feat(regen): flux stt numerals and aura-2 multilingual tts voices

.fern/metadata.json

@@ -16,8 +16,8 @@
       "skip_validation": true
     }
   },
-  "originGitCommit": "14c0ca94fcd0279d99570389d7b688dc3a18ea41",
+  "originGitCommit": "ff8fd2b74fdd5c081e9f111d59d3619f7e286dd8",
   "originGitCommitIsDirty": true,
   "invokedBy": "manual",
-  "sdkVersion": "7.5.0"
+  "sdkVersion": "7.5.1"
 }

.fernignore

@@ -101,6 +101,16 @@ src/deepgram/agent/v1/types/agent_v1settings_agent_context_listen_provider.py
 # resurrecting the invalid values. Frozen so Fern won't delete it again.
 src/deepgram/listen/v2/types/listen_v2close_stream_type.py
 
+# Backward-compat shim for the `stt_latency` removal from AgentV1LatencyReport (the
+# 2026-07-20 regen; API change in deepgram-docs #1006). LatencyReport is a
+# server-emitted (read-only) message, so re-adding the optional `stt_latency` field
+# has no request/wire impact — it keeps `report.stt_latency` resolving (to None, since
+# the server no longer emits it) instead of raising AttributeError, so the removal
+# stays a minor release rather than a major break. Remove the field and unfreeze when
+# it is intentionally retired in a future major release.
+src/deepgram/agent/v1/types/agent_v1latency_report.py
+src/deepgram/agent/v1/requests/agent_v1latency_report.py
+
 # Package __init__.py files that carry hand-applied legacy alias re-exports for the
 # above shims. Fern would otherwise regenerate these and strip the legacy entries on
 # every regen. Frozen to preserve the public-import surface for renamed types/params.
@@ -139,6 +149,8 @@ tests/custom/test_compat_aliases.py
 tests/custom/test_eot_thresholds_feature.py
 tests/custom/test_language_hint_compat.py
 tests/custom/test_language_hints_feature.py
+tests/custom/test_latency_report_stt_compat.py
+tests/custom/test_listen_v2_connect_wire.py
 tests/custom/test_listen_v2_regen_constraints.py
 tests/custom/test_query_encoder.py
 tests/custom/test_secure_logging.py

examples/14-transcription-live-websocket-v2.py

@@ -34,6 +34,9 @@
 try:
     with client.listen.v2.connect(
         model="flux-general-en",
+        # Flux STT numerals: format spoken numbers as digits (e.g. "one twenty" -> "120").
+        # Connection-time only — it cannot be toggled mid-stream via Configure.
+        numerals="true",
     ) as connection:
 
         def on_message(message: ListenV2SocketClientResponse) -> None:
@@ -46,7 +49,7 @@ def on_message(message: ListenV2SocketClientResponse) -> None:
                     print(f"  event: {message.get('event', '')}")
                     print(f"  turn_index: {message.get('turn_index', '')}")
             elif isinstance(message, ListenV2TurnInfo):
-                print(f"Received TurnInfo event")
+                print("Received TurnInfo event")
                 print(f"  transcript: {message.transcript}")
                 print(f"  event: {message.event}")
                 print(f"  turn_index: {message.turn_index}")

pyproject.toml

@@ -4,7 +4,7 @@ dynamic = ["version"]
 
 [tool.poetry]
 name = "deepgram-sdk"
-version = "7.5.0"
+version = "7.5.1"
 description = ""
 readme = "README.md"
 authors = []

reference.md

@@ -5465,6 +5465,30 @@ asyncio.run(main())
 <dl>
 <dd>
 
+**language_hint:** `typing.Optional[str]` — Language hint(s) to constrain and prioritize language detection (flux-general-multi)
+
+</dd>
+</dl>
+
+<dl>
+<dd>
+
+**profanity_filter:** `typing.Optional[str]` — Remove profanity from transcripts
+
+</dd>
+</dl>
+
+<dl>
+<dd>
+
+**numerals:** `typing.Optional[str]` — Convert numbers from written format to numerical format
+
+</dd>
+</dl>
+
+<dl>
+<dd>
+
 **mip_opt_out:** `typing.Optional[str]` — Opts out requests from the Deepgram Model Improvement Program
 
