agent-cli 0.70.4__tar.gz → 0.70.5__tar.gz

{agent_cli-0.70.4 → agent_cli-0.70.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agent-cli
-Version: 0.70.4
+Version: 0.70.5
 Summary: A suite of AI-powered command-line tools for text correction, audio transcription, and voice assistance.
 Project-URL: Homepage, https://github.com/basnijholt/agent-cli
 Author-email: Bas Nijholt <bas@nijho.lt>
@@ -80,6 +80,8 @@ Requires-Dist: pytest-timeout; extra == 'test'
 Requires-Dist: pytest>=7.0.0; extra == 'test'
 Provides-Extra: vad
 Requires-Dist: silero-vad>=5.1; extra == 'vad'
+Provides-Extra: wyoming
+Requires-Dist: wyoming>=1.5.2; extra == 'wyoming'
 Description-Content-Type: text/markdown
 
 # Agent CLI

{agent_cli-0.70.4 → agent_cli-0.70.5}/agent_cli/_extras.json

@@ -1,5 +1,6 @@
 {
-  "audio": ["Audio recording/playback with Wyoming protocol", ["numpy", "sounddevice", "wyoming"]],
+  "wyoming": ["Wyoming protocol for ASR/TTS servers", ["wyoming"]],
+  "audio": ["Local audio recording/playback", ["numpy", "sounddevice", "wyoming"]],
   "llm": ["LLM framework (pydantic-ai)", ["pydantic_ai"]],
   "memory": ["Long-term memory proxy", ["chromadb", "yaml"]],
   "rag": ["RAG proxy (ChromaDB, embeddings)", ["chromadb"]],

agent_cli-0.70.5/agent_cli/_requirements/wyoming.txt

@@ -0,0 +1,71 @@
+# This file was autogenerated by uv via the following command:
+#    uv export --extra wyoming --no-dev --no-emit-project --no-hashes
+annotated-types==0.7.0
+    # via pydantic
+anyio==4.12.1
+    # via httpx
+certifi==2026.1.4
+    # via
+    #   httpcore
+    #   httpx
+click==8.3.1
+    # via
+    #   typer
+    #   typer-slim
+colorama==0.4.6 ; sys_platform == 'win32'
+    # via click
+dotenv==0.9.9
+    # via agent-cli
+h11==0.16.0
+    # via httpcore
+httpcore==1.0.9
+    # via httpx
+httpx==0.28.1
+    # via agent-cli
+idna==3.11
+    # via
+    #   anyio
+    #   httpx
+markdown-it-py==4.0.0
+    # via rich
+mdurl==0.1.2
+    # via markdown-it-py
+psutil==7.2.1 ; sys_platform == 'win32'
+    # via agent-cli
+pydantic==2.12.5
+    # via agent-cli
+pydantic-core==2.41.5
+    # via pydantic
+pygments==2.19.2
+    # via rich
+pyperclip==1.11.0
+    # via agent-cli
+python-dotenv==1.2.1
+    # via dotenv
+rich==14.2.0
+    # via
+    #   agent-cli
+    #   typer
+    #   typer-slim
+setproctitle==1.3.7
+    # via agent-cli
+shellingham==1.5.4
+    # via
+    #   typer
+    #   typer-slim
+typer==0.21.1
+    # via agent-cli
+typer-slim==0.21.1
+    # via agent-cli
+typing-extensions==4.15.0
+    # via
+    #   anyio
+    #   pydantic
+    #   pydantic-core
+    #   typer
+    #   typer-slim
+    #   typing-inspection
+typing-inspection==0.4.2
+    # via pydantic
+wyoming==1.8.0
+    # via agent-cli

{agent_cli-0.70.4 → agent_cli-0.70.5}/agent_cli/server/cli.py

@@ -417,7 +417,7 @@ def whisper_cmd(  # noqa: PLR0912, PLR0915
 
 
 @app.command("transcribe-proxy")
-@requires_extras("server", "audio", "llm")
+@requires_extras("server", "wyoming", "llm")
 def transcribe_proxy_cmd(
     host: Annotated[
         str,

