vocence-plugins 0.1.0__tar.gz

vocence_plugins-0.1.0/.gitignore

@@ -0,0 +1,17 @@
+# Python
+__pycache__/
+*.py[cod]
+*.egg-info/
+*.egg
+.eggs/
+build/
+dist/
+.venv/
+
+# IDE
+.vscode/
+.idea/
+*.swp
+
+# OS
+.DS_Store

vocence_plugins-0.1.0/LICENSE

@@ -0,0 +1,200 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for describing the origin of the Work and
+      reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Support. While redistributing the Work or
+      Derivative Works thereof, You may accept support, warranty,
+      indemnity, or other liability obligations and/or rights consistent
+      with this License. However, in accepting such obligations, You may
+      act only on Your own behalf and on Your sole responsibility, not on
+      behalf of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or support.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright 2026 Vocence
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+   implied. See the License for the specific language governing permissions
+   and limitations under the License.

vocence_plugins-0.1.0/PKG-INFO

@@ -0,0 +1,106 @@
+Metadata-Version: 2.4
+Name: vocence-plugins
+Version: 0.1.0
+Summary: Vocence voice plug-ins — drop custom-cloned voices and streaming speech recognition into your real-time voice agents.
+Project-URL: Homepage, https://www.vocence.ai
+Project-URL: Documentation, https://www.vocence.ai/docs/sdk-agents
+Project-URL: Repository, https://github.com/concil859856/vocence-plugins
+Project-URL: Issues, https://github.com/concil859856/vocence-plugins/issues
+Author-email: Vocence <space@vocence.ai>
+License-Expression: Apache-2.0
+License-File: LICENSE
+Keywords: realtime,stt,tts,vocence,voice-agents,voice-cloning
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Multimedia :: Sound/Audio :: Speech
+Classifier: Topic :: Software Development :: Libraries :: Python Modules
+Requires-Python: >=3.11
+Requires-Dist: aiohttp>=3.9
+Requires-Dist: videosdk-agents>=0.1.0
+Description-Content-Type: text/markdown
+
+# vocence-plugins
+
+Vocence voice plug-ins for real-time agent pipelines — drop in **Vocence custom voices** for TTS and **Vocence streaming speech recognition** for STT.
+
+- **`VocenceTTS`** — streaming text-to-speech with the Vocence voice library (cloned, designed, and built-in speakers). One persistent connection per session, sub-second TTFB on warm connections, PCM16LE @ 24 kHz output.
+- **`VocenceSTT`** — streaming speech-to-text with interim + final transcripts, optional speech / silence events for VAD integration, and language auto-detect.
+
+> Status: 0.1.0 — public alpha.
+
+## Install
+
+```bash
+pip install vocence-plugins
+```
+
+The plug-ins conform to the standard TTS / STT abstract interfaces, so they slot into any compatible voice-agent framework.
+
+## API key
+
+Get one at https://www.vocence.ai/account/developer. Requires the Premium plan.
+
+```bash
+export VOCENCE_API_KEY=voc_live_...
+```
+
+Or pass it directly: `VocenceTTS(api_key="voc_live_...", voice="...")`.
+
+## Quickstart
+
+```python
+from vocence_plugins import VocenceTTS, VocenceSTT
+
+tts = VocenceTTS(voice="design-aria", language="English")
+stt = VocenceSTT(language="English")
+
+# Wire into your agent pipeline as the TTS / STT components.
+```
+
+The plug-ins handle the WebSocket lifecycle, reconnection, and audio framing — your code just sees text in and audio out (TTS), or audio in and transcript events out (STT).
+
+## Plugin reference
+
+### `VocenceTTS(*, api_key=None, voice, language=None, base_url=...)`
+
+Streaming TTS over the Vocence voice service. One WebSocket reused across many `synthesize()` calls in the same session, closed on `aclose()`.
+
+| Arg | Default | Notes |
+|---|---|---|
+| `api_key` | `VOCENCE_API_KEY` env | Required (`voc_live_…`). |
+| `voice` | — | Required. Either a built-in slug (`"design-aria"`, `"design-jasper"`, …) or the numeric id of a saved designed / cloned voice. |
+| `language` | `None` | Optional hint sent on every speak. |
+| `base_url` | `https://api.vocence.ai` | Override for staging / self-hosted. |
+
+Audio output: PCM16LE @ 24 kHz, mono.
+
+### `VocenceSTT(*, api_key=None, language="auto", sample_rate=16000, enable_partials=True, vad_events=True, base_url=...)`
+
+Streaming STT. Lazy-opens a WebSocket on the first audio frame, runs a background reader that translates events into the framework's standard transcript event shape (interim, final, speech-start, speech-end).
+
+| Arg | Default | Notes |
+|---|---|---|
+| `api_key` | `VOCENCE_API_KEY` env | Required. |
+| `language` | `"auto"` | ISO-639-1 (`"en"`), full name (`"English"`), or `"auto"`. Normalized to ISO before send. |
+| `sample_rate` | 16000 | PCM16LE mono input. |
+| `enable_partials` | `True` | Stream interim hypotheses as the model refines. |
+| `vad_events` | `True` | Emit speech-start / silence events from the internal VAD. |
+
+## Compared to the Vocence Python SDK
+
+| Use case | Use |
+|---|---|
+| Talk to a Vocence-hosted voice agent (REST + WebSocket to our service) | [`vocence`](https://pypi.org/project/vocence/) |
+| Build your own agent pipeline with Vocence voices + recognition | This package |
+
+The two don't overlap — different products for different use cases. Both authenticate with the same `voc_live_…` key.
+
+## License
+
+Apache-2.0.

