feat(backend): add Other model type (VAE/upscaler/text encoder) scanner, service and routes

This commit is contained in:
Will Miao
2026-09-12 11:25:51 +08:00
parent 3070838a42
commit 27da7b3ca3
22 changed files with 1771 additions and 19 deletions
+6 -1
View File
@@ -7,7 +7,7 @@ import logging
import os
import time
from ..utils.constants import VALID_LORA_SUB_TYPES, VALID_CHECKPOINT_SUB_TYPES
from ..utils.constants import VALID_LORA_SUB_TYPES, VALID_CHECKPOINT_SUB_TYPES, VALID_OTHER_SUB_TYPES
from ..utils.models import BaseModelMetadata
from ..utils.metadata_manager import MetadataManager
from ..utils.usage_stats import UsageStats
@@ -904,6 +904,11 @@ class BaseModelService(ABC):
and normalized_type not in VALID_CHECKPOINT_SUB_TYPES
):
continue
if (
self.model_type == "other"
and normalized_type not in VALID_OTHER_SUB_TYPES
):
continue
type_counts[normalized_type] = type_counts.get(normalized_type, 0) + 1
+1
View File
@@ -68,6 +68,7 @@ PAGE_TYPE_MAP = {
'lora': 'loras',
'checkpoint': 'checkpoints',
'embedding': 'embeddings',
'other': 'other',
}
+10 -5
View File
@@ -118,19 +118,24 @@ class ModelServiceFactory:
def register_default_model_types():
"""Register the default model types (LoRA, Checkpoint, and Embedding)"""
"""Register the default model types (LoRA, Checkpoint, Embedding, and Other)"""
from ..services.lora_service import LoraService
from ..services.checkpoint_service import CheckpointService
from ..services.embedding_service import EmbeddingService
from ..services.other_model_service import OtherModelService
from ..routes.lora_routes import LoraRoutes
from ..routes.checkpoint_routes import CheckpointRoutes
from ..routes.embedding_routes import EmbeddingRoutes
from ..routes.other_routes import OtherRoutes
# Register LoRA model type
ModelServiceFactory.register_model_type('lora', LoraService, LoraRoutes)
# Register Checkpoint model type
ModelServiceFactory.register_model_type('checkpoint', CheckpointService, CheckpointRoutes)
# Register Embedding model type
ModelServiceFactory.register_model_type('embedding', EmbeddingService, EmbeddingRoutes)
ModelServiceFactory.register_model_type('embedding', EmbeddingService, EmbeddingRoutes)
# Register Other model type (VAE, upscaler, text encoder, ...)
ModelServiceFactory.register_model_type('other', OtherModelService, OtherRoutes)
+79
View File
@@ -0,0 +1,79 @@
import os
import logging
from typing import Any, Dict, Optional
from .base_model_service import BaseModelService
from .auto_tag_service import extract_auto_tags
from ..utils.models import OtherModelMetadata
from ..config import config
logger = logging.getLogger(__name__)
class OtherModelService(BaseModelService):
"""Other-model-specific service implementation (VAE, upscaler, text encoder, ...)"""
def __init__(self, scanner, update_service=None):
"""Initialize Other-model service
Args:
scanner: Other-model scanner instance
update_service: Optional service for remote update tracking.
"""
super().__init__("other", scanner, OtherModelMetadata, update_service=update_service)
async def format_response(self, model_data: Dict[str, Any]) -> Optional[Dict[str, Any]]:
"""Format other-model data for API response.
Returns None when the entry is missing critical fields (corrupted cache
row), so the handler layer can filter it out. See issue #730.
