feat(core): AI 生成 Agent 化 — 多模型流式脚本 agent + 可插拔 Provider + gpt-image-2 参考图 + 模特库
- 后端·可插拔 Provider 层:通用 OpenAICompatibleProvider(tokenssr 等中转站,base_url+api_key,零改代码换站)+ ModelProvider.api_key - 后端·脚本 agent:结构化 ScriptDraft 契约 + 加载电商 skill + 出稿/改稿一体对话 agent(3模式/多模型)+ 流式 SSE 端点(DRF SSE renderer) - 后端·图像:gpt-image-2 参考图出图 + 故事板 @图1@图2@图3 多锚点合成(锁脸锁商品);Seedance 打开 generate_audio - 后端·模特库:gpt-image-2 生成器(9:16氛围图→16:9白底三视图)+ seed_demo_models 管理命令 - DB·迁移:tokenssr 中转站 + 多模型 seed(豆包/GPT-5.5/Gemini + gpt-image-2);ScriptSegment 结构化字段 - 前端·脚本趴:接真 SSE(工具卡 + 思考流)+ 模型下拉 + 3模式 + 改稿;agentScriptStream - skills/ecommerce-video-script 电商脚本技能(运行时依赖) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
seaislee1209
co-authored by
Claude Opus 4.8
Claude Opus 4.8
parent
a0ffb6fc8e
commit
6464001f84
@@ -0,0 +1,33 @@
|
||||
# Generated by Django 5.1.15 on 2026-06-16 17:58
|
||||
|
||||
from django.db import migrations, models
|
||||
|
||||
|
||||
class Migration(migrations.Migration):
|
||||
|
||||
dependencies = [
|
||||
('projects', '0001_initial'),
|
||||
]
|
||||
|
||||
operations = [
|
||||
migrations.AddField(
|
||||
model_name='scriptsegment',
|
||||
name='entity_refs',
|
||||
field=models.JSONField(blank=True, default=list),
|
||||
),
|
||||
migrations.AddField(
|
||||
model_name='scriptsegment',
|
||||
name='product_exposure',
|
||||
field=models.CharField(blank=True, max_length=64),
|
||||
),
|
||||
migrations.AddField(
|
||||
model_name='scriptsegment',
|
||||
name='role',
|
||||
field=models.CharField(blank=True, max_length=16),
|
||||
),
|
||||
migrations.AddField(
|
||||
model_name='scriptsegment',
|
||||
name='speaker',
|
||||
field=models.CharField(blank=True, max_length=32),
|
||||
),
|
||||
]
|
||||
@@ -80,6 +80,11 @@ class ScriptSegment(TimeStampedModel):
|
||||
narration = models.TextField(blank=True)
|
||||
visual_prompt = models.TextField(blank=True)
|
||||
product_points = models.JSONField(default=list, blank=True)
|
||||
# ScriptDraft 结构化契约字段(对话式脚本 agent 产出):
|
||||
role = models.CharField(max_length=16, blank=True) # 钩子|痛点|卖点|CTA
|
||||
speaker = models.CharField(max_length=32, blank=True) # 指向某 entity id;画外旁白为空
|
||||
product_exposure = models.CharField(max_length=64, blank=True) # 手持/特写/使用中…
|
||||
entity_refs = models.JSONField(default=list, blank=True) # 本镜引用的 entity id 列表(→ 故事板 @图N)
|
||||
|
||||
class Meta:
|
||||
ordering = ["sort_order", "created_at"]
|
||||
|
||||
@@ -228,7 +228,10 @@ class ExportJobSerializer(serializers.ModelSerializer):
|
||||
class ScriptSegmentSerializer(serializers.ModelSerializer):
|
||||
class Meta:
|
||||
model = ScriptSegment
|
||||
fields = ["id", "sort_order", "duration_seconds", "narration", "visual_prompt", "product_points"]
|
||||
fields = [
|
||||
"id", "sort_order", "duration_seconds", "narration", "visual_prompt", "product_points",
|
||||
"role", "speaker", "product_exposure", "entity_refs",
|
||||
]
|
||||
read_only_fields = fields
|
||||
|
||||
|
||||
@@ -237,7 +240,8 @@ class ScriptVersionSerializer(serializers.ModelSerializer):
|
||||
|
||||
class Meta:
|
||||
model = ScriptVersion
|
||||
fields = ["id", "title", "content", "source", "is_adopted", "segments", "created_at", "updated_at"]
|
||||
# metadata 携带 ScriptDraft 的 hook/tone/entities,供前端结构化渲染与下游故事板 @图N
|
||||
fields = ["id", "title", "content", "source", "is_adopted", "segments", "metadata", "created_at", "updated_at"]
|
||||
read_only_fields = fields
|
||||
|
||||
|
||||
|
||||
@@ -3,13 +3,17 @@ from pathlib import Path
|
||||
import uuid
|
||||
|
||||
from django.db import transaction
|
||||
from django.http import JsonResponse, StreamingHttpResponse
|
||||
from rest_framework import status
|
||||
from rest_framework.decorators import action
|
||||
from rest_framework.parsers import FormParser, MultiPartParser
|
||||
from rest_framework.renderers import BaseRenderer
|
||||
from rest_framework.response import Response
