feat(stepfun): add speech recognition, synthesis, and chat plugin

Add WebSocket and batch SSE speech recognition, streaming and HTTP SSE speech synthesis, and an optional OpenAI-compatible chat wrapper.

Support regional endpoints, provider error handling, and conversational stream lifecycle management with regression coverage.

Author: robinzsx

livekit-plugins/livekit-plugins-stepfun/README.md

@@ -0,0 +1,55 @@
+# StepFun plugin for LiveKit Agents
+
+Support for [StepFun](https://platform.stepfun.ai/docs) speech recognition,
+speech synthesis, and OpenAI-compatible chat models.
+
+## Installation
+
+```bash
+pip install livekit-plugins-stepfun
+# Include the OpenAI-compatible chat client:
+pip install 'livekit-plugins-stepfun[llm]'
+```
+
+## Configuration
+
+Set `STEPFUN_API_KEY`, or pass `api_key=` to the constructors. StepFun runs two
+regional clusters with separate endpoints and separate credentials:
+
+| Cluster | Endpoint | Value |
+| --- | --- | --- |
+| Global (default) | `api.stepfun.ai` | `"global"` |
+| Mainland China | `api.stepfun.com` | `"china"` |
+
+Select the cluster with `STEPFUN_CLUSTER` or `cluster=`. The two clusters use
+separate endpoints and separate credentials, so a key from one will not work on
+the other.
+
+- `stepfun.STT()` streams recognition with server endpointing.
+  `recognize()` uses the batch route; `use_streaming=False` selects batch
+  recognition for sessions and requires local VAD through `stt.StreamAdapter`.
+  Streaming input is intended for paced conversational audio, not file uploads;
+  use `recognize()` for recorded audio. At end of input, server-VAD streams send
+  endpointing silence and then a best-effort `input_audio_buffer.commit` (the
+  gateway currently ignores it with server VAD, but it will finalize the last
+  segment once commits are forwarded), and keep receiving: while a turn is pending this waits up
+  to `conn_options.timeout` and fails on expiry, otherwise it ends after a short
+  quiet period. The API has no acknowledgement that trailing silence was
+  processed, so completion is best-effort: it succeeds only after a final was
+  delivered, and otherwise reports that completion cannot be confirmed once a
+  bounded interval elapses. A turn first reported later than the quiet period is
+  not recovered. `aclose()` cancels without waiting.
+  Failures after consuming audio are non-retryable because a fresh session
+  cannot recover that input. `aclose()` cancels the stream without draining it.
+- `stepfun.TTS()` supports streaming and one-shot synthesis. Discover voice IDs
+  using `await tts.list_voices()`. One-shot synthesis streams PCM over SSE, so
+  audio starts arriving before the whole request is synthesized; inputs longer
+  than 1,000 characters are split into sequential requests and emitted as one
+  audio segment. Each request has an idle-read timeout plus a total deadline
+  that scales with its text length.
+- `stepfun.LLM()` uses the optional OpenAI plugin for chat and tool calling.
+
+Pass `language="zh"` for Chinese; the speech services default to English.
+Streaming recognition defaults to `stepaudio-2.5-asr-stream`, batch recognition
+to `stepaudio-3-asr-max`, synthesis to `stepaudio-3-tts`, and chat to
+`step-3.7-flash`. Custom model IDs are accepted.

livekit-plugins/livekit-plugins-stepfun/livekit/plugins/stepfun/__init__.py

@@ -0,0 +1,81 @@
+"""LiveKit Agents plugin for StepFun (阶跃星辰) ASR, TTS and chat models."""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, Any
+
+from livekit.agents import Plugin
+
+from .log import logger
+from .models import (
+    Cluster,
+    Language,
+    LLMModels,
+    ReasoningEffort,
+    STTBatchModels,
+    STTModels,
+    TTSModels,
+    TTSVoices,
+)
+from .stt import STT, RecognizeStream, SpeechStream
+from .tts import TTS, ChunkedStream, SynthesizeStream
+from .version import __version__
+
+# Register the optional OpenAI plugin on the importing thread, before jobs start.
+try:
+    from .llm import LLM
+
+    _LLM_IMPORT_ERROR: ImportError | None = None
+except ImportError as exc:
+    _LLM_IMPORT_ERROR = exc
+
+if TYPE_CHECKING:
+    from .llm import LLM
+
+__all__ = [
+    "LLM",
+    "STT",
+    "TTS",
+    "ChunkedStream",
+    "Cluster",
+    "Language",
+    "LLMModels",
+    "ReasoningEffort",
+    "RecognizeStream",
+    "STTBatchModels",
+    "STTModels",
+    "SpeechStream",
+    "SynthesizeStream",
+    "TTSModels",
+    "TTSVoices",
+    "__version__",
+]
+
+
+def __getattr__(name: str) -> Any:
+    """Explain a missing `llm` extra at access time.
+
+    Only reached when the eager import above failed, since a successful one
+    binds LLM as a module global and module __getattr__ is never consulted for
+    names that already exist.
+    """
+    if name == "LLM":
+        assert _LLM_IMPORT_ERROR is not None
+        raise _LLM_IMPORT_ERROR
+    raise AttributeError(f"module {__name__!r} has no attribute {name!r}")
+
+
+class StepFunPlugin(Plugin):
+    def __init__(self) -> None:
+        super().__init__(__name__, __version__, __package__, logger)
+
+
+Plugin.register_plugin(StepFunPlugin())
+
+# Hide symbols the doc generator would otherwise pick up
+_module = dir()
+NOT_IN_ALL = [m for m in _module if m not in __all__]
+
+__pdoc__ = {}
+for _n in NOT_IN_ALL:
+    __pdoc__[_n] = False