 </dd>

src/deepgram/__init__.py

@@ -171,6 +171,7 @@
         ListenV2LanguageHint,
         ListenV2MipOptOut,
         ListenV2Model,
+        ListenV2Numerals,
         ListenV2ProfanityFilter,
         ListenV2SampleRate,
         ListenV2Tag,
@@ -710,6 +711,7 @@
     "ListenV2LanguageHintParams": ".requests",
     "ListenV2MipOptOut": ".types",
     "ListenV2Model": ".types",
+    "ListenV2Numerals": ".types",
     "ListenV2ProfanityFilter": ".types",
     "ListenV2SampleRate": ".types",
     "ListenV2Tag": ".types",
@@ -1163,6 +1165,7 @@ def __dir__():
     "ListenV2LanguageHintParams",
     "ListenV2MipOptOut",
     "ListenV2Model",
+    "ListenV2Numerals",
     "ListenV2ProfanityFilter",
     "ListenV2SampleRate",
     "ListenV2Tag",

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

@@ -11,9 +11,13 @@ class AgentV1LatencyReportParams(typing_extensions.TypedDict):
     Message type identifier for the latency report
     """
 
+    # Backward-compat: mirrors the read-side shim in types/agent_v1latency_report.py.
+    # The API spec removed `stt_latency` (deepgram-docs #1006); kept here so the
+    # public request TypedDict shape is unchanged. No wire impact — the client never
+    # sends a LatencyReport. Remove and unfreeze in a future major.
     stt_latency: typing_extensions.NotRequired[float]
     """
-    Speech-to-text: time from audio received to transcript produced, in seconds
+    Deprecated. Speech-to-text latency, no longer reported by the server.
     """
 
     ttt_token_latency: typing_extensions.NotRequired[float]

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

@@ -13,9 +13,15 @@ class AgentV1LatencyReport(UncheckedBaseModel):
     Message type identifier for the latency report
     """
 
+    # Backward-compat: the API spec removed `stt_latency` from the LatencyReport
+    # schema (deepgram-docs #1006). LatencyReport is a server-emitted (read-only)
+    # message, so re-adding the field has no request/wire impact — it simply keeps
+    # `report.stt_latency` resolving (to None, since the server no longer emits it)
+    # instead of raising AttributeError, avoiding a major-version break. Remove this
+    # shim and unfreeze when the field is intentionally retired in a future major.
     stt_latency: typing.Optional[float] = pydantic.Field(default=None)
     """
-    Speech-to-text: time from audio received to transcript produced, in seconds
+    Deprecated. Speech-to-text latency, no longer reported by the server.
     """
 
     ttt_token_latency: typing.Optional[float] = pydantic.Field(default=None)

src/deepgram/core/client_wrapper.py

@@ -30,12 +30,12 @@ def get_headers(self) -> typing.Dict[str, str]:
         import platform
 
         headers: typing.Dict[str, str] = {
-            "User-Agent": "deepgram-sdk/7.5.0",  # x-release-please-version
+            "User-Agent": "deepgram-sdk/7.5.1",  # x-release-please-version
             "X-Fern-Language": "Python",
             "X-Fern-Runtime": f"python/{platform.python_version()}",
             "X-Fern-Platform": f"{platform.system().lower()}/{platform.release()}",
             "X-Fern-SDK-Name": "deepgram-sdk",
-            "X-Fern-SDK-Version": "7.5.0",  # x-release-please-version
+            "X-Fern-SDK-Version": "7.5.1",  # x-release-please-version
             **(self.get_custom_headers() or {}),
         }
         headers["Authorization"] = f"Token {self.api_key}"

src/deepgram/listen/v2/client.py

@@ -21,6 +21,7 @@
 from ...types.listen_v2eot_timeout_ms import ListenV2EotTimeoutMs
 from ...types.listen_v2mip_opt_out import ListenV2MipOptOut
 from ...types.listen_v2model import ListenV2Model
+from ...types.listen_v2numerals import ListenV2Numerals
 from ...types.listen_v2profanity_filter import ListenV2ProfanityFilter
 from ...types.listen_v2sample_rate import ListenV2SampleRate
 from ...types.listen_v2tag import ListenV2Tag
@@ -61,6 +62,7 @@ def connect(
         keyterm: typing.Optional[ListenV2KeytermParams] = None,
         language_hint: typing.Optional[ListenV2LanguageHintParams] = None,
         profanity_filter: typing.Optional[ListenV2ProfanityFilter] = None,
+        numerals: typing.Optional[ListenV2Numerals] = None,
         mip_opt_out: typing.Optional[ListenV2MipOptOut] = None,
         tag: typing.Optional[ListenV2Tag] = None,
         authorization: typing.Optional[str] = None,
@@ -90,6 +92,8 @@ def connect(
 
         profanity_filter : typing.Optional[ListenV2ProfanityFilter]
 