vocence_plugins-0.1.0/README.md

@@ -0,0 +1,79 @@
+# vocence-plugins
+
+Vocence voice plug-ins for real-time agent pipelines — drop in **Vocence custom voices** for TTS and **Vocence streaming speech recognition** for STT.
+
+- **`VocenceTTS`** — streaming text-to-speech with the Vocence voice library (cloned, designed, and built-in speakers). One persistent connection per session, sub-second TTFB on warm connections, PCM16LE @ 24 kHz output.
+- **`VocenceSTT`** — streaming speech-to-text with interim + final transcripts, optional speech / silence events for VAD integration, and language auto-detect.
+
+> Status: 0.1.0 — public alpha.
+
+## Install
+
+```bash
+pip install vocence-plugins
+```
+
+The plug-ins conform to the standard TTS / STT abstract interfaces, so they slot into any compatible voice-agent framework.
+
+## API key
+
+Get one at https://www.vocence.ai/account/developer. Requires the Premium plan.
+
+```bash
+export VOCENCE_API_KEY=voc_live_...
+```
+
+Or pass it directly: `VocenceTTS(api_key="voc_live_...", voice="...")`.
+
+## Quickstart
+
+```python
+from vocence_plugins import VocenceTTS, VocenceSTT
+
+tts = VocenceTTS(voice="design-aria", language="English")
+stt = VocenceSTT(language="English")
+
+# Wire into your agent pipeline as the TTS / STT components.
+```
+
+The plug-ins handle the WebSocket lifecycle, reconnection, and audio framing — your code just sees text in and audio out (TTS), or audio in and transcript events out (STT).
+
+## Plugin reference
+
+### `VocenceTTS(*, api_key=None, voice, language=None, base_url=...)`
+
+Streaming TTS over the Vocence voice service. One WebSocket reused across many `synthesize()` calls in the same session, closed on `aclose()`.
+
+| Arg | Default | Notes |
+|---|---|---|
+| `api_key` | `VOCENCE_API_KEY` env | Required (`voc_live_…`). |
+| `voice` | — | Required. Either a built-in slug (`"design-aria"`, `"design-jasper"`, …) or the numeric id of a saved designed / cloned voice. |
+| `language` | `None` | Optional hint sent on every speak. |
+| `base_url` | `https://api.vocence.ai` | Override for staging / self-hosted. |
+
+Audio output: PCM16LE @ 24 kHz, mono.
+
+### `VocenceSTT(*, api_key=None, language="auto", sample_rate=16000, enable_partials=True, vad_events=True, base_url=...)`
+
+Streaming STT. Lazy-opens a WebSocket on the first audio frame, runs a background reader that translates events into the framework's standard transcript event shape (interim, final, speech-start, speech-end).
+
+| Arg | Default | Notes |
+|---|---|---|
+| `api_key` | `VOCENCE_API_KEY` env | Required. |
+| `language` | `"auto"` | ISO-639-1 (`"en"`), full name (`"English"`), or `"auto"`. Normalized to ISO before send. |
+| `sample_rate` | 16000 | PCM16LE mono input. |
+| `enable_partials` | `True` | Stream interim hypotheses as the model refines. |
+| `vad_events` | `True` | Emit speech-start / silence events from the internal VAD. |
+
+## Compared to the Vocence Python SDK
+
+| Use case | Use |
+|---|---|
+| Talk to a Vocence-hosted voice agent (REST + WebSocket to our service) | [`vocence`](https://pypi.org/project/vocence/) |
+| Build your own agent pipeline with Vocence voices + recognition | This package |
+
+The two don't overlap — different products for different use cases. Both authenticate with the same `voc_live_…` key.
+
+## License
+
+Apache-2.0.

