`, allowing for both BYOK and managed credential configurations.
- Added validation logic to ensure required parameters are provided based on the selected credential mode.
- Updated documentation to reflect changes in parameter requirements and usage examples for both credential modes.
- Added tests to verify correct behavior for RimeTTS with managed and BYOK credential modes.

Author: seymourtang

docs/concepts/vendors.md

@@ -62,7 +62,7 @@ Used with `agent.with_tts()`. Each TTS vendor produces audio at a specific sampl
 | `GoogleTTS` | Google Cloud | `key`, `voice_name` | — |
 | `AmazonTTS` | Amazon Polly | `access_key`, `secret_key`, `region`, `voice_id`, `engine` | — |
 | `HumeAITTS` | Hume AI | `key`, `voice_id`, `provider` | — |
-| `RimeTTS` | Rime | `key`, `speaker`, `model_id` | — |
+| `RimeTTS` | Rime | `model_id`; BYOK also requires `key` and `speaker`, managed requires `base_url` | — |
 | `FishAudioTTS` | Fish Audio | `key`, `reference_id`, `backend` | — |
 | `MurfTTS` | Murf | `key`, `voice_id`, `model` | — |
 | `MiniMaxTTS` | MiniMax | `model` for supported Agora-managed global models; `key`, `group_id`, `model`, `voice_id`, `url` for BYOK | — |
@@ -71,6 +71,8 @@ Used with `agent.with_tts()`. Each TTS vendor produces audio at a specific sampl
 | `SarvamTTS` | Sarvam | `api_key` | — |
 | `XaiTTS` | xAI | `api_key`, `language` | Configurable |
 
+For `RimeTTS`, omitting `credential_mode` selects BYOK validation. Set `credential_mode=CredentialMode.MANAGED` to use managed credentials. See the [RimeTTS vendor reference](../reference/vendors.md#rimetts) for both configurations.
+
 ### CN TTS Vendors
 
 Used with `agent.with_tts()` when routing to `Area.CN`. Use `MiniMaxCNTTS` and `MicrosoftCNTTS` for CN-specific implementations that differ from the global classes.

docs/reference/vendors.md

@@ -318,14 +318,45 @@ The SDK also includes named helpers for the remaining Agora-supported LLM provid
 
 ### `RimeTTS`
 
+`CredentialMode` is exported from `agora_agent` as the shared credential mode constants for provider integrations.
+
 | Parameter | Type | Required | Default | Description |
 |---|---|---|---|---|
-| `key` | `str` | Yes | — | Rime API key |
-| `speaker` | `str` | Yes | — | Speaker ID |
+| `key` | `str` | BYOK: Yes; Managed: No | `None` | Rime API key |
+| `speaker` | `str` | BYOK: Yes; Managed: No | `None` | Speaker ID |
 | `model_id` | `str` | Yes | — | Model ID |
-| `base_url` | `str` | No | `None` | WebSocket URL |
+| `base_url` | `str` | Managed: Yes; BYOK: No | `None` | WebSocket URL |
+| `credential_mode` | `CredentialMode` | No | `None` | Shared credential mode (`"managed"` or `"byok"`); omission uses BYOK validation |
 | `skip_patterns` | `List[int]` | No | `None` | Skip patterns |
 
+When `credential_mode` is omitted or set to `"byok"`, provide `key`, `speaker`, and `model_id`:
+
+```python
+import os
+
+from agora_agent import RimeTTS
+
+tts = RimeTTS(
+    key=os.environ["RIME_API_KEY"],
+    speaker="your-speaker-id",
+    model_id="mist",
+)
+```
+
+For Agora-managed credentials, set `credential_mode=CredentialMode.MANAGED` and provide `base_url` and `model_id`:
+
+```python
+from agora_agent import CredentialMode, RimeTTS
+
+tts = RimeTTS(
+    credential_mode=CredentialMode.MANAGED,
+    base_url="wss://your-rime-endpoint.example.com",
+    model_id="mist",
+)
+```
+
+AgentKit serializes `credential_mode` at the top level of the Rime TTS configuration, alongside `vendor` and `params`. It omits the field when the option is not provided, preserving the existing BYOK request shape.
+
 ### `FishAudioTTS`
 
