@n1creator/openacp-cli 2026.712.3 → 2026.712.5

package/dist/speech/transcribe_audio.sh

@@ -0,0 +1,118 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# Keep the former wrapper cache as the fallback so existing model downloads and
+# virtual environments are reused after migration to the native provider.
+CACHE_ROOT="${OPENACP_LOCAL_WHISPER_CACHE:-${CODEX_TRANSCRIBE_VOICE_CACHE:-${XDG_CACHE_HOME:-$HOME/.cache}/codex/transcribe-voice}}"
+VENV_DIR="$CACHE_ROOT/venv"
+PYTHON="$VENV_DIR/bin/python"
+PACKAGE_MARKER="$CACHE_ROOT/.faster-whisper-1.2"
+
+usage() {
+  cat >&2 <<'EOF'
+Usage: transcribe_audio.sh [options] <audio-file>
+
+Options:
+  --model NAME          faster-whisper model name (default: base)
+  --language CODE       Whisper language code, for example ru or en (default: ru)
+  --beam-size N         Beam size for decoding (default: 5)
+  --vad-filter          Enable VAD filtering
+  --no-vad-filter       Disable VAD filtering (default)
+  --segments            Print timestamped segments instead of one plain transcript
+  --device NAME         Device for faster-whisper (default: cpu)
+  --compute-type NAME   Compute type (default: int8)
+  --quiet               Suppress metadata on stderr
+EOF
+}
+
+ensure_env() {
+  mkdir -p "$CACHE_ROOT"
+  if [[ ! -x "$PYTHON" ]]; then
+    if command -v uv >/dev/null 2>&1; then
+      uv venv "$VENV_DIR" >&2
+    else
+      local python_bin="${PYTHON_BIN:-python3}"
+      "$python_bin" -m venv "$VENV_DIR" >&2 || {
+        echo "Could not create a Python virtual environment. Install uv or python3-venv." >&2
+        exit 2
+      }
+    fi
+    rm -f "$PACKAGE_MARKER"
+  fi
+
+  if [[ ! -f "$PACKAGE_MARKER" ]]; then
+    if command -v uv >/dev/null 2>&1; then
+      uv pip install --python "$PYTHON" 'faster-whisper>=1.2,<2' >&2
+    else
+      "$PYTHON" -m pip install 'faster-whisper>=1.2,<2' >&2 || {
+        echo "Could not install faster-whisper. Install uv or ensure pip is available in the venv." >&2
+        exit 2
+      }
+    fi
+    touch "$PACKAGE_MARKER"
+  fi
+}
+
+if [[ $# -eq 0 ]]; then
+  usage
+  exit 2
+fi
+
+ensure_env
+
+"$PYTHON" - "$@" <<'PY'
+import argparse
+import os
+import sys
+from pathlib import Path
+
+from faster_whisper import WhisperModel
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Transcribe an audio file with faster-whisper.")
+    parser.add_argument("audio_file", help="Path to the audio file to transcribe")
+    parser.add_argument("--model", default=os.environ.get("OPENACP_LOCAL_WHISPER_MODEL", "base"))
+    parser.add_argument("--language", default=os.environ.get("OPENACP_LOCAL_WHISPER_LANGUAGE", "ru"))
+    parser.add_argument("--beam-size", type=int, default=int(os.environ.get("OPENACP_LOCAL_WHISPER_BEAM_SIZE", "5")))
+    parser.add_argument("--vad-filter", dest="vad_filter", action="store_true", default=False)
+    parser.add_argument("--no-vad-filter", dest="vad_filter", action="store_false")
+    parser.add_argument("--segments", action="store_true", help="Print timestamped segments")
+    parser.add_argument("--device", default=os.environ.get("OPENACP_LOCAL_WHISPER_DEVICE", "cpu"))
+    parser.add_argument("--compute-type", default=os.environ.get("OPENACP_LOCAL_WHISPER_COMPUTE_TYPE", "int8"))
+    parser.add_argument("--quiet", action="store_true", help="Suppress metadata on stderr")
+    args = parser.parse_args()
+
+    audio_path = Path(args.audio_file).expanduser()
+    if not audio_path.exists():
+        print(f"Audio file not found: {audio_path}", file=sys.stderr)
+        return 2
+
+    model = WhisperModel(args.model, device=args.device, compute_type=args.compute_type)
+    segments_iter, info = model.transcribe(
+        str(audio_path),
+        language=args.language,
+        beam_size=args.beam_size,
+        vad_filter=args.vad_filter,
+    )
+    segments = list(segments_iter)
+
+    if not args.quiet:
+        probability = getattr(info, "language_probability", None)
+        probability_text = f"{probability:.3f}" if probability is not None else "unknown"
+        print(
+            f"model={args.model} language={info.language} "
+            f"language_probability={probability_text} duration={info.duration:.3f}s",
+            file=sys.stderr,
+        )
+
+    if args.segments:
+        for segment in segments:
+            print(f"[{segment.start:.2f}-{segment.end:.2f}] {segment.text.strip()}")
+    else:
+        print(" ".join(segment.text.strip() for segment in segments).strip())
+    return 0
+
+
+raise SystemExit(main())
+PY

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@n1creator/openacp-cli",
-  "version": "2026.712.3",
+  "version": "2026.712.5",
   "description": "Self-hosted bridge for AI coding agents via ACP protocol",
   "type": "module",
   "bin": {