feat(regen): add Flux TTS batch (REST) endpoint and agent latency report

.fern/metadata.json

@@ -16,7 +16,7 @@
       "skip_validation": true
     }
   },
-  "originGitCommit": "335af96251466afae7a9f71badb65c630a48626e",
+  "originGitCommit": "14c0ca94fcd0279d99570389d7b688dc3a18ea41",
   "originGitCommitIsDirty": true,
   "invokedBy": "manual",
   "sdkVersion": "7.4.1"

src/deepgram/__init__.py

@@ -222,9 +222,11 @@
         SpeakV1Response,
         SpeakV1SampleRate,
         SpeakV1Speed,
+        SpeakV2AcceptedResponse,
         SpeakV2Encoding,
         SpeakV2MipOptOut,
         SpeakV2Model,
+        SpeakV2Response,
         SpeakV2SampleRate,
         SpeakV2Tag,
         ThinkSettingsV1,
@@ -403,6 +405,7 @@
         SpeakSettingsV1Provider_DeepgramParams,
         SpeakSettingsV1Provider_ElevenLabsParams,
         SpeakSettingsV1Provider_OpenAiParams,
+        SpeakV2AcceptedResponseParams,
         ThinkSettingsV1ContextLengthParams,
         ThinkSettingsV1EndpointParams,
         ThinkSettingsV1FunctionsItemEndpointParams,
@@ -796,9 +799,12 @@
     "SpeakV1Response": ".types",
     "SpeakV1SampleRate": ".types",
     "SpeakV1Speed": ".types",
+    "SpeakV2AcceptedResponse": ".types",
+    "SpeakV2AcceptedResponseParams": ".requests",
     "SpeakV2Encoding": ".types",
     "SpeakV2MipOptOut": ".types",
     "SpeakV2Model": ".types",
+    "SpeakV2Response": ".types",
     "SpeakV2SampleRate": ".types",
     "SpeakV2Tag": ".types",
     "ThinkSettingsV1": ".types",
@@ -1246,9 +1252,12 @@ def __dir__():
     "SpeakV1Response",
     "SpeakV1SampleRate",
     "SpeakV1Speed",
+    "SpeakV2AcceptedResponse",
+    "SpeakV2AcceptedResponseParams",
     "SpeakV2Encoding",
     "SpeakV2MipOptOut",
     "SpeakV2Model",
+    "SpeakV2Response",
     "SpeakV2SampleRate",
     "SpeakV2Tag",
     "ThinkSettingsV1",

src/deepgram/agent/__init__.py

@@ -41,6 +41,8 @@
         AgentV1InjectionRefusedParams,
         AgentV1KeepAlive,
         AgentV1KeepAliveParams,
+        AgentV1LatencyReport,
+        AgentV1LatencyReportParams,
         AgentV1ListenUpdated,
         AgentV1ListenUpdatedParams,
         AgentV1PromptUpdated,
@@ -181,6 +183,8 @@
     "AgentV1InjectionRefusedParams": ".v1",
     "AgentV1KeepAlive": ".v1",
     "AgentV1KeepAliveParams": ".v1",
+    "AgentV1LatencyReport": ".v1",
+    "AgentV1LatencyReportParams": ".v1",
     "AgentV1ListenUpdated": ".v1",
     "AgentV1ListenUpdatedParams": ".v1",
     "AgentV1PromptUpdated": ".v1",
@@ -345,6 +349,8 @@ def __dir__():
     "AgentV1InjectionRefusedParams",
     "AgentV1KeepAlive",
     "AgentV1KeepAliveParams",
+    "AgentV1LatencyReport",
+    "AgentV1LatencyReportParams",
     "AgentV1ListenUpdated",
     "AgentV1ListenUpdatedParams",
     "AgentV1PromptUpdated",

src/deepgram/agent/v1/__init__.py

@@ -25,6 +25,7 @@
         AgentV1InjectUserMessage,
         AgentV1InjectionRefused,
         AgentV1KeepAlive,
+        AgentV1LatencyReport,
         AgentV1ListenUpdated,
         AgentV1PromptUpdated,
         AgentV1ReceiveFunctionCallResponse,
@@ -98,6 +99,7 @@
         AgentV1InjectUserMessageParams,
         AgentV1InjectionRefusedParams,
         AgentV1KeepAliveParams,
+        AgentV1LatencyReportParams,
         AgentV1ListenUpdatedParams,
         AgentV1PromptUpdatedParams,
         AgentV1ReceiveFunctionCallResponseParams,
@@ -183,6 +185,8 @@
     "AgentV1InjectionRefusedParams": ".requests",
     "AgentV1KeepAlive": ".types",
     "AgentV1KeepAliveParams": ".requests",
+    "AgentV1LatencyReport": ".types",
+    "AgentV1LatencyReportParams": ".requests",
     "AgentV1ListenUpdated": ".types",
     "AgentV1ListenUpdatedParams": ".requests",
     "AgentV1PromptUpdated": ".types",
@@ -347,6 +351,8 @@ def __dir__():
     "AgentV1InjectionRefusedParams",
     "AgentV1KeepAlive",
     "AgentV1KeepAliveParams",
+    "AgentV1LatencyReport",
+    "AgentV1LatencyReportParams",
     "AgentV1ListenUpdated",
     "AgentV1ListenUpdatedParams",
     "AgentV1PromptUpdated",