 | Parameter | Type | Required | Default | Description |

src/agora_agent/agentkit/__init__.py

@@ -97,6 +97,7 @@
     validate_tts_sample_rate,
 )
 from .constants import (
+    CredentialMode,
     DataChannel,
     AudioScenario,
     SilenceActionValues,
@@ -284,6 +285,7 @@
     "MllmTurnDetectionMode",
     "Labels",
     # Type-safe constants
+    "CredentialMode",
     "DataChannel",
     "AudioScenario",
     "SilenceActionValues",

src/agora_agent/agentkit/constants.py

@@ -3,6 +3,10 @@
 Use these instead of raw strings to avoid typos and get IDE autocomplete.
 """
 
+class CredentialMode:
+    MANAGED = "managed"
+    BYOK = "byok"
+
 # Data channel: "rtm" | "datastream"
 class DataChannel:
     RTM = "rtm"

src/agora_agent/agentkit/vendors/tts.py

@@ -1,9 +1,10 @@
-from typing import Any, Dict, List, Optional
+from typing import Any, Dict, List, Literal, Optional
 from urllib.parse import urlsplit
 
 from pydantic import ConfigDict, Field, field_validator, model_validator
 
 from .base import BaseTTS, CartesiaSampleRate, ElevenLabsSampleRate, GoogleTTSSampleRate, MicrosoftSampleRate
+from ..constants import CredentialMode
 from ..presets import MiniMaxPresetModels, OpenAITtsPresetModels
 
 
@@ -297,26 +298,44 @@ def to_config(self) -> Dict[str, Any]:
 class RimeTTS(BaseTTS):
     model_config = ConfigDict(extra="forbid")
 
-    key: str = Field(..., description="Rime API key")
-    speaker: str = Field(..., description="Speaker ID")
-    model_id: str = Field(..., description="Model ID")
+    key: Optional[str] = Field(default=None, description="Rime API key")
+    speaker: Optional[str] = Field(default=None, description="Speaker ID")
+    model_id: Optional[str] = Field(default=None, description="Model ID")
     base_url: Optional[str] = Field(default=None, description="WebSocket URL")
+    credential_mode: Optional[Literal["managed", "byok"]] = Field(default=None, description="Credential mode")
     skip_patterns: Optional[List[int]] = Field(default=None)
 
+    @model_validator(mode="after")
+    def _validate_credential_mode(self) -> "RimeTTS":
+        required: Dict[str, Optional[str]]
+        if self.credential_mode == CredentialMode.MANAGED:
+            required = {"base_url": self.base_url, "model_id": self.model_id}
+            mode = "credential_mode='managed'"
+        else:
+            required = {"key": self.key, "speaker": self.speaker, "model_id": self.model_id}
+            mode = "credential_mode='byok' or when credential_mode is omitted"
+
+        missing = [name for name, value in required.items() if not value]
+        if missing:
+            raise ValueError(f"RimeTTS requires {', '.join(missing)} for {mode}")
+        return self
+
     @property
     def sample_rate(self) -> Optional[int]:
         return None
 
     def to_config(self) -> Dict[str, Any]:
-        params: Dict[str, Any] = {
-            "api_key": self.key,
-            "speaker": self.speaker,
-            "modelId": self.model_id,
-        }
+        params: Dict[str, Any] = {"modelId": self.model_id}
+        if self.key is not None:
+            params["api_key"] = self.key
+        if self.speaker is not None:
+            params["speaker"] = self.speaker
         if self.base_url is not None:
             params["base_url"] = self.base_url
 
         result: Dict[str, Any] = {"vendor": "rime", "params": params}
+        if self.credential_mode is not None:
+            result["credential_mode"] = self.credential_mode
         if self.skip_patterns is not None:
             result["skip_patterns"] = self.skip_patterns
         return result

tests/custom/test_request_body.py

@@ -30,6 +30,7 @@
     AssemblyAISTT,
     AzureOpenAI,
     CartesiaTTS,
+    CredentialMode,
     CustomLLM,
     DeepgramSTT,
     DeepgramTTS,
@@ -1179,6 +1180,25 @@ def test_start_session_rime_tts_preserves_wire_aliases() -> None:
     assert "model_id" not in params
 