+        numerals : typing.Optional[ListenV2Numerals]
+
         mip_opt_out : typing.Optional[ListenV2MipOptOut]
 
         tag : typing.Optional[ListenV2Tag]
@@ -132,6 +136,7 @@ def connect(
                             direction="write",
                         ),
                         "profanity_filter": profanity_filter,
+                        "numerals": numerals,
                         "mip_opt_out": mip_opt_out,
                         "tag": tag,
                         **(
@@ -196,6 +201,7 @@ async def connect(
         keyterm: typing.Optional[ListenV2KeytermParams] = None,
         language_hint: typing.Optional[ListenV2LanguageHintParams] = None,
         profanity_filter: typing.Optional[ListenV2ProfanityFilter] = None,
+        numerals: typing.Optional[ListenV2Numerals] = None,
         mip_opt_out: typing.Optional[ListenV2MipOptOut] = None,
         tag: typing.Optional[ListenV2Tag] = None,
         authorization: typing.Optional[str] = None,
@@ -225,6 +231,8 @@ async def connect(
 
         profanity_filter : typing.Optional[ListenV2ProfanityFilter]
 
+        numerals : typing.Optional[ListenV2Numerals]
+
         mip_opt_out : typing.Optional[ListenV2MipOptOut]
 
         tag : typing.Optional[ListenV2Tag]
@@ -267,6 +275,7 @@ async def connect(
                             direction="write",
                         ),
                         "profanity_filter": profanity_filter,
+                        "numerals": numerals,
                         "mip_opt_out": mip_opt_out,
                         "tag": tag,
                         **(

src/deepgram/listen/v2/raw_client.py

@@ -21,6 +21,7 @@
 from ...types.listen_v2eot_timeout_ms import ListenV2EotTimeoutMs
 from ...types.listen_v2mip_opt_out import ListenV2MipOptOut
 from ...types.listen_v2model import ListenV2Model
+from ...types.listen_v2numerals import ListenV2Numerals
 from ...types.listen_v2profanity_filter import ListenV2ProfanityFilter
 from ...types.listen_v2sample_rate import ListenV2SampleRate
 from ...types.listen_v2tag import ListenV2Tag
@@ -49,6 +50,7 @@ def connect(
         keyterm: typing.Optional[ListenV2KeytermParams] = None,
         language_hint: typing.Optional[ListenV2LanguageHintParams] = None,
         profanity_filter: typing.Optional[ListenV2ProfanityFilter] = None,
+        numerals: typing.Optional[ListenV2Numerals] = None,
         mip_opt_out: typing.Optional[ListenV2MipOptOut] = None,
         tag: typing.Optional[ListenV2Tag] = None,
         authorization: typing.Optional[str] = None,
@@ -78,6 +80,8 @@ def connect(
 
         profanity_filter : typing.Optional[ListenV2ProfanityFilter]
 
+        numerals : typing.Optional[ListenV2Numerals]
+
         mip_opt_out : typing.Optional[ListenV2MipOptOut]
 
         tag : typing.Optional[ListenV2Tag]
@@ -120,6 +124,7 @@ def connect(
                             direction="write",
                         ),
                         "profanity_filter": profanity_filter,
+                        "numerals": numerals,
                         "mip_opt_out": mip_opt_out,
                         "tag": tag,
                         **(
@@ -173,6 +178,7 @@ async def connect(
         keyterm: typing.Optional[ListenV2KeytermParams] = None,
         language_hint: typing.Optional[ListenV2LanguageHintParams] = None,
         profanity_filter: typing.Optional[ListenV2ProfanityFilter] = None,
+        numerals: typing.Optional[ListenV2Numerals] = None,
         mip_opt_out: typing.Optional[ListenV2MipOptOut] = None,
         tag: typing.Optional[ListenV2Tag] = None,
         authorization: typing.Optional[str] = None,
@@ -202,6 +208,8 @@ async def connect(
 
         profanity_filter : typing.Optional[ListenV2ProfanityFilter]
 
+        numerals : typing.Optional[ListenV2Numerals]
+
         mip_opt_out : typing.Optional[ListenV2MipOptOut]
 
         tag : typing.Optional[ListenV2Tag]
@@ -244,6 +252,7 @@ async def connect(
                             direction="write",
                         ),
                         "profanity_filter": profanity_filter,
+                        "numerals": numerals,
                         "mip_opt_out": mip_opt_out,
                         "tag": tag,
                         **(

