seaislee1209andClaude Opus 4.8
c8f3f91d38
后端: - 独立生图三类归类:model+product→model_tryon(模特上身图,引用模特库,不送审) / cover→platform_kit(平台套图) / image→free_create(自由创作);生成演员(model 无 product)仍→person(视频角色,留期3) - 成组:每次提交一个 batch_id 串起整批,落 asset.metadata;另记 mode + model_entity_id(上身图溯源模特库) - generate-image 端点透传 model_entity_id;资产库 _tab_q+summary 加 tryon/kits/creations 三类桶 - 单测 StandaloneCategoryTests 直跑 worker 验四态归类全绿(绕开异步 .delay) 前端: - 资产库 library.tsx 加三类 tab(模特上身图/平台套图/自由创作)+ 计数 + 筛选维度 - ai-tools 模特选择器数据源 person→模特库(listModels 映射,选中传形象图当参考 = 引用模特库),ActorLibrary 同源 - api.ts submitGenerateImage 加 model_entity_id 验收:tsc+build 全绿;无头 0 console error(资产库三类 tab 各归类正确、模特选择器 5 张取自模特库) 基线既有 3 失败(StandaloneImageReferenceTests 异步漂移)零新增 Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
279 lines
13 KiB
Python
279 lines
13 KiB
Python
from pathlib import Path
|
||
import uuid
|
||
|
||
import requests
|
||
from django.db import transaction
|
||
from django.db.models import Q
|
||
from django.http import StreamingHttpResponse
|
||
from rest_framework import status
|
||
from rest_framework.decorators import action
|
||
from rest_framework.exceptions import PermissionDenied, ValidationError
|
||
from rest_framework.pagination import PageNumberPagination
|
||
from rest_framework.parsers import FormParser, JSONParser, MultiPartParser
|
||
from rest_framework.response import Response
|
||
from rest_framework.views import APIView
|
||
from rest_framework.viewsets import ModelViewSet
|
||
|
||
from apps.common.api import TeamScopedViewSetMixin, get_current_team
|
||
|
||
from .models import Asset, AssetFile, Model
|
||
from .serializers import AssetSerializer, AssetUploadSerializer, ModelLibrarySerializer
|
||
from .storage import TosStorage
|
||
|
||
|
||
class AssetPagination(PageNumberPagination):
|
||
"""允许前端用 ?page_size= 覆盖(默认 20,上限 200)。前端「取全部资产」一次大页拉完,
|
||
不再逐页串行(原来 20/页 → 资产多时每次刷新要串十几个请求,整页很慢)。"""
|
||
|
||
page_size_query_param = "page_size"
|
||
max_page_size = 200
|
||
|
||
|
||
# 资产库 tab → 查询条件。与前端 assetTab(library.tsx)完全一致,供服务端按页过滤/计数。
|
||
# 图片趴三类(期2):tryon=模特上身图 / kits=平台套图 / creations=自由创作。
|
||
_KNOWN_CATS = [
|
||
"person", "scene", "product_image", "final_video", "upload",
|
||
"model_tryon", "platform_kit", "free_create",
|
||
]
|
||
|
||
|
||
def _tab_q(tab: str) -> Q:
|
||
if tab == "people":
|
||
return Q(category="person")
|
||
if tab == "scenes":
|
||
return Q(category="scene")
|
||
if tab == "products":
|
||
return Q(category="product_image")
|
||
if tab == "tryon": # 模特上身图(图片趴)
|
||
return Q(category="model_tryon")
|
||
if tab == "kits": # 平台套图(图片趴)
|
||
return Q(category="platform_kit")
|
||
if tab == "creations": # 自由创作(图片趴)
|
||
return Q(category="free_create")
|
||
if tab == "uploads":
|
||
return Q(category="upload")
|
||
if tab == "finals": # final_video,或「未归类但是视频」
|
||
return Q(category="final_video") | (~Q(category__in=_KNOWN_CATS) & Q(asset_type="video"))
|
||
if tab == "unclassified": # 未归类且非视频
|
||