vocence_plugins-0.1.0/pyproject.toml

@@ -0,0 +1,43 @@
+[build-system]
+requires = ["hatchling>=1.21"]
+build-backend = "hatchling.build"
+
+[project]
+name = "vocence-plugins"
+dynamic = ["version"]
+description = "Vocence voice plug-ins — drop custom-cloned voices and streaming speech recognition into your real-time voice agents."
+readme = "README.md"
+license = "Apache-2.0"
+requires-python = ">=3.11"
+authors = [{ name = "Vocence", email = "space@vocence.ai" }]
+keywords = ["vocence", "voice-agents", "tts", "stt", "voice-cloning", "realtime"]
+classifiers = [
+    "Development Status :: 4 - Beta",
+    "Intended Audience :: Developers",
+    "License :: OSI Approved :: Apache Software License",
+    "Operating System :: OS Independent",
+    "Programming Language :: Python :: 3",
+    "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
+    "Programming Language :: Python :: 3.13",
+    "Topic :: Multimedia :: Sound/Audio :: Speech",
+    "Topic :: Software Development :: Libraries :: Python Modules",
+]
+dependencies = [
+    "aiohttp>=3.9",
+    # The plug-ins implement the abstract interfaces of a compatible
+    # real-time agent framework. Required at runtime.
+    "videosdk-agents>=0.1.0",
+]
+
+[project.urls]
+Homepage = "https://www.vocence.ai"
+Documentation = "https://www.vocence.ai/docs/sdk-agents"
+Repository = "https://github.com/concil859856/vocence-plugins"
+Issues = "https://github.com/concil859856/vocence-plugins/issues"
+
+[tool.hatch.version]
+path = "src/vocence_plugins/__init__.py"
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/vocence_plugins"]

vocence_plugins-0.1.0/src/vocence_plugins/__init__.py

@@ -0,0 +1,37 @@
+"""Vocence voice plug-ins.
+
+Drop-in components that bring **Vocence custom voices** and
+**Vocence streaming speech recognition** into your real-time voice
+agent pipeline. Authenticated with the standard Vocence
+``voc_live_…`` developer key — the same one you use across the rest
+of the Vocence platform.
+
+The headline component is :class:`VocenceTTS` — streaming
+text-to-speech with the Vocence voice library: cloned voices,
+designed voices, and the built-in speaker catalog. :class:`VocenceSTT`
+streams audio in and transcripts out.
+
+Example
+-------
+
+>>> from vocence_plugins import VocenceTTS, VocenceSTT
+>>>
+>>> tts = VocenceTTS(api_key="voc_live_...", voice="design-aria")
+>>> stt = VocenceSTT(api_key="voc_live_...", language="English")
+
+Plug them into the agent framework of your choice — both classes
+conform to the standard TTS / STT abstract interfaces.
+
+See https://www.vocence.ai/docs/sdk-agents for the full guide.
+"""
+from __future__ import annotations
+
+__version__ = "0.1.0"
+
+from .tts import VocenceTTS
+from .stt import VocenceSTT
+
+__all__ = [
+    "VocenceTTS",
+    "VocenceSTT",
+]

vocence_plugins-0.1.0/src/vocence_plugins/_lang.py

@@ -0,0 +1,44 @@
+"""Language-name → ISO-639-1 mapping for the Vocence STT pod.
+
+The pod's wire protocol expects an ISO code (or ``"auto"``); full
+names like ``"English"`` silently degrade to auto-detect, which then
+mis-classifies short utterances. Mirrors the helper used inside the
+Vocence backend so plugin users get the same behavior the hosted
+service does.
+"""
+from __future__ import annotations
+
+from typing import Optional
+
+_NAME_TO_ISO = {
+    "English": "en",
+    "Chinese": "zh",
+    "Japanese": "ja",
+    "Korean": "ko",
+    "German": "de",
+    "French": "fr",
+    "Russian": "ru",
+    "Portuguese": "pt",
+    "Spanish": "es",
+    "Italian": "it",
+}
+
+
+def to_iso_639_1(language: Optional[str]) -> str:
+    """Resolve any input form to the ISO-639-1 code the STT pod wants.
+
+    Accepts agent-config full names (``"English"``), already-ISO codes
+    (``"en"``), and the sentinel ``"auto"`` / ``None``. Falls back to
+    ``"auto"`` for unknown input so the pod is never sent a string it
+    doesn't understand.
+    """
+    if not language:
+        return "auto"
+    s = language.strip()
+    if s.lower() == "auto":
+        return "auto"
+    if s in _NAME_TO_ISO:
+        return _NAME_TO_ISO[s]
+    if len(s) == 2 and s.isalpha():
+        return s.lower()
+    return "auto"

