media-engine 0.1.0__py3-none-any.whl

cli/transcript.py

@@ -0,0 +1,76 @@
+#!/usr/bin/env python3
+"""Transcribe audio from video file."""
+
+import argparse
+import json
+import logging
+import sys
+import time
+
+from media_engine.extractors import extract_transcript
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Transcribe audio from video file")
+    parser.add_argument("file", help="Path to video file")
+    parser.add_argument(
+        "--model",
+        type=str,
+        default="auto",
+        choices=["auto", "tiny", "small", "medium", "large-v3"],
+        help="Whisper model size (default: auto)",
+    )
+    parser.add_argument(
+        "--language",
+        type=str,
+        default=None,
+        help="Force language code (e.g., 'en', 'no')",
+    )
+    parser.add_argument(
+        "--fallback-language",
+        type=str,
+        default="en",
+        help="Fallback language for short clips (default: en)",
+    )
+    parser.add_argument("-v", "--verbose", action="store_true", help="Verbose output")
+    parser.add_argument("--json", action="store_true", help="Output as JSON")
+
+    args = parser.parse_args()
+
+    if args.verbose:
+        logging.basicConfig(level=logging.DEBUG)
+    else:
+        logging.basicConfig(level=logging.WARNING)
+
+    try:
+        start_time = time.perf_counter()
+        result = extract_transcript(
+            args.file,
+            model=args.model,
+            language=args.language,
+            fallback_language=args.fallback_language,
+        )
+        elapsed = time.perf_counter() - start_time
+
+        if args.json:
+            output = result.model_dump()
+            output["elapsed_seconds"] = round(elapsed, 2)
+            print(json.dumps(output, indent=2, default=str))
+        else:
+            print(f"File: {args.file}")
+            print(f"Language: {result.language} (confidence: {result.confidence:.2f})")
+            print(f"Segments: {len(result.segments)}")
+            print()
+            for seg in result.segments:
+                speaker = f"[{seg.speaker}] " if seg.speaker else ""
+                print(f"  [{seg.start:.2f}s - {seg.end:.2f}s] {speaker}{seg.text}")
+            print()
+            print(f"Elapsed: {elapsed:.2f}s")
+
+    except Exception as e:
+        print(f"Error: {e}", file=sys.stderr)
+        sys.exit(1)
+
+
+if __name__ == "__main__":
+    main()

media_engine/__init__.py

@@ -0,0 +1,7 @@
+"""Media Engine - AI-powered video extraction API."""
+
+try:
+    from media_engine._version import __version__
+except ImportError:
+    # Not installed, running from source without build
+    __version__ = "0.0.0.dev0"

media_engine/_version.py

@@ -0,0 +1,34 @@
+# file generated by setuptools-scm
+# don't change, don't track in version control
+
+__all__ = [
+    "__version__",
+    "__version_tuple__",
+    "version",
+    "version_tuple",
+    "__commit_id__",
+    "commit_id",
+]
+
+TYPE_CHECKING = False
+if TYPE_CHECKING:
+    from typing import Tuple
+    from typing import Union
+
+    VERSION_TUPLE = Tuple[Union[int, str], ...]
+    COMMIT_ID = Union[str, None]
+else:
+    VERSION_TUPLE = object
+    COMMIT_ID = object
+
+version: str
+__version__: str
+__version_tuple__: VERSION_TUPLE
+version_tuple: VERSION_TUPLE
+commit_id: COMMIT_ID
+__commit_id__: COMMIT_ID
+
+__version__ = version = '0.1.0'
+__version_tuple__ = version_tuple = (0, 1, 0)
+
+__commit_id__ = commit_id = None

media_engine/app.py

@@ -0,0 +1,80 @@
+"""FastAPI app factory for Media Engine."""
+
+import atexit
+import logging
+
+from fastapi import FastAPI
+from fastapi.middleware.cors import CORSMiddleware
+
+from media_engine import __version__
+from media_engine.batch.timing import save_timing_history
+from media_engine.extractors import (
+    shutdown_ffprobe_pool,
+    unload_clip_model,
+    unload_face_model,
+    unload_ocr_model,
+    unload_qwen_model,
+    unload_vad_model,
+    unload_whisper_model,
+    unload_yolo_model,
+)
+from media_engine.routers import (
+    batch_router,
+    health_router,
+    models_router,
+    settings_router,
+    utils_router,
+)
+
+logger = logging.getLogger(__name__)
+
+
+def cleanup_resources() -> None:
+    """Clean up all resources.
+
+    Note: This runs during Python shutdown via atexit, so we must be careful
+    not to import new modules or use logging (file handlers may be closed).
+    """
+    try:
+        # Save timing history before shutdown
+        save_timing_history()
+        shutdown_ffprobe_pool()
+        unload_whisper_model()
+        unload_qwen_model()
+        unload_yolo_model()
+        unload_clip_model()
+        unload_ocr_model()
+        unload_face_model()
+        unload_vad_model()
+    except Exception:
+        pass  # Suppress errors during shutdown
+
+
+def create_app() -> FastAPI:
+    """Create and configure the FastAPI application."""
+    app = FastAPI(
+        title="Media Engine",
+        description="AI-powered video extraction API",
+        version=__version__,
+    )
+
+    # Add CORS middleware
+    app.add_middleware(
+        CORSMiddleware,
+        allow_origins=["*"],
+        allow_credentials=True,
+        allow_methods=["*"],
+        allow_headers=["*"],
+    )
+
+    # Include routers
+    app.include_router(batch_router)
+    app.include_router(health_router)
+    app.include_router(settings_router)
+    app.include_router(models_router)
+    app.include_router(utils_router)
+
+    # Register cleanup on exit
+    atexit.register(cleanup_resources)
+
+    return app