return ~Q(category__in=_KNOWN_CATS) & ~Q(asset_type="video")
|
||
return Q()
|
||
|
||
|
||
class AssetViewSet(TeamScopedViewSetMixin, ModelViewSet):
|
||
# select_related 回溯链:asset → origin_task → project,供序列化器解析资产归属商品(避免 N+1)。
|
||
# ★ defer 掉 AITask 的两个巨型 JSON 列(request/response payload):列表序列化只需 project.product_id,
|
||
# 不 defer 的话 select_related 会把每个资产关联 AITask 的完整 AI 请求/响应 payload 全拖出来,
|
||
# 200 个资产实测 100s+(数据越多越慢);defer 后 <1s、product 仍正常解析、无额外查询。
|
||
queryset = (
|
||
Asset.objects.prefetch_related("files")
|
||
.select_related("origin_task__project")
|
||
.defer("origin_task__request_payload", "origin_task__response_payload")
|
||
.all()
|
||
)
|
||
serializer_class = AssetSerializer
|
||
pagination_class = AssetPagination
|
||
search_fields = ["name", "description"]
|
||
ordering_fields = ["created_at", "updated_at", "name"]
|
||
|
||
def get_queryset(self):
|
||
"""服务端过滤,支持各页按需懒加载(不再前端取全量后客户端切片)。
|
||
参数:tab(资产库分桶)/category/asset_type/source/product/q(搜索)/m_<key>(metadata 过滤)/ordering。
|
||
默认按 -created_at 排序——无 ORDER BY 时分页会重/漏。"""
|
||
qs = super().get_queryset()
|
||
p = self.request.query_params
|
||
if p.get("tab"):
|
||
qs = qs.filter(_tab_q(p["tab"]))
|
||
if p.get("category"):
|
||
qs = qs.filter(category=p["category"])
|
||
if p.get("asset_type"):
|
||
qs = qs.filter(asset_type=p["asset_type"])
|
||
if p.get("source"):
|
||
qs = qs.filter(source=p["source"])
|
||
if p.get("product"):
|
||
# 资产归属商品:独立生图写 metadata.product_id;项目内生成回溯 origin_task→project→product;
|
||
# 上传的商品图经 ProductImage 关联。三路并集 = 前端 ProductDetail 的 belongsToProduct。
|
||
pid = p["product"]
|
||
qs = qs.filter(
|
||
Q(metadata__product_id=pid)
|
||
| Q(origin_task__project__product_id=pid)
|
||
| Q(product_images__product_id=pid)
|
||
).distinct()
|
||
if p.get("q"):
|
||
qs = qs.filter(Q(name__icontains=p["q"]) | Q(description__icontains=p["q"]))
|
||
for key, val in p.items():
|
||
if key.startswith("m_") and val:
|
||
qs = qs.filter(**{f"metadata__{key[2:]}": val})
|
||
ordering = p.get("ordering") or "-created_at"
|
||
return qs.order_by(ordering)
|
||
|
||
@action(detail=False, methods=["get"])
|
||
def summary(self, request):
|
||
"""资产库 tab 计数(人物/场景/商品图/成片/我的上传/未分类),供 tab 徽标——不必取全量。"""
|
||
base = Asset.objects.filter(team=self.get_team())
|
||
tabs = ["people", "scenes", "products", "tryon", "kits", "creations", "finals", "uploads", "unclassified"]
|
||
return Response({t: base.filter(_tab_q(t)).count() for t in tabs})
|
||
|
||
@action(detail=False, methods=["get"])
|
||
def facets(self, request):
|
||
"""某 tab 下真实存在的筛选项(来源/类型/指定 metadata 键的取值),供下拉「只列真有的」。
|
||
参数:tab、meta_keys=gender,age,role,...(逗号分隔)。"""
|
||
base = Asset.objects.filter(team=self.get_team())
|
||
if request.query_params.get("tab"):
|
||
base = base.filter(_tab_q(request.query_params["tab"]))
|
||
sources = sorted(s for s in base.values_list("source", flat=True).distinct() if s)
|
||
kinds = sorted(k for k in base.values_list("asset_type", flat=True).distinct() if k)
|
||
meta = {}
|
||