vocence_plugins-0.1.0/src/vocence_plugins/stt.py

@@ -0,0 +1,313 @@
+"""VocenceSTT — streaming speech-to-text with Vocence recognition.
+
+Conforms to the standard STT abstract interface used by real-time
+agent pipelines, so it slots in alongside any compatible
+``Pipeline(stt=...)``. The plug-in handles connection lifecycle,
+audio framing, and event translation — callers just see audio in
+and standard transcript events out.
+
+Audio input: PCM16LE @ 16 kHz, mono. One persistent connection is
+lazily opened on the first audio frame and torn down on
+``aclose()``. A background reader task translates incoming events
+into the framework's standard transcript event shape.
+"""
+from __future__ import annotations
+
+import asyncio
+import json
+import logging
+import os
+from typing import Any, Optional
+from urllib.parse import urlparse
+
+import aiohttp
+
+from videosdk.agents import (  # type: ignore[import-not-found]
+    STT,
+    STTResponse,
+    SpeechData,
+    SpeechEventType,
+)
+
+from ._lang import to_iso_639_1
+
+logger = logging.getLogger(__name__)
+
+
+_DEFAULT_BASE_URL = os.environ.get("VOCENCE_BASE_URL", "https://api.vocence.ai")
+_DEFAULT_SAMPLE_RATE = 16_000
+_DEFAULT_TIMEOUT_SEC = 30.0
+
+
+def _ws_url_from_base(base_url: str) -> str:
+    """Translate ``https://api.vocence.ai`` → ``wss://api.vocence.ai/v1/stt/stream``."""
+    parsed = urlparse(base_url.rstrip("/"))
+    scheme = "wss" if parsed.scheme == "https" else "ws"
+    return f"{scheme}://{parsed.netloc}{parsed.path}/v1/stt/stream"
+
+
+class VocenceSTT(STT):
+    """Streaming STT plugin backed by the Vocence recognition service.
+
+    Parameters
+    ----------
+    api_key:
+        Vocence developer key (``voc_live_…``). Falls back to the
+        ``VOCENCE_API_KEY`` env var. Required.
+    language:
+        Spoken language. Accepts ISO-639-1 codes (``"en"``), full
+        names (``"English"``), or ``"auto"`` for auto-detect.
+        Default ``"auto"``. Normalized to ISO upfront to avoid
+        silent degradation to auto-detect for unrecognized forms.
+    sample_rate:
+        Audio sample rate. Default 16 kHz (mono PCM16LE). Only
+        16 kHz is accepted today; argument is kept for
+        forward-compatibility.
+    enable_partials:
+        Stream interim hypotheses as recognition refines. Default
+        ``True`` — needed for any responsive UI; disable only for
+        batch / archival use cases that just want finals.
+    vad_events:
+        Emit speech-start / silence events from the internal VAD so
+        the orchestrator can drive its interrupt / speech_started
+        hooks off these. Default ``True``. Independent of any
+        external VAD plugin you wire alongside.
+    base_url:
+        Override the default ``https://api.vocence.ai``.
+    forward_interim_transcripts:
+        Whether to surface interim text to the user UI (passed
+        through to the standard STT base initializer).
+    """
+
+    def __init__(
+        self,
+        *,
+        api_key: Optional[str] = None,
+        language: str = "auto",
+        sample_rate: int = _DEFAULT_SAMPLE_RATE,
+        enable_partials: bool = True,
+        vad_events: bool = True,
+        base_url: str = _DEFAULT_BASE_URL,
+        forward_interim_transcripts: bool = False,
+        **kwargs: Any,
+    ) -> None:
+        super().__init__(forward_interim_transcripts=forward_interim_transcripts)
+        self.api_key = api_key or os.environ.get("VOCENCE_API_KEY")
+        if not self.api_key:
+            raise ValueError(
+                "Vocence API key required — pass api_key= or set VOCENCE_API_KEY"
+            )
+        self.language = language
+        self.sample_rate = sample_rate
+        self.enable_partials = enable_partials
+        self.vad_events = vad_events
+        self.base_url = base_url
+        self._ws_url = _ws_url_from_base(base_url)
+
+        self._session: aiohttp.ClientSession | None = None
+        self._ws: aiohttp.ClientWebSocketResponse | None = None
+        self._reader_task: asyncio.Task | None = None
+        self._connect_lock = asyncio.Lock()
+        self._closed = False
+
+    # ----- abstract overrides ---------------------------------------------
+
+    async def process_audio(
+        self,
+        audio_frames: bytes,
+        language: Optional[str] = None,
+        **kwargs: Any,
+    ) -> None:
+        """Send one PCM16LE frame to the pod. Lazy-opens the WS on
+        first call. The framework calls this once per audio frame
+        (typically 20-40 ms), so the hot path is just a bytes send.
+
+        ``language`` is accepted for API symmetry but ignored once
+        the connection is established — the pod's ``start`` frame
+        binds the language for the session. Construct a new
+        ``VocenceSTT`` for a different language.
+        """
+        if self._closed:
+            return
+        if self._ws is None:
+            await self._ensure_connection()
+        ws = self._ws
+        if ws is None or ws.closed:
+            return
+        try:
+            await ws.send_bytes(audio_frames)
+        except Exception as exc:  # noqa: BLE001
+            logger.warning("VocenceSTT send_bytes failed: %s", exc)
+            await self._teardown_ws()
+
+    async def flush(self) -> None:
+        """Ask the pod to finalize its current partial as soon as
+        possible. Useful at end-of-utterance when the orchestrator
+        knows the turn is over but the pod hasn't auto-emitted a
+        final yet."""
+        ws = self._ws
+        if ws is None or ws.closed:
+            return
+        try:
+            await ws.send_str(json.dumps({"type": "commit"}))
+        except Exception as exc:  # noqa: BLE001
+            logger.debug("VocenceSTT commit failed: %s", exc)
+
+    async def aclose(self) -> None:
+        """Close the WebSocket + HTTP session. Idempotent."""
+        self._closed = True
+        if self._reader_task is not None and not self._reader_task.done():
+            self._reader_task.cancel()
+            try:
+                await self._reader_task
+            except (asyncio.CancelledError, Exception):
+                pass
+            self._reader_task = None
+        await self._teardown_ws()
+
+    # ----- internals ------------------------------------------------------
+
+    async def _ensure_connection(self) -> None:
+        """Open the WS, send ``start``, wait for ``ready``, kick the reader."""
+        async with self._connect_lock:
+            if self._ws is not None and not self._ws.closed:
+                return
+            if self._session is None or self._session.closed:
+                self._session = aiohttp.ClientSession(
+                    timeout=aiohttp.ClientTimeout(total=_DEFAULT_TIMEOUT_SEC),
+                )
+            headers = {"Authorization": f"Bearer {self.api_key}"}
+            self._ws = await self._session.ws_connect(self._ws_url, headers=headers)
+            start_payload = {
+                "type": "start",
+                "language": to_iso_639_1(self.language),
+                "sample_rate": self.sample_rate,
+                "encoding": "pcm_s16le",
+                "enable_partials": self.enable_partials,
+                "vad_events": self.vad_events,
+            }
+            await self._ws.send_str(json.dumps(start_payload))
+            ready = await self._ws.receive(timeout=_DEFAULT_TIMEOUT_SEC)
+            if ready.type != aiohttp.WSMsgType.TEXT:
+                raise RuntimeError(
+                    f"VocenceSTT: expected ready frame, got {ready.type}"
+                )
+            data = json.loads(ready.data)
+            mtype = (data.get("type") or "").lower()
+            if mtype == "error":
+                raise RuntimeError(
+                    f"VocenceSTT connect rejected: "
+                    f"{data.get('code')}: {data.get('message')}"
+                )
+            if mtype != "ready":
+                raise RuntimeError(
+                    f"VocenceSTT: unexpected first frame {mtype!r}"
+                )
+            self._reader_task = asyncio.create_task(
+                self._read_loop(), name="vocence_stt_reader"
+            )
+
+    async def _read_loop(self) -> None:
+        """Background task: translate pod events → STTResponse callbacks."""
+        ws = self._ws
+        if ws is None:
+            return
+        try:
+            async for msg in ws:
+                if msg.type != aiohttp.WSMsgType.TEXT:
+                    if msg.type in (
+                        aiohttp.WSMsgType.CLOSED,
+                        aiohttp.WSMsgType.CLOSE,
+                        aiohttp.WSMsgType.ERROR,
+                    ):
+                        return
+                    continue
+                try:
+                    data = json.loads(msg.data)
+                except json.JSONDecodeError:
+                    continue
+                response = self._translate_event(data)
+                if response is None:
+                    continue
+                cb = self._transcript_callback
+                if cb is None:
+                    continue
+                try:
+                    await cb(response)
+                except Exception as exc:  # noqa: BLE001
+                    logger.warning("VocenceSTT callback raised: %s", exc)
+        except asyncio.CancelledError:
+            return
+        except Exception as exc:  # noqa: BLE001
+            logger.warning("VocenceSTT reader loop crashed: %s", exc)
+            self.emit("error", str(exc))
+
+    def _translate_event(self, data: dict) -> STTResponse | None:
+        """Map one pod event to the the framework's STTResponse shape.
+
+        Returns ``None`` for events we drop (errors are logged via
+        EventEmitter instead of being passed to the orchestrator).
+        """
+        mtype = (data.get("type") or "").lower()
+        if mtype == "partial":
+            text = (data.get("text") or "").strip()
+            if not text:
+                return None
+            return STTResponse(
+                event_type=SpeechEventType.INTERIM,
+                data=SpeechData(text=text, language=self.language),
+            )
+        if mtype == "final":
+            text = (data.get("text") or "").strip()
+            if not text:
+                return None
+            return STTResponse(
+                event_type=SpeechEventType.FINAL,
+                data=SpeechData(
+                    text=text,
+                    language=data.get("language_detected") or self.language,
+                ),
+            )
+        if mtype == "vad_speech":
+            return STTResponse(
+                event_type=SpeechEventType.START,
+                data=SpeechData(text=""),
+            )
+        if mtype == "vad_silence":
+            return STTResponse(
+                event_type=SpeechEventType.END,
+                data=SpeechData(text=""),
+            )
+        if mtype == "error":
+            logger.warning(
+                "VocenceSTT pod error: %s: %s",
+                data.get("code"), data.get("message"),
+            )
+            self.emit("error", str(data.get("message") or data.get("code")))
+            return None
+        # ready / pong / unknown — drop silently
+        return None
+
+    async def _teardown_ws(self) -> None:
+        ws = self._ws
+        if ws is not None and not ws.closed:
+            with _suppress():
+                await ws.send_str(json.dumps({"type": "close"}))
+            with _suppress():
+                await ws.close(code=1000)
+        self._ws = None
+        if self._session is not None and not self._session.closed:
+            with _suppress():
+                await self._session.close()
+        self._session = None
+
+
+class _suppress:
+    """Inline contextlib.suppress(Exception) — keeps the module's
+    explicit import surface small."""
+
+    def __enter__(self) -> None:
+        return None
+
+    def __exit__(self, exc_type, exc, tb) -> bool:
+        return exc_type is not None and issubclass(exc_type, Exception)

