Release 16.4.0: add voice transcription

.fern/metadata.json

@@ -1,5 +1,5 @@
 {
-  "cliVersion": "5.50.1",
+  "cliVersion": "5.50.4",
   "generatorName": "fernapi/fern-python-sdk",
   "generatorVersion": "5.14.13",
   "generatorConfig": {
@@ -8,10 +8,10 @@
       "enabled": true
     }
   },
-  "originGitCommit": "fbaef4c66e97232edfaacb20d23d476aa286ecdd",
+  "originGitCommit": "e2c477c4583f8f9e7d0d1a5b5591c1aa8cabef10",
   "originGitCommitIsDirty": true,
   "invokedBy": "ci",
   "requestedVersion": "AUTO",
   "ciProvider": "unknown",
-  "sdkVersion": "16.3.0"
+  "sdkVersion": "16.4.0"
 }

changelog.md

@@ -1,3 +1,10 @@
+## [16.4.0] - 2026-06-23
+### Added
+- **`client.voice.voice.transcribe(...)`** — new sync and async method that accepts raw audio bytes (WAV, FLAC, MP3, OGG/WebM Opus) and returns a `TranscribeResponse` with the full transcript, supporting up to ~5 minutes of audio per request.
+- **`RawVoiceClient`** and **`AsyncRawVoiceClient`** — new clients under `phenoml.voice.voice` exposing `transcribe(...)` with support for `bytes`, `Iterator[bytes]`, or `AsyncIterator[bytes]` input and an optional BCP-47 `language` hint.
+- **`phenoml.voice.TranscribeResponse`** — new response type with a `transcript` field returned by the transcription endpoint.
+- **`phenoml.voice.errors`** — new typed error classes (`BadRequestError`, `UnauthorizedError`, `PaymentRequiredError`, `ContentTooLargeError`, `BadGatewayError`, `ServiceUnavailableError`, `GatewayTimeoutError`) raised by the voice service.
+
 ## [16.3.0] - 2026-06-18
 ### Added
 - **`phenoml.agent.errors.ConflictError`** — new `ApiError` subclass raised by `client.agent.chat.send(...)` and `client.agent.chat.stream(...)` for HTTP 409 responses when a session already has an active turn.

code-examples.json

@@ -2,8 +2,8 @@
   "metadata": {
     "language": "python",
     "packageName": "phenoml",
-    "sdkVersion": "16.3.0",
-    "specCommit": "fbaef4c66e97232edfaacb20d23d476aa286ecdd",
+    "sdkVersion": "16.4.0",
+    "specCommit": "e2c477c4583f8f9e7d0d1a5b5591c1aa8cabef10",
     "generatorName": "fernapi/fern-python-sdk"
   },
   "renderRules": {
@@ -3986,6 +3986,37 @@
         ]
       }
     },
+    "POST /transcribe": {
+      "httpMethod": "POST",
+      "httpPath": "/transcribe",
+      "request": {
+        "body": null
+      },
+      "response": {
+        "body": null
+      },
+      "render": {
+        "callTemplate": "client.voice.voice.transcribe({{__body__}})",
+        "params": [],
+        "body": {
+          "fieldSeparator": ", ",
+          "fields": [
+            {
+              "jsonKey": "language",
+              "fieldTemplate": "language={{value}}",
+              "kind": "list",
+              "required": false,
+              "items": {
+                "jsonKey": "",
+                "fieldTemplate": "{{value}}",
+                "kind": "string",
+                "required": true
+              }
+            }
+          ]
+        }
+      }
+    },
     "GET /workflows": {
       "httpMethod": "GET",
       "httpPath": "/workflows",

pyproject.toml

@@ -4,7 +4,7 @@ dynamic = ["version"]
 
 [tool.poetry]
 name = "phenoml"
-version = "16.3.0"
+version = "16.4.0"
 description = ""
 readme = "README.md"
 authors = []