"""
# Guard against corrupted cache entries missing critical fields
file_path = model_data.get("file_path")
if not file_path or not isinstance(file_path, str):
logger.warning(
"Skipping corrupted other-model entry (missing file_path): %s",
model_data.get("file_name", "<unknown>"),
)
return None
# Get sub_type from cache entry (new canonical field)
sub_type = model_data.get("sub_type", "vae")
file_name = model_data.get("file_name") or ""
model_name = model_data.get("model_name") or file_name
folder = model_data.get("folder") or ""
return {
"model_name": model_name,
"file_name": file_name,
"preview_url": config.get_preview_static_url(model_data.get("preview_url", "")),
"preview_nsfw_level": model_data.get("preview_nsfw_level", 0),
"base_model": model_data.get("base_model", ""),
"folder": folder,
"sha256": model_data.get("sha256", ""),
"autov3": model_data.get("autov3"),
"file_path": file_path.replace(os.sep, "/"),
"file_size": model_data.get("size", 0),
"modified": model_data.get("modified", ""),
"tags": model_data.get("tags", []),
"from_civitai": model_data.get("from_civitai", True),
"notes": model_data.get("notes", ""),
"sub_type": sub_type,
"favorite": model_data.get("favorite", False),
"exclude": bool(model_data.get("exclude", False)),
"update_available": bool(model_data.get("update_available", False)),
"skip_metadata_refresh": bool(model_data.get("skip_metadata_refresh", False)),
"civitai": self.filter_civitai_data(model_data.get("civitai", {}), minimal=True),
"auto_tags": model_data.get("auto_tags") or extract_auto_tags(model_data),
"version_count": model_data.get("version_count"),
"hf_url": model_data.get("hf_url", ""),
}
def find_duplicate_hashes(self) -> Dict[str, Any]:
"""Find other models with duplicate SHA256 hashes"""
return self.scanner._hash_index.get_duplicate_hashes()
def find_duplicate_filenames(self) -> Dict[str, Any]:
"""Find other models with conflicting filenames"""
return self.scanner._hash_index.get_duplicate_filenames()
+468
View File
@@ -0,0 +1,468 @@
# pyright: reportImportCycles=false
# Lazy (function-local) imports still count as static edges in basedpyright's
# reportImportCycles, so the ServiceRegistry singleton pattern necessarily forms
# import cycles. Breaking them would require an architectural refactor.
import asyncio
import json
import logging
import os
from datetime import datetime
from typing import Any, Dict, List, Optional
from ..utils.models import OtherModelMetadata
from ..utils.file_utils import find_preview_file, normalize_path, calculate_autov3
from ..utils.metadata_manager import MetadataManager
from ..config import config
from .model_scanner import ModelScanner, _is_excluded_dir
from .model_hash_index import ModelHashIndex
logger = logging.getLogger(__name__)
class OtherScanner(ModelScanner):
"""Service for scanning and managing "other" model files.
Aggregates every enabled folder_paths category from
OTHER_MODEL_FOLDER_SUBTYPES (VAE, upscalers, text encoders, CLIP vision,
opt-in ControlNet) into one scanner; sub_type is derived from the root
containing the file (mirrors CheckpointScanner's checkpoints/unet split).
Hashing is lazy (checkpoint-style): text encoders can be ~10 GB, so the
initial scan records hash_status="pending" and the SHA256 is computed
on-demand via calculate_hash_for_model (e.g. when fetching CivitAI
metadata).
"""
def __init__(self):
# Same extension set as CheckpointScanner (ComfyUI's
# supported_pt_extensions plus ".gguf").
file_extensions = {
".ckpt",
".pt",
".pt2",
".bin",
".pth",
".safetensors",
".pkl",
".sft",
".gguf",
}
super().__init__(
model_type="other",
model_class=OtherModelMetadata,
file_extensions=file_extensions,
hash_index=ModelHashIndex(),
)
if not hasattr(self, "_hash_calculation_lock"):
self._hash_calculation_lock = asyncio.Lock()
self._hash_calculation_tasks: dict[str, asyncio.Task[Optional[str]]] = {}
async def _create_default_metadata(
self, file_path: str
) -> Optional[OtherModelMetadata]:
"""Create default metadata without calculating hash (lazy hash).
Other models include multi-GB text encoders, so hash calculation is
deferred until on-demand (e.g. CivitAI metadata fetch).
