Add optional blake3 hashing during asset scanning

- Make blake3 import lazy in hashing.py (only imported when needed)
- Add compute_hashes parameter to AssetSeeder.start(), build_asset_specs(), and seed_assets()
- Fix missing tag clearing: include is_missing states in sync when update_missing_tags=True
- Clear is_missing flag on cache states when files are restored with matching mtime/size
- Fix validation error serialization in routes.py (use json.loads(ve.json()))

Amp-Thread-ID: https://ampcode.com/threads/T-019c3614-56d4-74a8-a717-19922d6dbbee
Co-authored-by: Amp <amp@ampcode.com>

app/assets/services/asset_management.py

@@ -272,7 +272,9 @@ def resolve_asset_for_download(
         states = list_cache_states_by_asset_id(session, asset_id=asset.id)
         abs_path = select_best_live_path(states)
         if not abs_path:
-            raise FileNotFoundError
+            raise FileNotFoundError(
+                f"No live path for AssetInfo {asset_info_id} (asset id={asset.id}, name={info.name})"
+            )
 
         update_asset_info_access_time(session, asset_info_id=asset_info_id)
         session.commit()

app/assets/services/bulk_ingest.py

@@ -36,6 +36,7 @@ class SeedAssetSpec(TypedDict):
     tags: list[str]
     fname: str
     metadata: ExtractedMetadata | None
+    hash: str | None
 
 
 class AssetRow(TypedDict):
@@ -163,7 +164,7 @@ def batch_insert_seed_assets(
         asset_rows.append(
             {
                 "id": asset_id,
-                "hash": None,
+                "hash": spec.get("hash"),
                 "size_bytes": spec["size_bytes"],
                 "mime_type": None,
                 "created_at": current_time,

app/assets/services/file_utils.py

@@ -23,15 +23,16 @@ def verify_file_unchanged(
 
     Returns True if the file's mtime and size match the database values.
     Returns False if mtime_db is None or values don't match.
+
+    size_db=None means don't check size; 0 is a valid recorded size.
     """
     if mtime_db is None:
         return False
     actual_mtime_ns = get_mtime_ns(stat_result)
     if int(mtime_db) != int(actual_mtime_ns):
         return False
-    sz = int(size_db or 0)
-    if sz > 0:
-        return int(stat_result.st_size) == sz
+    if size_db is not None:
+        return int(stat_result.st_size) == int(size_db)
     return True
 
 

app/assets/services/hashing.py

@@ -2,10 +2,23 @@ import asyncio
 import os
 from typing import IO
 
-from blake3 import blake3
-
 DEFAULT_CHUNK = 8 * 1024 * 1024
 
+_blake3 = None
+
+
+def _get_blake3():
+    global _blake3
+    if _blake3 is None:
+        try:
+            from blake3 import blake3 as _b3
+            _blake3 = _b3
+        except ImportError:
+            raise ImportError(
+                "blake3 is required for asset hashing. Install with: pip install blake3"
+            )
+    return _blake3
+
 
 def compute_blake3_hash(
     fp: str | IO[bytes],
@@ -42,7 +55,7 @@ def _hash_file_obj(file_obj: IO, chunk_size: int = DEFAULT_CHUNK) -> str:
         if orig_pos != 0:
             file_obj.seek(0)
 
-        h = blake3()
+        h = _get_blake3()()
         while True:
             chunk = file_obj.read(chunk_size)
             if not chunk: