Rename and publish Agent Voice

.github/scripts/verify_windows.py → .github/scripts/verify_package.py

@@ -2,13 +2,17 @@
 
 import json
 import os
+import platform
+import re
 import subprocess
 import sys
 import time
 import urllib.request
 import wave
 from pathlib import Path
 
+import imageio_ffmpeg
+
 SERVICE_URL = "http://127.0.0.1:18765"
 
 
@@ -22,51 +26,57 @@ def run_cli(cli: Path, *args: str) -> dict[str, object]:
     return json.loads(completed.stdout.strip().splitlines()[-1])
 
 
-def validate_wav(path: Path) -> None:
-    with wave.open(str(path), "rb") as audio:
-        assert audio.getnchannels() == 1
-        assert audio.getframerate() == 24_000
-        assert audio.getnframes() > 0
+def validate_decodable(path: Path) -> None:
+    ffmpeg = imageio_ffmpeg.get_ffmpeg_exe()
+    assert "imageio_ffmpeg" in Path(ffmpeg).as_posix()
     completed = subprocess.run(
         [
-            "ffprobe",
+            ffmpeg,
             "-v",
             "error",
-            "-select_streams",
-            "a:0",
-            "-show_entries",
-            "stream=codec_name,sample_rate,channels",
-            "-of",
-            "json",
+            "-i",
             str(path),
+            "-f",
+            "s16le",
+            "-acodec",
+            "pcm_s16le",
+            "pipe:1",
         ],
         check=True,
         capture_output=True,
-        text=True,
     )
-    stream = json.loads(completed.stdout)["streams"][0]
-    assert stream == {
-        "codec_name": "pcm_s16le",
-        "sample_rate": "24000",
-        "channels": 1,
-    }
+    assert completed.stdout
+
+
+def validate_wav(path: Path) -> None:
+    with wave.open(str(path), "rb") as audio:
+        assert audio.getnchannels() == 1
+        assert audio.getframerate() == 24_000
+        assert audio.getnframes() > 0
+    validate_decodable(path)
+
+
+def validate_mp3(path: Path) -> None:
+    assert path.suffix == ".mp3"
+    validate_decodable(path)
 
 
 def check_doctor(report: dict[str, object], service_status: str) -> None:
     assert report["ok"] is True
     checks = {check["name"]: check for check in report["checks"]}
     assert checks["model"]["status"] == "pass"
     assert checks["runtime"]["status"] == "pass"
-    assert checks["playback"]["status"] == "warn"
-    assert "experimental" in checks["playback"]["detail"]
+    assert checks["compressed audio"]["status"] == "pass"
+    assert "bundled by imageio-ffmpeg" in checks["compressed audio"]["detail"]
+    assert checks["playback"]["status"] in {"pass", "warn"}
+    assert "miniaudio" in checks["playback"]["detail"]
     assert checks["service"]["status"] == service_status
 
 
 def main() -> None:
-    if sys.platform != "win32":
-        raise RuntimeError("This verification is intentionally Windows-only")
     cli = Path(sys.argv[1]).resolve()
-    output_dir = Path(os.environ["RUNNER_TEMP"]) / "kokoro-windows-e2e"
+    system = platform.system()
+    output_dir = Path(os.environ["RUNNER_TEMP"]) / "agent-voice-package-e2e"
     output_dir.mkdir(parents=True, exist_ok=True)
 
     subprocess.run([str(cli), "setup", "--model", "int8"], check=True)
@@ -77,17 +87,42 @@ def main() -> None:
     local = run_cli(
         cli,
         "speak",
-        "Windows generation verification.",
+        f"{system} generation verification.",
         "--service",
         "off",
         "--output",
         str(local_wav),
-        "--json",
     )
     assert local["backend"] == "local"
     assert local["played"] is False
     validate_wav(local_wav)
 
+    labeled = run_cli(
+        cli,
+        "speak",
+        f"{system} labeled speed verification.",
+        "--service",
+        "off",
+        "--label",
+        "Package E2E",
+        "--format",
+        "mp3",
+        "--speed",
+        "1.5",
+    )
+    labeled_path = Path(str(labeled["path"]))
+    assert labeled["backend"] == "local"
+    assert labeled["speed"] == 1.5
+    assert re.fullmatch(
+        r"Package-E2E-\d{2}-\d{2}-\d{2}-at-\d{2}-\d{2}\.mp3",
+        labeled_path.name,
+    )
+    assert (
+        labeled_path.parent
+        == (Path(os.environ["AGENT_VOICE_HOME"]) / "recordings").resolve()
+    )
+    validate_mp3(labeled_path)
+
     log_path = output_dir / "service.log"
     with log_path.open("w", encoding="utf-8") as log:
         service = subprocess.Popen(
@@ -120,14 +155,13 @@ def main() -> None:
             remote = run_cli(
                 cli,
                 "speak",
-                "Windows localhost service verification.",
+                f"{system} localhost service verification.",
                 "--service",
-                "required",
+                "on",
                 "--service-url",
                 SERVICE_URL,
                 "--output",
                 str(service_wav),
-                "--json",
             )
             assert remote["backend"] == "service"
             assert remote["played"] is False
@@ -140,6 +174,7 @@ def main() -> None:
                 service.kill()
                 service.wait(timeout=10)
     print(log_path.read_text(encoding="utf-8", errors="replace"))
+    print(f"Verified installed package on {system}")
 
 
 if __name__ == "__main__":