media_engine/batch/__init__.py

@@ -0,0 +1,56 @@
+"""Batch processing system for Polybos Media Engine."""
+
+from media_engine.batch.models import (
+    BatchFileStatus,
+    BatchJobStatus,
+    BatchRequest,
+    ExtractorTiming,
+    JobProgress,
+)
+from media_engine.batch.processor import run_batch_job
+from media_engine.batch.queue import (
+    cleanup_expired_batch_jobs,
+    start_next_batch,
+    update_queue_positions,
+)
+from media_engine.batch.state import (
+    batch_jobs,
+    batch_jobs_lock,
+    batch_queue,
+    batch_queue_lock,
+    is_batch_running,
+    set_batch_running,
+)
+from media_engine.batch.timing import (
+    calculate_queue_eta,
+    get_enabled_extractors_from_request,
+    load_timing_history,
+    save_timing_history,
+)
+
+__all__ = [
+    # Models
+    "BatchFileStatus",
+    "BatchJobStatus",
+    "BatchRequest",
+    "ExtractorTiming",
+    "JobProgress",
+    # State
+    "batch_jobs",
+    "batch_jobs_lock",
+    "batch_queue",
+    "batch_queue_lock",
+    "is_batch_running",
+    "set_batch_running",
+    # Queue
+    "cleanup_expired_batch_jobs",
+    "start_next_batch",
+    "update_queue_positions",
+    # Timing
+    "calculate_queue_eta",
+    "get_enabled_extractors_from_request",
+    "load_timing_history",
+    "save_timing_history",
+    # Processor
+    "run_batch_job",
+]

media_engine/batch/models.py

@@ -0,0 +1,99 @@
+"""Pydantic models for batch processing."""
+
+from datetime import datetime
+from typing import Any
+
+from pydantic import BaseModel
+
+
+class JobProgress(BaseModel):
+    """Progress within current extraction step."""
+
+    message: str  # e.g., "Loading model...", "Processing frame 2/5"
+    current: int | None = None
+    total: int | None = None
+    # ETA tracking
+    stage_elapsed_seconds: float | None = None  # Time spent in current stage
+    eta_seconds: float | None = None  # Estimated seconds remaining for current stage
+    # Total ETA fields (for full batch/queue visibility)
+    total_eta_seconds: float | None = None  # Total time remaining for entire batch
+    queue_eta_seconds: float | None = None  # Total time remaining for all queued batches
+    queued_batches: int | None = None  # Number of batches waiting in queue
+
+
+# Job TTL for automatic cleanup (1 hour)
+JOB_TTL_SECONDS = 3600
+
+
+class BatchRequest(BaseModel):
+    """Request for batch extraction.
+
+    Model selection is configured via global settings (GET/PUT /settings).
+    This keeps batch requests simple and hardware config in one place.
+    """
+
+    files: list[str]
+    enable_metadata: bool = True
+    enable_vad: bool = False  # Voice Activity Detection
+    enable_scenes: bool = False
+    enable_transcript: bool = False
+    enable_faces: bool = False
+    enable_objects: bool = False  # YOLO object detection (fast, bounding boxes)
+    enable_visual: bool = False  # Qwen VLM scene descriptions (slower, richer)
+    enable_clip: bool = False
+    enable_ocr: bool = False
+    enable_motion: bool = False
+
+    # Context for Whisper
+    language: str | None = None  # Force specific language (ISO 639-1 code, e.g., "en", "no")
+    language_hints: list[str] | None = None  # Hints (currently unused by Whisper)
+    context_hint: str | None = None
+    # Context for Qwen VLM - per-file context mapping (file path -> context dict)
+    # Example: {"/path/video1.mp4": {"location": "Oslo"}, "/path/video2.mp4": {"location": "Bergen"}}
+    contexts: dict[str, dict[str, str]] | None = None
+    # Per-file timestamps for visual/VLM analysis (file path -> list of timestamps)
+    # Example: {"/path/video1.mp4": [10.0, 30.0], "/path/video2.mp4": [5.0, 15.0, 25.0]}
+    visual_timestamps: dict[str, list[float]] | None = None
+    # Optional LUT path for visual analysis (e.g., for log footage color correction)
+    # Applied to extracted frames before sending to Qwen
+    lut_path: str | None = None
+
+
+class BatchFileStatus(BaseModel):
+    """Status for a single file in a batch."""
+
+    file: str
+    filename: str
+    status: str  # pending, running, completed, failed
+    results: dict[str, Any] = {}
+    error: str | None = None
+    timings: dict[str, float] = {}  # extractor -> seconds
+    extractor_status: dict[str, str] = {}  # extractor -> pending/active/completed/failed/skipped
+
+
+class ExtractorTiming(BaseModel):
+    """Timing for a single extractor stage."""
+
+    extractor: str
+    started_at: datetime
+    completed_at: datetime | None = None
+    duration_seconds: float | None = None
+    files_processed: int = 0
+
+
+class BatchJobStatus(BaseModel):
+    """Status of a batch extraction job."""
+
+    batch_id: str
+    status: str  # queued, pending, running, completed, failed
+    queue_position: int | None = None  # Position in queue (1 = next to run), None if not queued
+    current_extractor: str | None = None
+    progress: JobProgress | None = None
+    files: list[BatchFileStatus] = []
+    created_at: datetime
+    completed_at: datetime | None = None
+    # Timing and resource metrics
+    extractor_timings: list[ExtractorTiming] = []
+    elapsed_seconds: float | None = None
+    memory_mb: int | None = None  # Current process memory
+    peak_memory_mb: int | None = None  # Peak process memory during batch