converse-framework 0.2.2__tar.gz → 0.2.3__tar.gz

{converse_framework-0.2.2 → converse_framework-0.2.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: converse-framework
-Version: 0.2.2
+Version: 0.2.3
 Summary: Provider-agnostic speech stack for speech-to-speech applications
 License: MIT
 License-File: LICENSE
@@ -22,12 +22,15 @@ Requires-Dist: nvidia-cublas-cu12; (platform_system == 'Windows') and extra == '
 Provides-Extra: all-llm
 Requires-Dist: httpx>=0.28; extra == 'all-llm'
 Provides-Extra: all-tts
+Requires-Dist: httpx>=0.28; extra == 'all-tts'
 Requires-Dist: kokoro-onnx>=0.5; (python_version < '3.14') and extra == 'all-tts'
 Requires-Dist: misaki>=0.7; extra == 'all-tts'
 Requires-Dist: pocket-tts>=2.1; extra == 'all-tts'
 Provides-Extra: all-vad
 Requires-Dist: onnxruntime>=1.20; extra == 'all-vad'
 Requires-Dist: silero-vad>=6.0; extra == 'all-vad'
+Provides-Extra: audio-cpp
+Requires-Dist: httpx>=0.28; extra == 'audio-cpp'
 Provides-Extra: faster-whisper
 Requires-Dist: faster-whisper>=1.2; extra == 'faster-whisper'
 Requires-Dist: nvidia-cublas-cu12; (platform_system == 'Windows') and extra == 'faster-whisper'
@@ -94,6 +97,7 @@ providers live behind optional extras:
 pip install converse-framework[silero]          # Silero VAD
 pip install converse-framework[faster-whisper]  # faster-whisper ASR
 pip install converse-framework[whisper-cpp]     # whisper.cpp HTTP ASR
+pip install converse-framework[audio-cpp]       # audio.cpp HTTP ASR + TTS
 pip install converse-framework[llamacpp]        # llama.cpp HTTP LLM
 pip install converse-framework[kokoro]          # Kokoro ONNX TTS
 pip install converse-framework[pocket-tts]      # Pocket TTS
@@ -147,12 +151,13 @@ own constraints (the table below mirrors the markers in
 | `faster-whisper` | 3.11+ | The `nvidia-cublas-cu12` wheel pins Windows. |
 | `llamacpp` | 3.11+ | `httpx` itself supports 3.9+, so 3.11+ is the only constraint. |
 | `whisper-cpp` | 3.11+ | Only needs `httpx`, which supports 3.9+. |
+| `audio-cpp` | 3.11+ | Only needs `httpx`. Talks to a user-managed `audiocpp_server`. |
 | `kokoro` | 3.11 to <3.14 | `kokoro-onnx` 0.5.0 requires Python <3.14. The wheel build fails fast on 3.14+. |
 | `pocket-tts` | 3.11+ | No known upper bound. |
 
 The `kokoro` extra is the only one with an upper-bound marker today.
-If you are on Python 3.14+ and need a TTS provider, use `pocket-tts`
-or a mock provider. New providers should add their own
+If you are on Python 3.14+ and need a TTS provider, use `pocket-tts`,
+`audio-cpp`, or a mock provider. New providers should add their own
 `python_version` markers in `pyproject.toml` when their backend has a
 known limit.
 

{converse_framework-0.2.2 → converse_framework-0.2.3}/README.md

@@ -47,6 +47,7 @@ providers live behind optional extras:
 pip install converse-framework[silero]          # Silero VAD
 pip install converse-framework[faster-whisper]  # faster-whisper ASR
 pip install converse-framework[whisper-cpp]     # whisper.cpp HTTP ASR
+pip install converse-framework[audio-cpp]       # audio.cpp HTTP ASR + TTS
 pip install converse-framework[llamacpp]        # llama.cpp HTTP LLM
 pip install converse-framework[kokoro]          # Kokoro ONNX TTS
 pip install converse-framework[pocket-tts]      # Pocket TTS
@@ -100,12 +101,13 @@ own constraints (the table below mirrors the markers in
 | `faster-whisper` | 3.11+ | The `nvidia-cublas-cu12` wheel pins Windows. |
 | `llamacpp` | 3.11+ | `httpx` itself supports 3.9+, so 3.11+ is the only constraint. |
 | `whisper-cpp` | 3.11+ | Only needs `httpx`, which supports 3.9+. |
+| `audio-cpp` | 3.11+ | Only needs `httpx`. Talks to a user-managed `audiocpp_server`. |
 | `kokoro` | 3.11 to <3.14 | `kokoro-onnx` 0.5.0 requires Python <3.14. The wheel build fails fast on 3.14+. |
 | `pocket-tts` | 3.11+ | No known upper bound. |
 