vocence_plugins-0.1.0/src/vocence_plugins/tts.py

@@ -0,0 +1,294 @@
+"""VocenceTTS — streaming text-to-speech with the Vocence voice library.
+
+Conforms to the standard TTS abstract interface used by real-time
+agent pipelines, so it slots in alongside any compatible
+``Pipeline(tts=...)``. The plug-in handles all of the network
+plumbing — connection lifecycle, audio framing, and reconnection —
+so callers just see text in and audio out.
+
+Audio output: PCM16LE @ 24 kHz, mono. One persistent connection is
+reused across many ``synthesize()`` calls in the same session;
+lazily opened on the first call and torn down on ``aclose()``.
+"""
+from __future__ import annotations
+
+import asyncio
+import json
+import logging
+import os
+from typing import Any, AsyncIterator, Optional, Union
+from urllib.parse import urlparse
+
+import aiohttp
+
+from videosdk.agents import TTS, FlushMarker  # type: ignore[import-not-found]
+
+logger = logging.getLogger(__name__)
+
+
+_DEFAULT_BASE_URL = os.environ.get("VOCENCE_BASE_URL", "https://api.vocence.ai")
+_DEFAULT_SAMPLE_RATE = 24_000
+_DEFAULT_CHANNELS = 1
+_DEFAULT_TIMEOUT_SEC = 30.0
+# Per-speak text cap. Mirrors the dev-API limit so the server doesn't
+# reject mid-stream. Plugin users sending longer text should split
+# their input — the framework's sentence chunker already handles this when
+# wired in front of the plugin.
+_MAX_TEXT_CHARS_PER_SPEAK = 4000
+
+
+def _ws_url_from_base(base_url: str, voice_id: str) -> str:
+    """Translate ``https://api.vocence.ai`` → ``wss://api.vocence.ai/...``."""
+    parsed = urlparse(base_url.rstrip("/"))
+    scheme = "wss" if parsed.scheme == "https" else "ws"
+    return f"{scheme}://{parsed.netloc}{parsed.path}/v1/voices/{voice_id}/stream"
+
+
+class VocenceTTS(TTS):
+    """Streaming TTS plugin backed by the Vocence voice service.
+
+    Parameters
+    ----------
+    api_key:
+        Vocence developer key (``voc_live_…``). Falls back to the
+        ``VOCENCE_API_KEY`` env var. Required.
+    voice:
+        Voice slug (built-in like ``"design-aria"``) or the numeric id
+        of a saved designed / cloned voice. Required at construction
+        because the WS endpoint is voice-scoped.
+    language:
+        Optional language hint passed on every ``speak`` frame.
+    base_url:
+        Override the default ``https://api.vocence.ai`` (set for
+        staging / self-hosted deployments).
+    """
+
+    def __init__(
+        self,
+        *,
+        api_key: Optional[str] = None,
+        voice: Union[str, int],
+        language: Optional[str] = None,
+        base_url: str = _DEFAULT_BASE_URL,
+        sample_rate: int = _DEFAULT_SAMPLE_RATE,
+        **kwargs: Any,
+    ) -> None:
+        super().__init__(sample_rate=sample_rate, num_channels=_DEFAULT_CHANNELS)
+        self.api_key = api_key or os.environ.get("VOCENCE_API_KEY")
+        if not self.api_key:
+            raise ValueError(
+                "Vocence API key required — pass api_key= or set VOCENCE_API_KEY"
+            )
+        self.voice = str(voice)
+        self.language = language
+        self.base_url = base_url
+        self._ws_url = _ws_url_from_base(base_url, self.voice)
+
+        self._session: aiohttp.ClientSession | None = None
+        self._ws: aiohttp.ClientWebSocketResponse | None = None
+        self._connect_lock = asyncio.Lock()
+        # Per-speak state: cleared at the start of each synthesize() call.
+        # The receiver loop runs INLINE with the speak request so we can
+        # cancel cleanly on interrupt without a separate task.
+        self._interrupted = False
+        self._first_chunk_sent = False
+
+    # ----- abstract overrides ---------------------------------------------
+
+    async def synthesize(
+        self,
+        text: AsyncIterator[Union[str, FlushMarker]] | str,
+        voice_id: Optional[str] = None,
+        **kwargs: Any,
+    ) -> None:
+        """Send text → push PCM frames to ``self.audio_track``.
+
+        ``text`` may be a plain string OR an async iterator of strings
