Implement audio.cpp source synchronization and build enhancements

- Introduced `_sync_git_checkout` method in `AudioCppManager` for managing source synchronization with Git, including fetching and resetting branches.
- Added `_compile_tree` method to streamline the build process for audio.cpp, encapsulating configuration, building, and validation of binaries.
- Enhanced API routes to support audio.cpp source synchronization, including task management and progress tracking.
- Updated frontend components to reflect new synchronization capabilities and improve user interaction during build processes.
- Added tests to validate the synchronization and build functionalities for audio.cpp.

Author: lapy

backend/audio_align_profiles.py

@@ -0,0 +1,48 @@
+"""Curated forced-alignment family guidance."""
+
+from __future__ import annotations
+
+from typing import Any, Dict, List, Optional
+
+from backend.audio_profile_fields import field_spec
+
+_ALIGN_TASKS = frozenset({"align"})
+
+
+def is_align_task(task: Optional[str]) -> bool:
+    return str(task or "").strip().lower() in _ALIGN_TASKS
+
+
+def align_profile_for_family(family: Optional[str]) -> Optional[Dict[str, Any]]:
+    key = str(family or "").strip().lower()
+    return _FAMILY_PROFILES.get(key) if key else None
+
+
+def alignment_request_field_groups(family: Optional[str]) -> List[Dict[str, Any]]:
+    profile = align_profile_for_family(family) or {}
+    groups: List[Dict[str, Any]] = []
+    for group_id, label, description in _GROUP_META:
+        keys = profile.get(f"{group_id}_fields") or []
+        fields = [field_spec(key) for key in keys]
+        if fields:
+            groups.append({"id": group_id, "label": label, "description": description, "fields": fields})
+    return groups
+
+
+_GROUP_META = [
+    ("audio", "Audio & transcript", "Speech audio and the exact transcript to align."),
+    ("context", "Language", "Transcript language hint."),
+    ("chunking", "Chunking", "Chunking mode for long audio."),
+]
+
+_FAMILY_PROFILES: Dict[str, Dict[str, Any]] = {
+    "qwen3_forced_aligner": {
+        "label": "Qwen3 Forced Aligner",
+        "workflows": ["offline"],
+        "summary": "Map an exact transcript onto speech audio to produce word timestamps.",
+        "audio_fields": ["audio", "transcript"],
+        "context_fields": ["language"],
+        "chunking_fields": ["audio_chunk_mode"],
+        "api_hint": "Not an ASR route — the transcript is required input. For long audio timestamps, use Qwen3 ASR with words_out.",
+    },
+}

backend/audio_analysis_profiles.py

@@ -0,0 +1,97 @@
+"""Curated VAD and diarization family guidance."""
+
+from __future__ import annotations
+
+from typing import Any, Dict, List, Optional
+
+from backend.audio_profile_fields import field_spec
+
+_VAD_TASKS = frozenset({"vad"})
+_DIAR_TASKS = frozenset({"diar"})
+
+
+def is_vad_task(task: Optional[str]) -> bool:
+    return str(task or "").strip().lower() in _VAD_TASKS
+
+
+def is_diar_task(task: Optional[str]) -> bool:
+    return str(task or "").strip().lower() in _DIAR_TASKS
+
+
+def is_analysis_task(task: Optional[str]) -> bool:
+    return is_vad_task(task) or is_diar_task(task)
+
+
+def analysis_profile_for_family(family: Optional[str]) -> Optional[Dict[str, Any]]:
+    key = str(family or "").strip().lower()
+    return _FAMILY_PROFILES.get(key) if key else None
+
+
+def analysis_request_field_groups(family: Optional[str]) -> List[Dict[str, Any]]:
+    profile = analysis_profile_for_family(family) or {}
+    groups: List[Dict[str, Any]] = []
+    for group_id, label, description in _GROUP_META:
+        keys = profile.get(f"{group_id}_fields") or []
+        fields = [field_spec(key) for key in keys]
+        if fields:
+            groups.append({"id": group_id, "label": label, "description": description, "fields": fields})
+    return groups
+
+
+_GROUP_META = [
+    ("audio", "Audio input", "Input audio for analysis."),
+    ("session", "Session mode", "Streaming requires server mode=streaming."),
+    ("chunking", "VAD chunk planning", "Offline VAD chunk window controls."),
+    ("options", "Detection options", "Threshold and segment timing controls."),
+]
+
+_FAMILY_PROFILES: Dict[str, Dict[str, Any]] = {
+    "silero_vad": {
+        "label": "Silero VAD",
+        "workflows": ["offline", "streaming"],
+        "summary": "Speech activity detection with offline and streaming modes.",
+        "audio_fields": ["audio"],
+        "session_fields": ["stream"],
+        "chunking_fields": [
+            "vad_chunk_max_seconds",
+            "vad_chunk_merge_gap_seconds",
+            "vad_chunk_padding_seconds",
+        ],
+        "options_fields": [
+            "threshold",
+            "neg_threshold",
+            "min_speech_duration_ms",
+            "min_silence_duration_ms",
+            "speech_pad_ms",
+        ],
+        "api_hint": "Use 16 kHz WAV. Streaming mode emits partial segments.",
+    },
+    "marblenet_vad": {
+        "label": "MarbleNet VAD",
+        "workflows": ["offline"],
+        "summary": "Offline speech activity detection.",
+        "audio_fields": ["audio"],
+        "options_fields": ["threshold"],
+        "api_hint": "Offline only; writes segment JSON.",
+    },
+    "marblenet": {
+        "label": "MarbleNet VAD",
+        "workflows": ["offline"],
+        "summary": "Offline speech activity detection.",
+        "audio_fields": ["audio"],
+        "options_fields": ["threshold"],
+    },
+    "sortformer_diar": {
+        "label": "Sortformer Diarization",
+        "workflows": ["offline"],
+        "summary": "Speaker diarization for meetings and conversations.",
+        "audio_fields": ["audio"],
+        "api_hint": "Default package supports up to 4 speakers.",
+    },
+    "sortformer": {
+        "label": "Sortformer Diarization",
+        "workflows": ["offline"],
+        "summary": "Speaker diarization for meetings and conversations.",
+        "audio_fields": ["audio"],
+    },
+}