{agent_cli-0.70.4 → agent_cli-0.70.5}/agent_cli/server/proxy/api.py

@@ -121,6 +121,7 @@ async def health_check() -> HealthResponse:
 
 async def _transcribe_with_provider(
     audio_data: bytes,
+    filename: str,
     provider_cfg: config.ProviderSelection,
     wyoming_asr_cfg: config.WyomingASR,
     openai_asr_cfg: config.OpenAIASR,
@@ -128,6 +129,7 @@ async def _transcribe_with_provider(
 ) -> str:
     """Transcribe audio using the configured provider."""
     transcriber = asr.create_recorded_audio_transcriber(provider_cfg)
+    file_suffix = Path(filename).suffix.lower() or ".wav"
 
     if provider_cfg.asr_provider == "wyoming":
         return await transcriber(
@@ -140,12 +142,14 @@ async def _transcribe_with_provider(
             audio_data=audio_data,
             openai_asr_cfg=openai_asr_cfg,
             logger=LOGGER,
+            file_suffix=file_suffix,
         )
     if provider_cfg.asr_provider == "gemini":
         return await transcriber(
             audio_data=audio_data,
             gemini_asr_cfg=gemini_asr_cfg,
             logger=LOGGER,
+            file_suffix=file_suffix,
         )
     msg = f"Unsupported ASR provider: {provider_cfg.asr_provider}"
     raise NotImplementedError(msg)
@@ -356,8 +360,14 @@ async def transcribe_audio(
             defaults,
         ) = _load_transcription_configs()
 
-        # Save uploaded file
+        # Read uploaded file
         audio_data = await audio_file.read()
+        LOGGER.info(
+            "Received audio: filename=%s, size=%d bytes, content_type=%s",
+            audio_file.filename,
+            len(audio_data),
+            audio_file.content_type,
+        )
 
         # Convert audio to Wyoming format if using local ASR
         if provider_cfg.asr_provider == "wyoming":
@@ -366,6 +376,7 @@ async def transcribe_audio(
         # Transcribe audio using the configured provider
         raw_transcript = await _transcribe_with_provider(
             audio_data,
+            audio_file.filename or "audio.wav",
             provider_cfg,
             wyoming_asr_cfg,
             openai_asr_cfg,

{agent_cli-0.70.4 → agent_cli-0.70.5}/agent_cli/services/__init__.py

@@ -17,6 +17,7 @@ if TYPE_CHECKING:
 
 
 _RIFF_HEADER = b"RIFF"
+_LOG_TRUNCATE_LENGTH = 100
 
 
 def _is_wav_file(data: bytes) -> bool:
@@ -112,9 +113,17 @@ async def transcribe_audio_gemini(
     # Determine MIME type from file suffix
     mime_type = _GEMINI_MIME_TYPES.get(file_suffix.lower(), "audio/wav")
 
+    logger.debug(
+        "Received audio: size=%d bytes, file_suffix=%s, is_wav=%s",
+        len(audio_data),
+        file_suffix,
+        _is_wav_file(audio_data),
+    )
+
     # If raw PCM (no recognized format header), convert to WAV
+    # Only do this if file_suffix is .wav but data doesn't have WAV header (indicating raw PCM)
     if not _is_wav_file(audio_data) and file_suffix.lower() == ".wav":
-        logger.debug("Converting raw PCM to WAV format for Gemini")
+        logger.debug("Wrapping raw PCM data with WAV header (16kHz, 16-bit, mono)")
         audio_data = pcm_to_wav(
             audio_data,
             sample_rate=constants.AUDIO_RATE,
@@ -141,7 +150,19 @@ async def transcribe_audio_gemini(
             types.Part.from_bytes(data=audio_data, mime_type=mime_type),
         ],
     )
-    return response.text.strip()
+    text = response.text.strip()
+
+    if text:
+        logger.info(
+            "Transcription result: %s",
+            text[:_LOG_TRUNCATE_LENGTH] + "..." if len(text) > _LOG_TRUNCATE_LENGTH else text,
+        )
+    else:
+        logger.warning(
+            "Empty transcription returned - audio may be silent, corrupted, or in wrong format",
+        )
+
+    return text
 
 
 def _get_openai_client(api_key: str | None, base_url: str | None = None) -> AsyncOpenAI:
@@ -197,9 +218,17 @@ async def transcribe_audio_openai(
         base_url=openai_asr_cfg.openai_base_url,
     )
 