|
||||
from rest_framework.viewsets import ModelViewSet
|
||||
|
||||
from apps.ai.models import ModelConfig
|
||||
from apps.ai.providers import TtsNotConfigured
|
||||
from apps.ai.script_agent import stream_script_agent
|
||||
from apps.ai.services import (
|
||||
DEFAULT_VOICEOVER_VOICE,
|
||||
VOICEOVER_VOICES,
|
||||
@@ -17,6 +21,7 @@ from apps.ai.services import (
|
||||
generate_base_asset,
|
||||
generate_project_script,
|
||||
generate_storyboard_frame,
|
||||
get_default_model,
|
||||
poll_video_segment,
|
||||
regenerate_script_segment,
|
||||
submit_storyboard,
|
||||
@@ -58,6 +63,18 @@ from .tasks import poll_video_segment_task
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
class ServerSentEventRenderer(BaseRenderer):
|
||||
"""让 DRF 内容协商接受 Accept: text/event-stream(否则流式端点直接 406)。
|
||||
实际响应由视图返回 StreamingHttpResponse 直接下发,这个 renderer 只用于通过协商。"""
|
||||
|
||||
media_type = "text/event-stream"
|
||||
format = "event-stream"
|
||||
charset = None
|
||||
|
||||
def render(self, data, accepted_media_type=None, renderer_context=None):
|
||||
return data
|
||||
|
||||
|
||||
def _store_uploaded_asset(*, team, user, upload, asset_type: str, category: str, name: str) -> Asset:
|
||||
"""把上传的文件落到 TOS,建 Asset+AssetFile(主文件)。供上传视频段 / 上传 BGM 复用。"""
|
||||
suffix = Path(upload.name).suffix.lower() or (".mp4" if asset_type == Asset.Type.VIDEO else ".mp3")
|
||||
@@ -133,6 +150,53 @@ class ProjectViewSet(TeamScopedViewSetMixin, ModelViewSet):
|
||||
)
|
||||
return Response(ScriptVersionSerializer(script).data, status=status.HTTP_201_CREATED)
|
||||
|
||||
@action(detail=True, methods=["post"], url_path="script-agent-stream", renderer_classes=[ServerSentEventRenderer])
|
||||
def script_agent_stream(self, request, pk=None):
|
||||
"""对话式脚本 agent · 流式(SSE)。出稿 + 改稿一体,多模型可选。
|
||||
请求体:mode(auto|theme|revise)、prompt、model_config_id、selling_point_ids、
|
||||
base_version_id(改稿)、aspect_ratio、total_duration。
|
||||
响应:text/event-stream,逐帧吐 tool/delta/draft/saved/done/error。"""
|
||||
project = self.get_object()
|
||||
mode = str(request.data.get("mode") or "auto")
|
||||
prompt = str(request.data.get("prompt") or "")
|
||||
selling_point_ids = request.data.get("selling_point_ids") or []
|
||||
base_version_id = request.data.get("base_version_id") or None
|
||||
aspect_ratio = str(request.data.get("aspect_ratio") or "9:16")
|
||||
try:
|
||||
total_duration = int(request.data.get("total_duration") or 60)
|
||||
except (TypeError, ValueError):
|
||||
total_duration = 60
|
||||
|
||||
model_config = None
|
||||
requested = request.data.get("model_config_id")
|
||||
if requested:
|
||||
model_config = (
|
||||
ModelConfig.objects.select_related("provider")
|
||||
.filter(id=requested, capability=ModelConfig.Capability.TEXT, status=ModelConfig.Status.ACTIVE)
|
||||
.first()
|
||||
)
|
||||
if model_config is None:
|
||||
model_config = get_default_model(ModelConfig.Capability.TEXT)
|
||||
if model_config is None:
|
||||
# 纯 Django 响应:绕开 DRF 渲染(此 action 只挂了 SSE renderer)
|
||||
return JsonResponse({"detail": "没有可用的文本模型,请先在模型库配置"}, status=400)
|
||||
|
||||
stream = stream_script_agent(
|
||||
project=project,
|
||||
user=request.user,
|
||||
model_config=model_config,
|
||||
mode=mode,
|
||||
user_prompt=prompt,
|
||||
selling_point_ids=selling_point_ids,
|
||||
base_version_id=base_version_id,
|
||||
aspect_ratio=aspect_ratio,
|
||||
total_duration=total_duration,
|
||||
)
|
||||
response = StreamingHttpResponse(stream, content_type="text/event-stream")
|
||||
response["Cache-Control"] = "no-cache"
|
||||
response["X-Accel-Buffering"] = "no" # 关 nginx 缓冲,保证逐帧下发
|
||||
return response
|
||||
|
||||
@action(detail=True, methods=["post"], url_path="adopt-script")
|
||||
@transaction.atomic
|
||||
def adopt_script(self, request, pk=None):
|
||||
|
||||
Reference in New Issue
Block a user