src/deepgram/agent/v1/requests/__init__.py

@@ -21,6 +21,7 @@
     from .agent_v1inject_user_message import AgentV1InjectUserMessageParams
     from .agent_v1injection_refused import AgentV1InjectionRefusedParams
     from .agent_v1keep_alive import AgentV1KeepAliveParams
+    from .agent_v1latency_report import AgentV1LatencyReportParams
     from .agent_v1listen_updated import AgentV1ListenUpdatedParams
     from .agent_v1prompt_updated import AgentV1PromptUpdatedParams
     from .agent_v1receive_function_call_response import AgentV1ReceiveFunctionCallResponseParams
@@ -103,6 +104,7 @@
     "AgentV1InjectUserMessageParams": ".agent_v1inject_user_message",
     "AgentV1InjectionRefusedParams": ".agent_v1injection_refused",
     "AgentV1KeepAliveParams": ".agent_v1keep_alive",
+    "AgentV1LatencyReportParams": ".agent_v1latency_report",
     "AgentV1ListenUpdatedParams": ".agent_v1listen_updated",
     "AgentV1PromptUpdatedParams": ".agent_v1prompt_updated",
     "AgentV1ReceiveFunctionCallResponseParams": ".agent_v1receive_function_call_response",
@@ -193,6 +195,7 @@ def __dir__():
     "AgentV1InjectUserMessageParams",
     "AgentV1InjectionRefusedParams",
     "AgentV1KeepAliveParams",
+    "AgentV1LatencyReportParams",
     "AgentV1ListenUpdatedParams",
     "AgentV1PromptUpdatedParams",
     "AgentV1ReceiveFunctionCallResponseParams",

src/deepgram/agent/v1/requests/agent_v1inject_agent_message.py

@@ -23,4 +23,5 @@ class AgentV1InjectAgentMessageParams(typing_extensions.TypedDict):
 
     * `default` — The agent speaks only if neither the user nor the agent is mid-turn. If a turn is in progress, the server replies with `InjectionRefused`.
     * `queue` — The message is appended after any already-queued `ConversationText` without interrupting the current agent turn or think response. If nothing is queued, the message plays immediately.
+    * `interrupt` — The agent immediately speaks. If the agent was already speaking, it interrupts the current speech and replaces it with the new message. If the user is speaking, the agent interrupts with the new message, but the user's continued speech triggers `UserStartedSpeaking`, which quickly interrupts the agent.
     """

src/deepgram/agent/v1/requests/agent_v1latency_report.py

@@ -0,0 +1,47 @@
+# This file was auto-generated by Fern from our API Definition.
+
+import typing
+
+import typing_extensions
+
+
+class AgentV1LatencyReportParams(typing_extensions.TypedDict):
+    type: typing.Literal["LatencyReport"]
+    """
+    Message type identifier for the latency report
+    """
+
+    stt_latency: typing_extensions.NotRequired[float]
+    """
+    Speech-to-text: time from audio received to transcript produced, in seconds
+    """
+
+    ttt_token_latency: typing_extensions.NotRequired[float]
+    """
+    Time to first token of any type (text, tool call, or thinking), in seconds
+    """
+
+    ttt_text_latency: typing_extensions.NotRequired[float]
+    """
+    Time to first text token from the LLM, in seconds
+    """
+
+    ttt_tool_latency: typing_extensions.NotRequired[float]
+    """
+    Time to first tool-call token from the LLM, in seconds
+    """
+
+    ttt_thinking_latency: typing_extensions.NotRequired[float]
+    """
+    Time to first thinking token from the LLM, in seconds
+    """
+
+    tts_latency: typing_extensions.NotRequired[float]
+    """
+    Text-to-speech: time from first text token to first audio byte, in seconds
+    """
+
+    total_latency: typing_extensions.NotRequired[float]
+    """
+    End-to-end: time from user utterance end to first audio byte, in seconds
+    """