+        (with optional ``FlushMarker`` segment boundaries). The async
+        iterator path lets the pipeline pump LLM tokens directly here
+        instead of waiting for the full reply. We don't yet support
+        the WebSocket re-flush semantics that some upstream sentence
+        chunkers rely on, so we concatenate iterator chunks up to a
+        flush / end-of-stream marker and send them as one ``speak``
+        frame each. This produces good audio with one network
+        round-trip per sentence.
+
+        ``voice_id`` is accepted for API symmetry but ignored — the
+        endpoint is voice-scoped at the WS URL, so changing voice
+        mid-session would require a fresh connection. Construct a new
+        ``VocenceTTS`` for a different voice.
+        """
+        self._interrupted = False
+        self._first_chunk_sent = False
+        await self._ensure_connection()
+        if isinstance(text, str):
+            await self._speak_once(text)
+            return
+        # Async iterator path — collect into segments separated by FlushMarker
+        # (or end-of-iterator), send each segment as one speak.
+        buf: list[str] = []
+        async for chunk in text:
+            if self._interrupted:
+                return
+            if isinstance(chunk, FlushMarker):
+                segment = "".join(buf).strip()
+                buf = []
+                if segment:
+                    await self._speak_once(segment)
+                    if self._interrupted:
+                        return
+                continue
+            if chunk:
+                buf.append(chunk)
+        tail = "".join(buf).strip()
+        if tail and not self._interrupted:
+            await self._speak_once(tail)
+
+    async def interrupt(self) -> None:
+        """Stop the in-flight ``synthesize()`` ASAP. Doesn't close the
+        WebSocket — the connection stays warm for the next call."""
+        self._interrupted = True
+        ws = self._ws
+        if ws is not None and not ws.closed:
+            # Best-effort: cancel any in-flight read. The receiver loop
+            # checks self._interrupted between chunks and bails out.
+            # No control frame to send — the pod will move on once we
+            # send a fresh ``speak`` next time.
+            pass
+
+    async def aclose(self) -> None:
+        """Tear down the WebSocket + HTTP session. Idempotent."""
+        ws = self._ws
+        if ws is not None and not ws.closed:
+            with _suppress():
+                await ws.send_str(json.dumps({"type": "stop"}))
+            with _suppress():
+                await ws.close(code=1000)
+        self._ws = None
+        if self._session is not None and not self._session.closed:
+            with _suppress():
+                await self._session.close()
+        self._session = None
+
+    def reset_first_audio_tracking(self) -> None:
+        self._first_chunk_sent = False
+
+    # ----- internals ------------------------------------------------------
+
+    async def _ensure_connection(self) -> None:
+        """Open the WS if not already open; reopen if it died."""
+        async with self._connect_lock:
+            if self._ws is not None and not self._ws.closed:
+                return
+            if self._session is None or self._session.closed:
+                self._session = aiohttp.ClientSession(
+                    timeout=aiohttp.ClientTimeout(total=_DEFAULT_TIMEOUT_SEC),
+                )
+            headers = {"Authorization": f"Bearer {self.api_key}"}
+            self._ws = await self._session.ws_connect(self._ws_url, headers=headers)
+            # First message from the server is ``ready``. Wait for it so
+            # subsequent ``speak`` frames aren't sent into a half-open
+            # connection. Surface any auth / not-found error here.
+            msg = await self._ws.receive(timeout=_DEFAULT_TIMEOUT_SEC)
+            if msg.type != aiohttp.WSMsgType.TEXT:
+                raise RuntimeError(
+                    f"VocenceTTS: expected text ready frame, got {msg.type}"
+                )