+    logger.debug(
+        "Received audio: size=%d bytes, file_suffix=%s, is_wav=%s",
+        len(audio_data),
+        file_suffix,
+        _is_wav_file(audio_data),
+    )
+
     # Convert raw PCM to WAV if needed (custom endpoints like faster-whisper require proper format)
+    # Only do this if file_suffix is .wav but data doesn't have WAV header (indicating raw PCM)
     if not _is_wav_file(audio_data) and file_suffix.lower() == ".wav":
-        logger.debug("Converting raw PCM to WAV format")
+        logger.debug("Wrapping raw PCM data with WAV header (16kHz, 16-bit, mono)")
         audio_data = pcm_to_wav(
             audio_data,
             sample_rate=constants.AUDIO_RATE,
@@ -211,7 +240,7 @@ async def transcribe_audio_openai(
     # Use the correct file extension so OpenAI knows the format
     audio_file.name = f"audio{file_suffix}"
 
-    logger.debug("Using filename: %s", audio_file.name)
+    logger.debug("Sending to OpenAI with filename: %s", audio_file.name)
 
     transcription_params: dict[str, object] = {
         "model": openai_asr_cfg.asr_openai_model,
@@ -225,7 +254,19 @@ async def transcribe_audio_openai(
         logger.debug("Using OpenAI ASR with prompt")
 
     response = await client.audio.transcriptions.create(**transcription_params)
-    return response.text
+    text = response.text
+
+    if text:
+        logger.info(
+            "Transcription result: %s",
+            text[:_LOG_TRUNCATE_LENGTH] + "..." if len(text) > _LOG_TRUNCATE_LENGTH else text,
+        )
+    else:
+        logger.warning(
+            "Empty transcription returned - audio may be silent, corrupted, or in wrong format",
+        )
+
+    return text
 
 
 async def synthesize_speech_openai(

{agent_cli-0.70.4 → agent_cli-0.70.5}/docker/transcribe-proxy.Dockerfile

@@ -38,7 +38,7 @@ COPY pyproject.toml uv.lock README.md ./
 COPY .git ./.git
 COPY agent_cli ./agent_cli
 COPY scripts ./scripts
-RUN uv sync --frozen --no-dev --no-editable --extra server --extra audio --extra llm
+RUN uv sync --frozen --no-dev --no-editable --extra server --extra wyoming --extra llm
 
 # =============================================================================
 # Runtime stage - minimal image

{agent_cli-0.70.4 → agent_cli-0.70.5}/docker/tts.Dockerfile

@@ -26,7 +26,7 @@ COPY pyproject.toml uv.lock README.md ./
 COPY .git ./.git
 COPY agent_cli ./agent_cli
 COPY scripts ./scripts
-RUN uv sync --frozen --no-dev --no-editable --extra server --extra kokoro --extra audio && \
+RUN uv sync --frozen --no-dev --no-editable --extra server --extra kokoro --extra wyoming && \
     /app/.venv/bin/python -m spacy download en_core_web_sm
 
 # =============================================================================
@@ -46,7 +46,7 @@ COPY pyproject.toml uv.lock README.md ./
 COPY .git ./.git
 COPY agent_cli ./agent_cli
 COPY scripts ./scripts
-RUN uv sync --frozen --no-dev --no-editable --extra server --extra piper --extra audio
+RUN uv sync --frozen --no-dev --no-editable --extra server --extra piper --extra wyoming
 