for key in (k for k in request.query_params.get("meta_keys", "").split(",") if k):
|
||
vals = base.values_list(f"metadata__{key}", flat=True)
|
||
meta[key] = sorted({str(v) for v in vals if v not in (None, "")})
|
||
return Response({"sources": sources, "kinds": kinds, "metadata": meta})
|
||
|
||
@action(detail=True, methods=["get"], url_path="raw")
|
||
def raw(self, request, pk=None):
|
||
"""同源流式代理资产主文件。TOS 桶未配 CORS,浏览器 JS 读不到跨域媒体数据——
|
||
编辑器抽视频缩略图(canvas 会被 taint)/解码音频波形(fetch 被拦)都需要走这里。"""
|
||
asset = self.get_object()
|
||
primary = asset.files.filter(is_primary=True).first() or asset.files.first()
|
||
if primary is None:
|
||
return Response({"detail": "asset has no file"}, status=status.HTTP_404_NOT_FOUND)
|
||
url = TosStorage().presigned_get_url(object_key=primary.object_key, expires_in=600)
|
||
upstream_headers = {}
|
||
if request.META.get("HTTP_RANGE"):
|
||
upstream_headers["Range"] = request.META["HTTP_RANGE"]
|
||
upstream = requests.get(url, headers=upstream_headers, stream=True, timeout=120)
|
||
response = StreamingHttpResponse(
|
||
upstream.iter_content(chunk_size=256 * 1024),
|
||
status=upstream.status_code,
|
||
content_type=primary.content_type or "application/octet-stream",
|
||
)
|
||
for header in ("Content-Length", "Content-Range", "Accept-Ranges"):
|
||
if header in upstream.headers:
|
||
response[header] = upstream.headers[header]
|
||
return response
|
||
|
||
|
||
class ModelLibraryViewSet(ModelViewSet):
|
||
"""模特库(顶级实体)· 团队级 + 官方模板跨团队可见。
|
||
list 返回「本团队模特 ∪ 官方模板」;?tab=official 只看官方、?tab=mine 只看自建。
|
||
真人上传走 upload action(传图 → 建 model_portrait 资产 + Model 实体)。删除 = 软删(官方不可删)。"""
|
||
|
||
serializer_class = ModelLibrarySerializer
|
||
pagination_class = AssetPagination
|
||
parser_classes = [JSONParser, MultiPartParser, FormParser]
|
||
search_fields = ["name", "description"]
|
||
|
||
def get_team(self):
|
||
return get_current_team(self.request.user)
|
||
|
||
def get_queryset(self):
|
||
team = self.get_team()
|
||
qs = (
|
||
Model.objects.filter(Q(team=team) | Q(is_official=True), is_deleted=False)
|
||
.prefetch_related("portrait_asset__files", "triview_asset__files")
|
||
)
|
||
tab = self.request.query_params.get("tab")
|
||
if tab == "official":
|
||
qs = qs.filter(is_official=True)
|
||
elif tab == "mine":
|
||
qs = qs.filter(team=team, is_official=False)
|
||
if self.request.query_params.get("q"):
|
||
q = self.request.query_params["q"]
|
||
qs = qs.filter(Q(name__icontains=q) | Q(description__icontains=q))
|
||
# 官方模板靠前,再按新→旧
|
||
return qs.order_by("-is_official", "-created_at")
|
||
|
||
def perform_create(self, serializer):
|
||
serializer.save(team=self.get_team(), created_by=self.request.user)
|
||
|
||
def perform_destroy(self, instance):
|
||
if instance.is_official:
|
||
raise PermissionDenied("官方模特不可删除")
|
||
if instance.team_id != self.get_team().id:
|
||
raise PermissionDenied("无权删除其他团队的模特")
|
||
instance.is_deleted = True
|
||
instance.save(update_fields=["is_deleted", "updated_at"])