 
+def test_start_session_rime_tts_managed_credentials() -> None:
+    agent = full_agent_with_tts(
+        RimeTTS(
+            credential_mode=CredentialMode.MANAGED,
+            base_url="wss://users.rime.ai/ws",
+            model_id="mist",
+        )
+    )
+
+    call = start_session(agent)
+    properties = dump_wire(call["properties"])
+
+    assert properties["tts"]["credential_mode"] == "managed"
+    assert properties["tts"]["params"] == {
+        "modelId": "mist",
+        "base_url": "wss://users.rime.ai/ws",
+    }
+
+
 def test_start_session_murf_tts_preserves_wire_aliases() -> None:
     agent = full_agent_with_tts(MurfTTS(key="murf-key", voice_id="Ariana"))
 

tests/custom/test_root_exports.py

@@ -17,6 +17,7 @@ def test_root_exports_match_agentkit_for_common_symbols() -> None:
         "SpatiusAvatar",
         "AgentPresets",
         "generate_rtc_token",
+        "CredentialMode",
         "DataChannel",
     ):
         assert getattr(agora_agent, name) is getattr(agentkit, name)
@@ -29,6 +30,11 @@ def test_root_exports_fern_client_symbols() -> None:
     assert agora_agent.AsyncAgora is not None
 
 
+def test_credential_mode_constants() -> None:
+    assert agora_agent.CredentialMode.MANAGED == "managed"
+    assert agora_agent.CredentialMode.BYOK == "byok"
+
+
 def test_unknown_root_export_raises_attribute_error() -> None:
     with pytest.raises(AttributeError):
         _ = agora_agent.NotARealExportName
@@ -43,6 +49,7 @@ def test_dir_includes_agentkit_vendor_exports() -> None:
 
 
 def test_all_includes_agentkit_vendor_exports() -> None:
+    assert "CredentialMode" in agora_agent.__all__
     assert "DeepgramSTT" in agora_agent.__all__
     assert "MiniMaxCNTTS" in agora_agent.__all__
     assert "TencentSTT" in agora_agent.__all__

tests/custom/test_tts_vendors.py

@@ -1,6 +1,6 @@
 import pytest
 
-from agora_agent import AmazonTTS, CartesiaTTS, DeepgramTTS, ElevenLabsTTS, FishAudioTTS, GoogleTTS, HumeAITTS, MicrosoftTTS, MiniMaxTTS, MurfTTS, OpenAITTS, RimeTTS, SarvamTTS
+from agora_agent import AmazonTTS, CartesiaTTS, CredentialMode, DeepgramTTS, ElevenLabsTTS, FishAudioTTS, GoogleTTS, HumeAITTS, MicrosoftTTS, MiniMaxTTS, MurfTTS, OpenAITTS, RimeTTS, SarvamTTS
 from agora_agent.agents.types.start_agents_request_properties import StartAgentsRequestProperties
 from agora_agent.core.jsonable_encoder import jsonable_encoder
 from agora_agent.core.pydantic_utilities import parse_obj_as
@@ -170,6 +170,78 @@ def test_tts_managed_mode_validation_matches_core_shapes() -> None:
         )
 