 # =============================================================================
 # CUDA target: GPU-accelerated with Kokoro TTS

{agent_cli-0.70.4 → agent_cli-0.70.5}/docker/whisper.Dockerfile

@@ -26,7 +26,7 @@ COPY pyproject.toml uv.lock README.md ./
 COPY .git ./.git
 COPY agent_cli ./agent_cli
 COPY scripts ./scripts
-RUN uv sync --frozen --no-dev --no-editable --extra server --extra faster-whisper --extra audio
+RUN uv sync --frozen --no-dev --no-editable --extra server --extra faster-whisper --extra wyoming
 
 # =============================================================================
 # CUDA target: GPU-accelerated with faster-whisper

{agent_cli-0.70.4 → agent_cli-0.70.5}/docs/commands/install-extras.md

@@ -31,7 +31,8 @@ Available extras:
 <!-- ⚠️ This content is auto-generated by `markdown-code-runner`. -->
 | Extra | Description |
 |-------|-------------|
-| `audio` | Audio recording/playback with Wyoming protocol |
+| `wyoming` | Wyoming protocol for ASR/TTS servers |
+| `audio` | Local audio recording/playback |
 | `llm` | LLM framework (pydantic-ai) |
 | `memory` | Long-term memory proxy |
 | `rag` | RAG proxy (ChromaDB, embeddings) |

{agent_cli-0.70.4 → agent_cli-0.70.5}/pyproject.toml

@@ -31,7 +31,8 @@ Homepage = "https://github.com/basnijholt/agent-cli"
 
 [project.optional-dependencies]
 # Provider extras (previously base dependencies, now optional)
-audio = ["numpy", "sounddevice>=0.4.6", "wyoming>=1.5.2"]
+wyoming = ["wyoming>=1.5.2"]  # Wyoming protocol only (for servers)
+audio = ["numpy", "sounddevice>=0.4.6", "agent-cli[wyoming]"]  # Local audio recording
 llm = ["pydantic-ai-slim[openai,google,duckduckgo,vertexai]>=0.1.1"]
 
 # Feature extras

{agent_cli-0.70.4 → agent_cli-0.70.5}/tests/test_api_integration.py

@@ -52,6 +52,7 @@ async def test_full_transcription_workflow() -> None:
             # Create mock upload file
             class MockUploadFile:
                 filename = "test.wav"
+                content_type = "audio/wav"
 
                 async def read(self) -> bytes:
                     return audio_data
@@ -215,6 +216,7 @@ async def test_concurrent_requests() -> None:
         class MockUploadFile:
             def __init__(self, idx: int) -> None:
                 self.filename = f"test{idx}.wav"
+                self.content_type = "audio/wav"
                 self.idx = idx
 
             async def read(self) -> bytes:

{agent_cli-0.70.4 → agent_cli-0.70.5}/uv.lock

@@ -113,6 +113,9 @@ test = [
 vad = [
     { name = "silero-vad" },
 ]
+wyoming = [
+    { name = "wyoming" },
+]
 
 [package.dev-dependencies]
 dev = [
@@ -194,9 +197,10 @@ requires-dist = [
     { name = "watchfiles", marker = "extra == 'memory'", specifier = ">=0.21.0" },
     { name = "watchfiles", marker = "extra == 'rag'", specifier = ">=0.21.0" },
     { name = "wyoming", marker = "extra == 'audio'", specifier = ">=1.5.2" },
+    { name = "wyoming", marker = "extra == 'wyoming'", specifier = ">=1.5.2" },
     { name = "zensical", marker = "extra == 'dev'" },
 ]
-provides-extras = ["audio", "dev", "faster-whisper", "kokoro", "llm", "memory", "mlx-whisper", "piper", "rag", "server", "speed", "test", "vad"]
+provides-extras = ["audio", "dev", "faster-whisper", "kokoro", "llm", "memory", "mlx-whisper", "piper", "rag", "server", "speed", "test", "vad", "wyoming"]
 
 [package.metadata.requires-dev]
 dev = [