|
||
|
||
@action(detail=False, methods=["post"], url_path="upload", parser_classes=[MultiPartParser, FormParser])
|
||
def upload(self, request):
|
||
"""真人上传:一张人像图 → 建 model_portrait 资产 + Model(source=upload)。三视图/声线后续补。"""
|
||
upload = request.FILES.get("file")
|
||
if upload is None:
|
||
raise ValidationError({"file": "请上传一张模特形象图"})
|
||
team = self.get_team()
|
||
name = (request.data.get("name") or Path(upload.name).stem or "模特")[:255]
|
||
asset_id = uuid.uuid4()
|
||
suffix = Path(upload.name).suffix.lower() or ".png"
|
||
object_key = f"teams/{team.id}/models/{asset_id}{suffix}"
|
||
with transaction.atomic():
|
||
stored = TosStorage().upload_fileobj(
|
||
fileobj=upload.file,
|
||
object_key=object_key,
|
||
content_type=upload.content_type or "image/png",
|
||
)
|
||
portrait = Asset.objects.create(
|
||
id=asset_id,
|
||
team=team,
|
||
created_by=request.user,
|
||
name=f"{name}·形象图",
|
||
asset_type=Asset.Type.IMAGE,
|
||
source=Asset.Source.UPLOAD,
|
||
category=Asset.Category.MODEL_PORTRAIT,
|
||
metadata={"kind": "model", "view": "frontal", "source": "upload"},
|
||
)
|
||
AssetFile.objects.create(
|
||
asset=portrait,
|
||
object_key=stored.object_key,
|
||
bucket=stored.bucket,
|
||
content_type=stored.content_type,
|
||
size_bytes=stored.size_bytes,
|
||
is_primary=True,
|
||
)
|
||
model = Model.objects.create(
|
||
team=team,
|
||
created_by=request.user,
|
||
name=name,
|
||
source=Model.Source.UPLOAD,
|
||
portrait_asset=portrait,
|
||
metadata={"source": "upload"},
|
||
)
|
||
return Response(ModelLibrarySerializer(model).data, status=status.HTTP_201_CREATED)
|
||
|
||
|
||
class AssetUploadView(APIView):
|
||
parser_classes = [MultiPartParser, FormParser]
|
||
|
||
@transaction.atomic
|
||
def post(self, request):
|
||
serializer = AssetUploadSerializer(data=request.data)
|
||
serializer.is_valid(raise_exception=True)
|
||
team = get_current_team(request.user)
|
||
upload = serializer.validated_data["file"]
|
||
suffix = Path(upload.name).suffix.lower()
|
||
asset_id = uuid.uuid4()
|
||
object_key = f"teams/{team.id}/uploads/{asset_id}{suffix}"
|
||
|
||
stored = TosStorage().upload_fileobj(
|
||
fileobj=upload.file,
|
||
object_key=object_key,
|
||
content_type=upload.content_type or "application/octet-stream",
|
||
)
|
||
asset = Asset.objects.create(
|
||
id=asset_id,
|
||
team=team,
|
||
created_by=request.user,
|
||
name=serializer.validated_data.get("name") or upload.name,
|
||
asset_type=serializer.validated_data["asset_type"],
|
||
source=Asset.Source.UPLOAD,
|
||
category=serializer.validated_data["category"],
|
||
description=serializer.validated_data.get("description", ""),
|
||
)
|
||
AssetFile.objects.create(
|
||
asset=asset,
|
||
object_key=stored.object_key,
|
||
bucket=stored.bucket,
|
||
content_type=stored.content_type,
|
||
size_bytes=stored.size_bytes,
|
||
is_primary=True,
|
||
)
|
||
return Response(AssetSerializer(asset).data, status=status.HTTP_201_CREATED)
|