 
+def test_rime_tts_managed_credential_mode_params() -> None:
+    config = RimeTTS(
+        credential_mode=CredentialMode.MANAGED,
+        base_url="wss://users.rime.ai/ws",
+        model_id="mist",
+    ).to_config()
+    assert config == {
+        "vendor": "rime",
+        "credential_mode": "managed",
+        "params": {
+            "modelId": "mist",
+            "base_url": "wss://users.rime.ai/ws",
+        },
+    }
+
+
+@pytest.mark.parametrize("credential_mode", [None, CredentialMode.BYOK])
+def test_rime_tts_byok_credential_mode_params(credential_mode) -> None:
+    config = RimeTTS(
+        credential_mode=credential_mode,
+        key="rime-key",
+        speaker="speaker",
+        model_id="mist",
+    ).to_config()
+    expected = {
+        "modelId": "mist",
+        "api_key": "rime-key",
+        "speaker": "speaker",
+    }
+    assert config["params"] == expected
+    if credential_mode is None:
+        assert "credential_mode" not in config
+    else:
+        assert config["credential_mode"] == credential_mode
+
+
+@pytest.mark.parametrize(
+    ("kwargs", "missing"),
+    [
+        ({"model_id": "mist"}, "base_url"),
+        ({"base_url": "wss://users.rime.ai/ws"}, "model_id"),
+        ({}, "base_url, model_id"),
+    ],
+)
+def test_rime_tts_managed_mode_requires_base_url_and_model_id(kwargs: dict, missing: str) -> None:
+    with pytest.raises(Exception, match=rf"RimeTTS requires {missing} for credential_mode='managed'"):
+        RimeTTS(credential_mode=CredentialMode.MANAGED, **kwargs)
+
+
+@pytest.mark.parametrize("credential_mode", [None, CredentialMode.BYOK])
+@pytest.mark.parametrize(
+    ("kwargs", "missing"),
+    [
+        ({"speaker": "speaker", "model_id": "mist"}, "key"),
+        ({"key": "rime-key", "model_id": "mist"}, "speaker"),
+        ({"key": "rime-key", "speaker": "speaker"}, "model_id"),
+    ],
+)
+def test_rime_tts_byok_mode_requires_key_speaker_and_model_id(
+    credential_mode,
+    kwargs: dict,
+    missing: str,
+) -> None:
+    with pytest.raises(Exception, match=rf"RimeTTS requires {missing}"):
+        RimeTTS(credential_mode=credential_mode, **kwargs)
+
+
+def test_rime_tts_rejects_unknown_credential_mode() -> None:
+    with pytest.raises(Exception, match="credential_mode"):
+        RimeTTS(credential_mode="unknown", base_url="wss://users.rime.ai/ws", model_id="mist")  # type: ignore[arg-type]
+
+
 def test_tts_wire_serialization_applies_fern_aliases() -> None:
     """Verify alias-sensitive TTS params keep the exact provider wire keys."""
     _BASE = dict(channel="ch", token="tok", agent_rtc_uid="1", remote_rtc_uids=["100"])
@@ -192,6 +264,23 @@ def test_tts_wire_serialization_applies_fern_aliases() -> None:
     assert "modelId" in rime_params, f"wire missing modelId, got: {list(rime_params)}"
     assert "model_id" not in rime_params
 
+    managed_rime_config = RimeTTS(
+        credential_mode=CredentialMode.MANAGED,
+        base_url="wss://users.rime.ai/ws",
+        model_id="mist",
+    ).to_config()
+    managed_rime_wire = jsonable_encoder(
+        parse_obj_as(StartAgentsRequestProperties, {**_BASE, "tts": managed_rime_config})
+    )
+    assert managed_rime_wire["tts"] == {
+        "vendor": "rime",
+        "credential_mode": "managed",
+        "params": {
+            "modelId": "mist",
+            "base_url": "wss://users.rime.ai/ws",
+        },
+    }
+
     murf_config = MurfTTS(key="murf-key", voice_id="Ariana").to_config()
     assert "voiceId" in murf_config["params"]
     murf_wire = jsonable_encoder(parse_obj_as(StartAgentsRequestProperties, {**_BASE, "tts": murf_config}))