 The `kokoro` extra is the only one with an upper-bound marker today.
-If you are on Python 3.14+ and need a TTS provider, use `pocket-tts`
-or a mock provider. New providers should add their own
+If you are on Python 3.14+ and need a TTS provider, use `pocket-tts`,
+`audio-cpp`, or a mock provider. New providers should add their own
 `python_version` markers in `pyproject.toml` when their backend has a
 known limit.
 

{converse_framework-0.2.2 → converse_framework-0.2.3}/converse_framework/audio_utils.py

@@ -175,6 +175,53 @@ def float_audio_to_wav_bytes(audio, sample_rate: int) -> bytes:
     return buffer.getvalue()
 
 
+def wav_bytes_to_pcm_s16le(
+    wav_bytes: bytes,
+) -> tuple[bytes, int, int]:
+    """Decode a WAV byte string back into raw PCM s16le bytes and shape.
+
+    The inverse of :func:`float_audio_to_wav_bytes`: this reads a
+    complete 16-bit PCM WAV file from ``bytes`` and returns the raw
+    signed little-endian PCM body along with the sample rate and channel
+    count declared in the header. Providers that fetch WAV audio from
+    an HTTP backend (e.g. the audio.cpp ``/v1/audio/speech`` endpoint)
+    use this to turn the response into a wire-ready
+    :class:`~converse_framework.protocols.AudioChunk`.
+
+    Args:
+        wav_bytes: A complete WAV file as ``bytes`` (RIFF header + data).
+
+    Returns:
+        A ``(pcm_s16le, sample_rate, channels)`` tuple. ``pcm_s16le`` is
+        the raw 16-bit signed LE PCM bytes with no header;
+        ``sample_rate`` and ``channels`` come from the ``fmt `` chunk.
+        An empty or non-WAV input returns ``(b"", 0, 0)`` rather than
+        raising so callers can treat a failed decode as "no audio".
+
+    Raises:
+        ValueError: If ``wav_bytes`` starts like a WAV stream but cannot
+            be parsed, or if it is not 16-bit PCM audio.
+    """
+    if not wav_bytes or wav_bytes[:4] != b"RIFF":
+        return b"", 0, 0
+    buffer = BytesIO(wav_bytes)
+    try:
+        with wave.open(buffer, "rb") as wav:
+            sample_rate = wav.getframerate()
+            channels = wav.getnchannels()
+            sample_width = wav.getsampwidth()
+            if wav.getcomptype() != "NONE" or sample_width != 2:
+                bits = sample_width * 8
+                raise ValueError(
+                    f"unsupported WAV format: expected 16-bit PCM, got {bits}-bit "
+                    f"{wav.getcompname()}"
+                )
+            pcm = wav.readframes(wav.getnframes())
+    except wave.Error as exc:
+        raise ValueError(f"invalid WAV stream: {exc}") from exc
+    return pcm, sample_rate, channels
+
+
 def float_audio_to_pcm_s16le_bytes(audio) -> bytes:
     """Encode a float audio buffer as raw 16-bit signed LE PCM bytes.
 

{converse_framework-0.2.2 → converse_framework-0.2.3}/converse_framework/providers/__init__.py

@@ -2,9 +2,10 @@
 
 Mock and unavailable providers are imported eagerly because they have no
 heavy dependencies. The concrete providers (``silero``, ``faster-whisper``,
-``llamacpp``, ``kokoro-onnx``, ``pocket-tts``) are not imported here --
-they are registered with :func:`converse_framework.registry.register_provider`
-by import string and loaded lazily on first use.
+``whisper-cpp``, ``llamacpp``, ``kokoro-onnx``, ``pocket-tts``,
+``audio-cpp``) are not imported here -- they are registered with
+:func:`converse_framework.registry.register_provider` by import string
+and loaded lazily on first use.
 """
 
 from converse_framework.providers.mock import (