.github/workflows/ci.yml

@@ -2,11 +2,16 @@ name: CI
 
 on:
   push:
+    branches: [main]
   pull_request:
 
 permissions:
   contents: read
 
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
+  cancel-in-progress: true
+
 jobs:
   test:
     name: ${{ matrix.os }} / Python ${{ matrix.python-version }}
@@ -26,28 +31,39 @@ jobs:
           enable-cache: true
           python-version: ${{ matrix.python-version }}
 
-      - name: Install FFmpeg on Linux
-        if: runner.os == 'Linux'
-        run: |
-          sudo apt-get update
-          sudo apt-get install --yes ffmpeg
-
-      - name: Install FFmpeg on macOS
-        if: runner.os == 'macOS'
-        run: brew install ffmpeg
+      - name: Cache verified Kokoro model
+        if: matrix.python-version == '3.13'
+        uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6.1.0
+        with:
+          path: ${{ runner.temp }}/agent-voice-data/models
+          key: kokoro-v1-int8-${{ runner.os }}-${{ runner.arch }}
 
       - name: Lint
-        run: uv run --frozen ruff check src tests
+        if: matrix.os == 'ubuntu-latest' && matrix.python-version == '3.11'
+        run: uv run --frozen ruff check src tests .github/scripts
 
       - name: Test
         run: uv run --frozen pytest -q
 
-      - name: Build packages
-        if: matrix.os == 'ubuntu-latest' && matrix.python-version == '3.11'
+      - name: Build package
+        if: matrix.python-version == '3.13'
         run: uv build
 
+      - name: Install built wheel
+        if: matrix.python-version == '3.13'
+        run: |
+          uv venv --python ${{ matrix.python-version }} .ci-venv
+          uv pip install --python .ci-venv/bin/python dist/*.whl
+
+      - name: Verify installed package with real model, audio, doctor, and service
+        if: matrix.python-version == '3.13'
+        env:
+          AGENT_VOICE_HOME: ${{ runner.temp }}/agent-voice-data
+        run: |
+          .ci-venv/bin/python .github/scripts/verify_package.py .ci-venv/bin/agent-voice
+
   windows:
-    name: windows-latest / Python ${{ matrix.python-version }} / package E2E
+    name: windows-latest / Python ${{ matrix.python-version }}
     runs-on: windows-latest
     strategy:
       fail-fast: false
@@ -63,34 +79,32 @@ jobs:
           enable-cache: true
           python-version: ${{ matrix.python-version }}
 
-      - name: Install FFmpeg
-        run: choco install ffmpeg --version=7.1.1 --yes --no-progress --allow-downgrade
-
       - name: Cache verified Kokoro model
-        uses: actions/cache@v4
+        if: matrix.python-version == '3.13'
+        uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6.1.0
         with:
-          path: ${{ runner.temp }}/kokoro-data/models
-          key: kokoro-v1-int8-windows
-
-      - name: Lint
-        run: uv run --frozen ruff check src tests .github/scripts
+          path: ${{ runner.temp }}/agent-voice-data/models
+          key: kokoro-v1-int8-${{ runner.os }}-${{ runner.arch }}
 
       - name: Test
         run: uv run --frozen pytest -q
 
       - name: Build package
+        if: matrix.python-version == '3.13'
         run: uv build
 
       - name: Install built wheel
+        if: matrix.python-version == '3.13'
         shell: pwsh
         run: |
           uv venv --python ${{ matrix.python-version }} .ci-venv
           $wheel = (Get-ChildItem dist\*.whl).FullName
           uv pip install --python .ci-venv\Scripts\python.exe $wheel
 
       - name: Verify installed package with real model, audio, doctor, and service
+        if: matrix.python-version == '3.13'
         shell: pwsh
         env:
-          KOKORO_HOME: ${{ runner.temp }}/kokoro-data
+          AGENT_VOICE_HOME: ${{ runner.temp }}/agent-voice-data
         run: |
-          .ci-venv\Scripts\python.exe .github\scripts\verify_windows.py .ci-venv\Scripts\kokoro.exe
+          .ci-venv\Scripts\python.exe .github\scripts\verify_package.py .ci-venv\Scripts\agent-voice.exe

.github/workflows/publish.yml

@@ -36,7 +36,7 @@ jobs:
     runs-on: ubuntu-latest
     environment:
       name: pypi
-      url: https://pypi.org/p/kokoro-cli
+      url: https://pypi.org/p/agent-voice
     permissions:
       id-token: write
 

.gitignore

@@ -1,14 +1,21 @@
 .venv/
+.DS_Store
 __pycache__/
 .pytest_cache/
 .ruff_cache/
+.agents/
+skills-lock.json
 *.pyc
 *.egg-info/
 dist/
 config.json
+service-start.lock
+viewer.lock
+viewer.json
 models/*.onnx
 models/*.bin
 models/*.lock
 recordings/*
+IDEAS.md
 !models/.gitkeep
 !recordings/.gitkeep