src/deepgram/agent/v1/socket_client.py

@@ -18,6 +18,7 @@
 from .types.agent_v1inject_user_message import AgentV1InjectUserMessage
 from .types.agent_v1injection_refused import AgentV1InjectionRefused
 from .types.agent_v1keep_alive import AgentV1KeepAlive
+from .types.agent_v1latency_report import AgentV1LatencyReport
 from .types.agent_v1listen_updated import AgentV1ListenUpdated
 from .types.agent_v1prompt_updated import AgentV1PromptUpdated
 from .types.agent_v1receive_function_call_response import AgentV1ReceiveFunctionCallResponse
@@ -73,6 +74,7 @@ def _sanitize_numeric_types(obj: typing.Any) -> typing.Any:
     AgentV1ConversationText,
     AgentV1UserStartedSpeaking,
     AgentV1AgentThinking,
+    AgentV1LatencyReport,
     AgentV1FunctionCallRequest,
     AgentV1AgentStartedSpeaking,
     AgentV1AgentAudioDone,

src/deepgram/agent/v1/types/__init__.py

@@ -24,6 +24,7 @@
     from .agent_v1inject_user_message import AgentV1InjectUserMessage
     from .agent_v1injection_refused import AgentV1InjectionRefused
     from .agent_v1keep_alive import AgentV1KeepAlive
+    from .agent_v1latency_report import AgentV1LatencyReport
     from .agent_v1listen_updated import AgentV1ListenUpdated
     from .agent_v1prompt_updated import AgentV1PromptUpdated
     from .agent_v1receive_function_call_response import AgentV1ReceiveFunctionCallResponse
@@ -114,6 +115,7 @@
     "AgentV1InjectUserMessage": ".agent_v1inject_user_message",
     "AgentV1InjectionRefused": ".agent_v1injection_refused",
     "AgentV1KeepAlive": ".agent_v1keep_alive",
+    "AgentV1LatencyReport": ".agent_v1latency_report",
     "AgentV1ListenUpdated": ".agent_v1listen_updated",
     "AgentV1PromptUpdated": ".agent_v1prompt_updated",
     "AgentV1ReceiveFunctionCallResponse": ".agent_v1receive_function_call_response",
@@ -212,6 +214,7 @@ def __dir__():
     "AgentV1InjectUserMessage",
     "AgentV1InjectionRefused",
     "AgentV1KeepAlive",
+    "AgentV1LatencyReport",
     "AgentV1ListenUpdated",
     "AgentV1PromptUpdated",
     "AgentV1ReceiveFunctionCallResponse",

src/deepgram/agent/v1/types/agent_v1inject_agent_message.py

@@ -25,6 +25,7 @@ class AgentV1InjectAgentMessage(UncheckedBaseModel):
 
     * `default` — The agent speaks only if neither the user nor the agent is mid-turn. If a turn is in progress, the server replies with `InjectionRefused`.
     * `queue` — The message is appended after any already-queued `ConversationText` without interrupting the current agent turn or think response. If nothing is queued, the message plays immediately.
+    * `interrupt` — The agent immediately speaks. If the agent was already speaking, it interrupts the current speech and replaces it with the new message. If the user is speaking, the agent interrupts with the new message, but the user's continued speech triggers `UserStartedSpeaking`, which quickly interrupts the agent.
     """
 
     if IS_PYDANTIC_V2:

src/deepgram/agent/v1/types/agent_v1inject_agent_message_behavior.py

@@ -2,4 +2,4 @@
 
 import typing
 
-AgentV1InjectAgentMessageBehavior = typing.Union[typing.Literal["default", "queue"], typing.Any]
+AgentV1InjectAgentMessageBehavior = typing.Union[typing.Literal["default", "queue", "interrupt"], typing.Any]

src/deepgram/agent/v1/types/agent_v1latency_report.py

@@ -0,0 +1,58 @@
+# This file was auto-generated by Fern from our API Definition.
+
+import typing
+
+import pydantic
+from ....core.pydantic_utilities import IS_PYDANTIC_V2
+from ....core.unchecked_base_model import UncheckedBaseModel
+
+
+class AgentV1LatencyReport(UncheckedBaseModel):
+    type: typing.Literal["LatencyReport"] = pydantic.Field(default="LatencyReport")
+    """
+    Message type identifier for the latency report
+    """
+
+    stt_latency: typing.Optional[float] = pydantic.Field(default=None)
+    """
+    Speech-to-text: time from audio received to transcript produced, in seconds
+    """
+
+    ttt_token_latency: typing.Optional[float] = pydantic.Field(default=None)
+    """
+    Time to first token of any type (text, tool call, or thinking), in seconds
+    """
+
+    ttt_text_latency: typing.Optional[float] = pydantic.Field(default=None)
+    """
+    Time to first text token from the LLM, in seconds
+    """
+
+    ttt_tool_latency: typing.Optional[float] = pydantic.Field(default=None)
+    """
+    Time to first tool-call token from the LLM, in seconds
+    """
+
+    ttt_thinking_latency: typing.Optional[float] = pydantic.Field(default=None)
+    """
+    Time to first thinking token from the LLM, in seconds
+    """
+
+    tts_latency: typing.Optional[float] = pydantic.Field(default=None)
+    """
+    Text-to-speech: time from first text token to first audio byte, in seconds
+    """
+
+    total_latency: typing.Optional[float] = pydantic.Field(default=None)
+    """
+    End-to-end: time from user utterance end to first audio byte, in seconds
+    """
+
+    if IS_PYDANTIC_V2:
+        model_config: typing.ClassVar[pydantic.ConfigDict] = pydantic.ConfigDict(extra="allow", frozen=True)  # type: ignore # Pydantic v2
+    else:
+
+        class Config:
+            frozen = True
+            smart_union = True
+            extra = pydantic.Extra.allow