src/deepgram/speak/v1/audio/types/audio_generate_request_model.py

@@ -4,17 +4,17 @@
 
 AudioGenerateRequestModel = typing.Union[
     typing.Literal[
+        "aura-angus-en",
+        "aura-arcas-en",
         "aura-asteria-en",
-        "aura-luna-en",
-        "aura-stella-en",
         "aura-athena-en",
+        "aura-helios-en",
         "aura-hera-en",
+        "aura-luna-en",
         "aura-orion-en",
-        "aura-arcas-en",
-        "aura-perseus-en",
-        "aura-angus-en",
         "aura-orpheus-en",
-        "aura-helios-en",
+        "aura-perseus-en",
+        "aura-stella-en",
         "aura-zeus-en",
         "aura-2-amalthea-en",
         "aura-2-andromeda-en",
@@ -26,8 +26,8 @@
         "aura-2-atlas-en",
         "aura-2-aurora-en",
         "aura-2-callista-en",
-        "aura-2-cordelia-en",
         "aura-2-cora-en",
+        "aura-2-cordelia-en",
         "aura-2-delia-en",
         "aura-2-draco-en",
         "aura-2-electra-en",
@@ -57,16 +57,56 @@
         "aura-2-theia-en",
         "aura-2-vesta-en",
         "aura-2-zeus-en",
-        "aura-2-sirio-es",
-        "aura-2-nestor-es",
+        "aura-2-agustina-es",
+        "aura-2-alvaro-es",
+        "aura-2-antonia-es",
+        "aura-2-aquila-es",
         "aura-2-carina-es",
         "aura-2-celeste-es",
-        "aura-2-alvaro-es",
         "aura-2-diana-es",
-        "aura-2-aquila-es",
-        "aura-2-selena-es",
         "aura-2-estrella-es",
+        "aura-2-gloria-es",
         "aura-2-javier-es",
+        "aura-2-luciano-es",
+        "aura-2-nestor-es",
+        "aura-2-olivia-es",
+        "aura-2-selena-es",
+        "aura-2-silvia-es",
+        "aura-2-sirio-es",
+        "aura-2-valerio-es",
+        "aura-2-aurelia-de",
+        "aura-2-elara-de",
+        "aura-2-fabian-de",
+        "aura-2-julius-de",
+        "aura-2-kara-de",
+        "aura-2-lara-de",
+        "aura-2-viktoria-de",
+        "aura-2-beatrix-nl",
+        "aura-2-cornelia-nl",
+        "aura-2-daphne-nl",
+        "aura-2-hestia-nl",
+        "aura-2-lars-nl",
+        "aura-2-leda-nl",
+        "aura-2-rhea-nl",
+        "aura-2-roman-nl",
+        "aura-2-sander-nl",
+        "aura-2-agathe-fr",
+        "aura-2-hector-fr",
+        "aura-2-cesare-it",
+        "aura-2-cinzia-it",
+        "aura-2-demetra-it",
+        "aura-2-dionisio-it",
+        "aura-2-elio-it",
+        "aura-2-flavio-it",
+        "aura-2-livia-it",
+        "aura-2-maia-it",
+        "aura-2-melia-it",
+        "aura-2-perseo-it",
+        "aura-2-ama-ja",
+        "aura-2-ebisu-ja",
+        "aura-2-fujin-ja",
+        "aura-2-izanami-ja",
+        "aura-2-uzume-ja",
     ],
     typing.Any,
 ]

src/deepgram/speak/v2/audio/client.py

@@ -50,7 +50,7 @@ def generate(
         Parameters
         ----------
         model : str
-            Flux TTS model used to synthesize the submitted text, in the form `flux-{voice}-{language}` (for example, `flux-alexis-en`). Required; unlike the v1 (Aura) endpoint there is no default and only flux models are accepted. English-only at launch. The voice catalog is not yet finalized, so this is typed as a free-form string rather than a closed enum; it can be locked to an enum (matching STT Flux `ListenV2Model`) once the GA catalog is confirmed.
+            Flux TTS model used to synthesize the submitted text, in the form `flux-{voice}-{language}` (for example, `flux-alexis-en`). Required; unlike the v1 (Aura) endpoint there is no default and only flux models are accepted. English-only at launch.
 
         text : str
             The text content to be converted to speech. The server normalizes and preprocesses the text (e.g. stripping inline controls) before synthesis.
@@ -156,7 +156,7 @@ async def generate(
         Parameters
         ----------
         model : str
-            Flux TTS model used to synthesize the submitted text, in the form `flux-{voice}-{language}` (for example, `flux-alexis-en`). Required; unlike the v1 (Aura) endpoint there is no default and only flux models are accepted. English-only at launch. The voice catalog is not yet finalized, so this is typed as a free-form string rather than a closed enum; it can be locked to an enum (matching STT Flux `ListenV2Model`) once the GA catalog is confirmed.
+            Flux TTS model used to synthesize the submitted text, in the form `flux-{voice}-{language}` (for example, `flux-alexis-en`). Required; unlike the v1 (Aura) endpoint there is no default and only flux models are accepted. English-only at launch.
 
         text : str
             The text content to be converted to speech. The server normalizes and preprocesses the text (e.g. stripping inline controls) before synthesis.