+            data = json.loads(msg.data)
+            mtype = (data.get("type") or "").lower()
+            if mtype == "error":
+                raise RuntimeError(
+                    f"VocenceTTS connect rejected: "
+                    f"{data.get('code')}: {data.get('message')}"
+                )
+            if mtype != "ready":
+                raise RuntimeError(
+                    f"VocenceTTS: unexpected first frame {mtype!r}"
+                )
+
+    async def _speak_once(self, text: str) -> None:
+        """Send one ``speak`` and drain audio frames until ``end``."""
+        if not text:
+            return
+        if len(text) > _MAX_TEXT_CHARS_PER_SPEAK:
+            # Truncate rather than fail: a sentence chunker in front
+            # of us should keep segments well under the cap.
+            logger.warning(
+                "VocenceTTS: truncating %d-char segment to %d (cap)",
+                len(text), _MAX_TEXT_CHARS_PER_SPEAK,
+            )
+            text = text[:_MAX_TEXT_CHARS_PER_SPEAK]
+        ws = self._ws
+        if ws is None:
+            return
+        payload: dict[str, Any] = {"type": "speak", "text": text}
+        if self.language:
+            payload["language"] = self.language
+        await ws.send_str(json.dumps(payload))
+        await self._drain_until_end(ws)
+
+    async def _drain_until_end(self, ws: aiohttp.ClientWebSocketResponse) -> None:
+        """Read frames until ``{"type":"end"}`` or interruption."""
+        while True:
+            if self._interrupted:
+                return
+            try:
+                msg = await ws.receive(timeout=_DEFAULT_TIMEOUT_SEC)
+            except asyncio.TimeoutError:
+                logger.warning("VocenceTTS: receive timed out")
+                return
+            if msg.type == aiohttp.WSMsgType.BINARY:
+                if not msg.data:
+                    continue
+                # First-byte callback for TTFB metrics — the pipeline
+                # uses this to fire its ``first_audio_byte`` event.
+                if not self._first_chunk_sent:
+                    self._first_chunk_sent = True
+                    if self._first_audio_callback is not None:
+                        with _suppress():
+                            await self._first_audio_callback()
+                if self.audio_track is not None:
+                    with _suppress():
+                        await self.audio_track.add_new_bytes(msg.data)
+                continue
+            if msg.type == aiohttp.WSMsgType.TEXT:
+                try:
+                    data = json.loads(msg.data)
+                except json.JSONDecodeError:
+                    continue
+                mtype = (data.get("type") or "").lower()
+                if mtype == "end":
+                    return
+                if mtype == "meta":
+                    # Sample rate / encoding info — we know our pod
+                    # output is PCM16LE @ 24 kHz mono so don't need to
+                    # do anything with this on the plugin side. The
+                    # the audio_track already expects this format
+                    # because we set ``sample_rate=24000`` on the base
+                    # class at __init__.
+                    continue
+                if mtype == "error":
+                    logger.warning(
+                        "VocenceTTS pod error: %s: %s",
+                        data.get("code"), data.get("message"),
+                    )
+                    return
+                continue
+            if msg.type in (
+                aiohttp.WSMsgType.CLOSED,
+                aiohttp.WSMsgType.CLOSE,
+                aiohttp.WSMsgType.ERROR,
+            ):
+                self._ws = None
+                return
+
+
+class _suppress:
+    """Tiny ``contextlib.suppress(Exception)`` clone — kept inline to
+    keep the module's import surface small."""
+
+    def __enter__(self) -> None:
+        return None
+
+    def __exit__(self, exc_type, exc, tb) -> bool:
+        return exc_type is not None and issubclass(exc_type, Exception)