reference.md

@@ -6448,6 +6448,85 @@ client.tools.mcp_tools.delete(
 </dl>
 
 
+</dd>
+</dl>
+</details>
+
+## Voice
+<details><summary><code>client.voice.voice.<a href="src/phenoml/voice/voice/client.py">transcribe</a>(...) -> TranscribeResponse</code></summary>
+<dl>
+<dd>
+
+#### 📝 Description
+
+<dl>
+<dd>
+
+<dl>
+<dd>
+
+Transcribes an uploaded audio recording and returns the transcript.
+Send the raw audio bytes as the request body; the audio format is
+detected automatically (WAV, FLAC, MP3, OGG/WebM Opus).
+
+Supports up to ~5 minutes of audio per request. This limit is on audio
+duration regardless of file size or format, so a compressed recording
+within the size limit can still be rejected for being too long. Pair the
+transcript with a downstream text step (e.g. `POST /lang2fhir/create`)
+to turn it into a FHIR resource.
+</dd>
+</dl>
+</dd>
+</dl>
+
+#### 🔌 Usage
+
+<dl>
+<dd>
+
+<dl>
+<dd>
+
+```python
+client.voice.voice.transcribe(...)
+```
+</dd>
+</dl>
+</dd>
+</dl>
+
+#### ⚙️ Parameters
+
+<dl>
+<dd>
+
+<dl>
+<dd>
+
+**request:** `typing.Union[bytes, typing.Iterator[bytes], typing.AsyncIterator[bytes]]` — Raw audio bytes (WAV, FLAC, MP3, or OGG/WebM Opus).
+
+</dd>
+</dl>
+
+<dl>
+<dd>
+
+**language:** `typing.Optional[typing.Union[str, typing.Sequence[str]]]` — BCP-47 language tag, repeatable for up to 4 candidate languages. Defaults to `en-US`.
+
+</dd>
+</dl>
+
+<dl>
+<dd>
+
+**request_options:** `typing.Optional[RequestOptions]` — Request-specific configuration.
+
+</dd>
+</dl>
+</dd>
+</dl>
+
+
 </dd>
 </dl>
 </details>

src/phenoml/__init__.py

@@ -17,6 +17,7 @@
         lang2fhir,
         summary,
         tools,
+        voice,
         workflows,
     )
     from ._default_clients import DefaultAioHttpClient, DefaultAsyncHttpxClient
@@ -40,6 +41,7 @@
     "lang2fhir": ".lang2fhir",
     "summary": ".summary",
     "tools": ".tools",
+    "voice": ".voice",
     "workflows": ".workflows",
 }
 
@@ -82,5 +84,6 @@ def __dir__():
     "lang2fhir",
     "summary",
     "tools",
+    "voice",
     "workflows",
 ]

src/phenoml/client.py

@@ -23,6 +23,7 @@
     from .lang2fhir.client import AsyncLang2FhirClient, Lang2FhirClient
     from .summary.client import AsyncSummaryClient, SummaryClient
     from .tools.client import AsyncToolsClient, ToolsClient
+    from .voice.client import AsyncVoiceClient, VoiceClient
     from .workflows.client import AsyncWorkflowsClient, WorkflowsClient
 
 
@@ -206,6 +207,7 @@ def __init__(
         self._lang2fhir: typing.Optional[Lang2FhirClient] = None
         self._summary: typing.Optional[SummaryClient] = None
         self._tools: typing.Optional[ToolsClient] = None
+        self._voice: typing.Optional[VoiceClient] = None
         self._workflows: typing.Optional[WorkflowsClient] = None
 
     @property
@@ -288,6 +290,14 @@ def tools(self):
             self._tools = ToolsClient(client_wrapper=self._client_wrapper)
         return self._tools
 
+    @property
+    def voice(self):
+        if self._voice is None:
+            from .voice.client import VoiceClient  # noqa: E402
+
+            self._voice = VoiceClient(client_wrapper=self._client_wrapper)
+        return self._voice
+
     @property
     def workflows(self):
         if self._workflows is None:
@@ -492,6 +502,7 @@ def __init__(
         self._lang2fhir: typing.Optional[AsyncLang2FhirClient] = None
         self._summary: typing.Optional[AsyncSummaryClient] = None
         self._tools: typing.Optional[AsyncToolsClient] = None
+        self._voice: typing.Optional[AsyncVoiceClient] = None
         self._workflows: typing.Optional[AsyncWorkflowsClient] = None
 
     @property
@@ -574,6 +585,14 @@ def tools(self):
             self._tools = AsyncToolsClient(client_wrapper=self._client_wrapper)
         return self._tools
 
+    @property
+    def voice(self):
+        if self._voice is None:
+            from .voice.client import AsyncVoiceClient  # noqa: E402
+
+            self._voice = AsyncVoiceClient(client_wrapper=self._client_wrapper)
+        return self._voice
+
     @property
     def workflows(self):
         if self._workflows is None:

src/phenoml/core/client_wrapper.py

@@ -29,12 +29,12 @@ def get_headers(self) -> typing.Dict[str, str]:
         import platform
 
         headers: typing.Dict[str, str] = {
-            "User-Agent": "phenoml/16.3.0",
+            "User-Agent": "phenoml/16.4.0",
             "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": "phenoml",
-            "X-Fern-SDK-Version": "16.3.0",
+            "X-Fern-SDK-Version": "16.4.0",
             **(self.get_custom_headers() or {}),
         }
         token = self._get_token()