"""
try:
real_path = os.path.realpath(file_path)
if not os.path.exists(real_path):
logger.error(f"File not found: {file_path}")
return None
base_name = os.path.splitext(os.path.basename(file_path))[0]
dir_path = os.path.dirname(file_path)
# Find preview image
preview_url = find_preview_file(base_name, dir_path)
# AutoV3 reads only the safetensors header, so it is cheap even for
# large files; record the checked state at creation time ("" =
# checked but unavailable).
autov3 = calculate_autov3(real_path)
# Create metadata WITHOUT calculating hash
metadata = OtherModelMetadata(
file_name=base_name,
model_name=base_name,
file_path=normalize_path(file_path),
size=os.path.getsize(real_path),
modified=datetime.now().timestamp(),
sha256="", # Empty hash - will be calculated on-demand
base_model="Unknown",
preview_url=normalize_path(preview_url),
tags=[],
modelDescription="",
sub_type=self.resolve_sub_type_for_path(file_path) or "vae",
from_civitai=False, # Mark as local model since no hash yet
hash_status="pending", # Mark hash as pending
autov3=autov3 or "",
)
# Save the created metadata
logger.info(f"Creating other-model metadata (hash pending) for {file_path}")
await MetadataManager.save_metadata(file_path, metadata)
return metadata
except Exception as e:
logger.error(
f"Error creating default other-model metadata for {file_path}: {e}"
)
return None
async def calculate_hash_for_model(self, file_path: str) -> Optional[str]:
"""Calculate hash for a model on-demand with per-file singleflight.
Args:
file_path: Path to the model file
Returns:
SHA256 hash string, or None if calculation failed
"""
try:
real_path = os.path.realpath(file_path)
if not os.path.exists(real_path):
logger.error(f"File not found for hash calculation: {file_path}")
return None
metadata, _ = await MetadataManager.load_metadata(
file_path, self.model_class
)
if (
metadata is not None
and metadata.hash_status == "completed"
and metadata.sha256
):
# Ensure the in-memory hash index is populated even when
# the hash was already computed and persisted to the metadata
# file. Without this, usage tracking (and any other caller
# that queries get_hash_by_filename first) will miss on every
# lookup and keep calling back into this method, creating a
# tight loop that never populates the index.
self._hash_index.add_entry(
metadata.sha256.lower(),
file_path,
getattr(metadata, "autov3", None) or None,
)
return metadata.sha256
async with self._hash_calculation_lock:
metadata, _ = await MetadataManager.load_metadata(
file_path, self.model_class
)
if (
metadata is not None
and metadata.hash_status == "completed"
and metadata.sha256
):
self._hash_index.add_entry(
metadata.sha256.lower(),
file_path,
getattr(metadata, "autov3", None) or None,
)
return metadata.sha256
task = self._hash_calculation_tasks.get(real_path)
if task is None:
task = asyncio.create_task(
self._run_hash_calculation_task(file_path, real_path)
)
self._hash_calculation_tasks[real_path] = task
return await asyncio.shield(task)
except Exception as e:
logger.error(f"Error calculating hash for {file_path}: {e}")
return None
async def _run_hash_calculation_task(
self, file_path: str, real_path: str
) -> Optional[str]:
"""Run a hash calculation task and remove it from the in-flight map."""
try:
return await self._calculate_hash_for_model_uncached(file_path, real_path)
finally:
task = asyncio.current_task()
async with self._hash_calculation_lock:
if self._hash_calculation_tasks.get(real_path) is task:
del self._hash_calculation_tasks[real_path]
async def _calculate_hash_for_model_uncached(
self, file_path: str, real_path: str
) -> Optional[str]:
"""Calculate hash for a model without checking in-flight tasks."""
from ..utils.file_utils import calculate_sha256
try:
# Load current metadata
metadata, should_skip = await MetadataManager.load_metadata(
file_path, self.model_class
)
if metadata is None:
if should_skip:
logger.error(f"Invalid metadata found for {file_path}")
return None
created_metadata = await self._create_default_metadata(file_path)
if created_metadata is None:
logger.error(f"No metadata found for {file_path}")
return None
metadata = created_metadata
# Check if hash is already calculated
if metadata.hash_status == "completed" and metadata.sha256:
# Populate the in-memory hash index even for pre-computed
# hashes, mirroring the fix in calculate_hash_for_model.
self._hash_index.add_entry(
metadata.sha256.lower(),
file_path,
getattr(metadata, "autov3", None) or None,
)
return metadata.sha256
# Update status to calculating
metadata.hash_status = "calculating"
await MetadataManager.save_metadata(file_path, metadata)
# Calculate hash
logger.info(f"Calculating hash for other model: {file_path}")
sha256 = await calculate_sha256(real_path)
# Update metadata with hash
metadata.sha256 = sha256
metadata.hash_status = "completed"
await MetadataManager.save_metadata(file_path, metadata)
# Update hash index
self._hash_index.add_entry(
sha256.lower(),
file_path,
getattr(metadata, "autov3", None) or None,
)
# Update the in-memory cache entry so that subsequent
# _persist_current_cache / _save_persistent_cache calls
# write the hash back to the SQLite models table. Without
# this the hash only lives in the metadata file and the
# in-memory hash index, both of which are lost across
# restarts, causing the same re-computation loop on the
# next session.
if self._cache is not None and self._cache.raw_data:
for entry in self._cache.raw_data:
if entry.get("file_path") == file_path:
entry["sha256"] = sha256.lower()
entry["hash_status"] = "completed"
self.bump_cache_version()
break
logger.info(f"Hash calculated for other model: {file_path}")
return sha256
except Exception as e:
logger.error(f"Error calculating hash for {file_path}: {e}")
# Update status to failed
try:
metadata, _ = await MetadataManager.load_metadata(
file_path, self.model_class
)
if metadata:
metadata.hash_status = "failed"
await MetadataManager.save_metadata(file_path, metadata)
except Exception:
pass
return None
async def calculate_all_pending_hashes(
self, progress_callback=None
) -> Dict[str, int]:
"""Calculate hashes for all other models with pending hash status.
If cache is not initialized, scans filesystem directly for metadata files
with hash_status != 'completed'.
Args:
progress_callback: Optional callback(progress, total, current_file)
Returns:
Dict with 'completed', 'failed', 'total' counts
"""
# Try to get from cache first
cache = await self.get_cached_data()
if cache and cache.raw_data:
# Use cache if available
pending_models = [
item
for item in cache.raw_data
if item.get("hash_status") != "completed" or not item.get("sha256")
]
else:
# Cache not initialized, scan filesystem directly
pending_models = await self._find_pending_models_from_filesystem()
if not pending_models:
return {"completed": 0, "failed": 0, "total": 0}
total = len(pending_models)
completed = 0
failed = 0
for i, model_data in enumerate(pending_models):
file_path = model_data.get("file_path")
if not file_path:
continue
try:
sha256 = await self.calculate_hash_for_model(file_path)
if sha256:
completed += 1
else:
failed += 1
except Exception as e:
logger.error(f"Error calculating hash for {file_path}: {e}")
failed += 1
if progress_callback:
try:
await progress_callback(i + 1, total, file_path)
except Exception:
pass
return {"completed": completed, "failed": failed, "total": total}
async def _find_pending_models_from_filesystem(self) -> List[Dict[str, Any]]:
"""Scan filesystem for other-model metadata files with pending hash status."""
pending_models = []
for root_path in self.get_model_roots():
if not os.path.exists(root_path):
continue
for dirpath, dirnames, filenames in os.walk(root_path):
dirnames[:] = [d for d in dirnames if not _is_excluded_dir(d)]
for filename in filenames:
if not filename.endswith(".metadata.json"):
continue
metadata_path = os.path.join(dirpath, filename)
try:
with open(metadata_path, "r", encoding="utf-8") as f:
data = json.load(f)
# Check if hash is pending
hash_status = data.get("hash_status", "completed")
sha256 = data.get("sha256", "")
if hash_status != "completed" or not sha256:
# Find corresponding model file
model_name = filename.replace(".metadata.json", "")
model_path = None
# Look for model file with matching name
for ext in self.file_extensions:
potential_path = os.path.join(dirpath, model_name + ext)
if os.path.exists(potential_path):
model_path = potential_path
break
if model_path:
pending_models.append(
{
"file_path": model_path.replace(os.sep, "/"),
"hash_status": hash_status,
"sha256": sha256,
**{
k: v
for k, v in data.items()
if k
not in [
"file_path",
"hash_status",
"sha256",
]
},
}
)
except (json.JSONDecodeError, Exception) as e:
logger.debug(
f"Error reading metadata file {metadata_path}: {e}"
)
continue
return pending_models
def _root_sub_type_map(self) -> Dict[str, str]:
"""Return the configured business root -> sub_type map."""