src/deepgram/requests/__init__.py

@@ -195,6 +195,7 @@
         SpeakSettingsV1Provider_ElevenLabsParams,
         SpeakSettingsV1Provider_OpenAiParams,
     )
+    from .speak_v2accepted_response import SpeakV2AcceptedResponseParams
     from .think_settings_v1 import ThinkSettingsV1Params
     from .think_settings_v1context_length import ThinkSettingsV1ContextLengthParams
     from .think_settings_v1endpoint import ThinkSettingsV1EndpointParams
@@ -368,6 +369,7 @@
     "SpeakSettingsV1Provider_DeepgramParams": ".speak_settings_v1provider",
     "SpeakSettingsV1Provider_ElevenLabsParams": ".speak_settings_v1provider",
     "SpeakSettingsV1Provider_OpenAiParams": ".speak_settings_v1provider",
+    "SpeakV2AcceptedResponseParams": ".speak_v2accepted_response",
     "ThinkSettingsV1ContextLengthParams": ".think_settings_v1context_length",
     "ThinkSettingsV1EndpointParams": ".think_settings_v1endpoint",
     "ThinkSettingsV1FunctionsItemEndpointParams": ".think_settings_v1functions_item_endpoint",
@@ -563,6 +565,7 @@ def __dir__():
     "SpeakSettingsV1Provider_DeepgramParams",
     "SpeakSettingsV1Provider_ElevenLabsParams",
     "SpeakSettingsV1Provider_OpenAiParams",
+    "SpeakV2AcceptedResponseParams",
     "ThinkSettingsV1ContextLengthParams",
     "ThinkSettingsV1EndpointParams",
     "ThinkSettingsV1FunctionsItemEndpointParams",

src/deepgram/requests/speak_v2accepted_response.py

@@ -0,0 +1,14 @@
+# This file was auto-generated by Fern from our API Definition.
+
+import typing_extensions
+
+
+class SpeakV2AcceptedResponseParams(typing_extensions.TypedDict):
+    """
+    Accepted response returned when a callback URL is supplied; the audio is delivered asynchronously to that URL.
+    """
+
+    request_id: str
+    """
+    Unique identifier for tracking the asynchronous request
+    """

src/deepgram/speak/v2/__init__.py

@@ -20,6 +20,8 @@
         SpeakV2SpeechStarted,
         SpeakV2Warning,
     )
+    from . import audio
+    from .audio import AudioGenerateRequestCallbackMethod, AudioGenerateRequestContainer, AudioGenerateRequestEncoding
     from .requests import (
         SpeakV2CloseParams,
         SpeakV2ConnectedParams,
@@ -34,6 +36,9 @@
         SpeakV2WarningParams,
     )
 _dynamic_imports: typing.Dict[str, str] = {
+    "AudioGenerateRequestCallbackMethod": ".audio",
+    "AudioGenerateRequestContainer": ".audio",
+    "AudioGenerateRequestEncoding": ".audio",
     "SpeakV2Close": ".types",
     "SpeakV2CloseParams": ".requests",
     "SpeakV2Connected": ".types",
@@ -57,6 +62,7 @@
     "SpeakV2SpeechStartedParams": ".requests",
     "SpeakV2Warning": ".types",
     "SpeakV2WarningParams": ".requests",
+    "audio": ".audio",
 }
 
 
@@ -82,6 +88,9 @@ def __dir__():
 
 
 __all__ = [
+    "AudioGenerateRequestCallbackMethod",
+    "AudioGenerateRequestContainer",
+    "AudioGenerateRequestEncoding",
     "SpeakV2Close",
     "SpeakV2CloseParams",
     "SpeakV2Connected",
@@ -105,4 +114,5 @@ def __dir__():
     "SpeakV2SpeechStartedParams",
     "SpeakV2Warning",
     "SpeakV2WarningParams",
+    "audio",
 ]