refactor storage minio for k8s

core/storage/blob.py

@@ -0,0 +1,112 @@
+"""
+Cloud-agnostic blob storage interface.
+
+All file-based sources (chunks, uploads, checkpoints) go through MinIO.
+Local dev runs MinIO in docker-compose — same code path as production.
+Production changes S3_ENDPOINT_URL; nothing else changes.
+
+Single bucket, multiple prefixes:
+  in/              — source media
+  out/             — transcoded chunks
+  checkpoints/     — detection intermediate blobs (frames, crops)
+
+Each prefix is independently configurable via env vars so they can
+be split into separate buckets later if needed.
+
+Nothing outside core/storage/ should import boto3 directly.
+"""
+
+from __future__ import annotations
+
+import os
+from dataclasses import dataclass
+from typing import Optional
+
+
+# Single bucket, prefix-based layout
+BUCKET = os.environ.get("S3_BUCKET", "mpr")
+PREFIX_IN = os.environ.get("S3_PREFIX_IN", "in/")
+PREFIX_OUT = os.environ.get("S3_PREFIX_OUT", "out/")
+PREFIX_CHECKPOINTS = os.environ.get("S3_PREFIX_CHECKPOINTS", "checkpoints/")
+
+
+@dataclass
+class BlobObject:
+    key: str
+    filename: str
+    size_bytes: int
+
+
+class BlobStore:
+    """
+    Thin wrapper over the S3-compatible storage backend (MinIO / AWS S3).
+
+    All configuration (endpoint URL, credentials, region) is read from
+    environment variables by the underlying s3 module.
+    """
+
+    def __init__(self, bucket: str, prefix: str = ""):
+        self.bucket = bucket
+        self.prefix = prefix
+
+    def _full_prefix(self, prefix: str) -> str:
+        """Combine store prefix with caller prefix."""
+        return self.prefix + prefix
+
+    def list(
+        self,
+        prefix: str = "",
+        extensions: Optional[set[str]] = None,
+    ) -> list[BlobObject]:
+        """List objects in the bucket, optionally filtered by extension."""
+        from core.storage.s3 import list_objects
+
+        full = self._full_prefix(prefix)
+        raw = list_objects(self.bucket, prefix=full, extensions=extensions)
+        objects = []
+        for obj in raw:
+            blob = BlobObject(
+                key=obj["key"],
+                filename=obj["filename"],
+                size_bytes=obj["size"],
+            )
+            objects.append(blob)
+        return objects
+
+    def download_to_temp(self, key: str) -> str:
+        """Download a blob to a temp file. Caller is responsible for cleanup."""
+        from core.storage.s3 import download_to_temp
+
+        return download_to_temp(self.bucket, key)
+
+    def upload(self, local_path: str, key: str) -> None:
+        """Upload a local file to the bucket."""
+        from core.storage.s3 import upload_file
+
+        upload_file(local_path, self.bucket, key)
+
+    def get_url(self, key: str, expires: int = 3600) -> str:
+        """Return a presigned URL for the given key."""
+        from core.storage.s3 import get_presigned_url
+
+        return get_presigned_url(self.bucket, key, expires=expires)
+
+
+def get_store(purpose: str = "out") -> BlobStore:
+    """
+    Return a BlobStore for the given purpose.
+
+    Purposes map to prefixes:
+      "in"          → source media     (S3_PREFIX_IN)
+      "out"         → transcoded output (S3_PREFIX_OUT)
+      "checkpoints" → detection blobs   (S3_PREFIX_CHECKPOINTS)
+
+    All share the same bucket (S3_BUCKET), each scoped to its prefix.
+    """
+    prefix_map = {
+        "in": PREFIX_IN,
+        "out": PREFIX_OUT,
+        "checkpoints": PREFIX_CHECKPOINTS,
+    }
+    prefix = prefix_map.get(purpose, "")
+    return BlobStore(BUCKET, prefix=prefix)