root_map = getattr(config, "other_root_subtypes", None)
return root_map if isinstance(root_map, dict) else {}
def _resolve_sub_type(self, root_path: Optional[str]) -> Optional[str]:
"""Resolve the sub_type for a configured root path."""
if not root_path:
return None
normalized_root = self._normalize_path_value(root_path)
for root, sub_type in self._root_sub_type_map().items():
if self._normalize_path_value(root) == normalized_root:
return sub_type
return None
def resolve_sub_type_for_path(self, file_path: Optional[str]) -> Optional[str]:
"""Resolve sub_type from the configured root that contains the file.
Uses the longest-prefix match so nested roots (e.g. a controlnet root
inside a vae root) resolve to the most specific category.
"""
normalized_path = self._normalize_path_value(file_path)
if not normalized_path:
return None
best_length = 0
best_sub_type: Optional[str] = None
for root, sub_type in self._root_sub_type_map().items():
normalized_root = self._normalize_path_value(root)
if not normalized_root:
continue
if (
normalized_path == normalized_root
or normalized_path.startswith(f"{normalized_root}/")
) and len(normalized_root) > best_length:
best_length = len(normalized_root)
best_sub_type = sub_type
return best_sub_type
def adjust_metadata(self, metadata, file_path, root_path):
"""Adjust metadata during scanning to set sub_type."""
sub_type = self._resolve_sub_type(root_path) or self.resolve_sub_type_for_path(
file_path
)
if sub_type:
metadata.sub_type = sub_type
return metadata
def adjust_cached_entry(self, entry: Dict[str, Any]) -> Dict[str, Any]:
"""Adjust entries loaded from the persisted cache to ensure sub_type is set.
sub_type is location-derived: it is re-derived on cache load, never
trusted from the persisted snapshot.
"""
sub_type = self.resolve_sub_type_for_path(entry.get("file_path"))
if sub_type:
entry["sub_type"] = sub_type
return entry
def get_model_roots(self) -> List[str]:
"""Get other-model root directories"""
roots: List[str] = []
roots.extend(config.other_roots or [])
# Remove duplicates while preserving order
seen: set[str] = set()
unique_roots: List[str] = []
for root in roots:
if root and root not in seen:
seen.add(root)
unique_roots.append(root)
return unique_roots
+2
View File
@@ -59,6 +59,7 @@ _MODEL_TYPE_PAGE_MAP = {
"lora": "loras",
"checkpoint": "checkpoints",
"embedding": "embeddings",
"other": "other",
}
# Module-level alias so tests can spy on timer task creation without patching
@@ -983,6 +984,7 @@ class PendingDeleteService:
"get_lora_scanner",
"get_checkpoint_scanner",
"get_embedding_scanner",
"get_other_scanner",
):
getter = getattr(ServiceRegistry, getter_name, None)
if not callable(getter):
+26 -5
View File
@@ -297,23 +297,44 @@ class ServiceRegistry:
async def get_embedding_scanner(cls):
"""Get or create Embedding scanner instance"""
service_name = "embedding_scanner"
if service_name in cls._services:
return cls._services[service_name]
async with cls._get_lock(service_name):
# Double-check after acquiring lock
if service_name in cls._services:
return cls._services[service_name]
# Import here to avoid circular imports
from .embedding_scanner import EmbeddingScanner
scanner = await EmbeddingScanner.get_instance()
cls._services[service_name] = scanner
logger.debug(f"Created and registered {service_name}")
return scanner
@classmethod
async def get_other_scanner(cls):
"""Get or create Other-model scanner instance"""
service_name = "other_scanner"
if service_name in cls._services:
return cls._services[service_name]
async with cls._get_lock(service_name):
# Double-check after acquiring lock
if service_name in cls._services:
return cls._services[service_name]
# Import here to avoid circular imports
from .other_scanner import OtherScanner
scanner = await OtherScanner.get_instance()
cls._services[service_name] = scanner
logger.debug(f"Created and registered {service_name}")
return scanner
@classmethod
def clear_services(cls):
"""Clear all registered services - mainly for testing"""