Yuandu AI Backend

Industrial AI Visual Asset Production Platform · Backend Control Plane

Live Website: https://www.emoji.icu

1) 定位（Positioning）

元都AI（Yuandu AI）是工业级 AI 视觉资产生产平台。
本仓库是平台后端中枢，负责 API、任务编排、Worker 调度、质量复审、资产交付与可观测。

• 业务定位：从视频自动生产高价值视觉资产
• 系统定位：高并发调度中枢 + 多格式生产流水线
• 支持格式：GIF / PNG / JPG / WebP / MP4 / Live

---

2) 架构总览图（Architecture Overview）

---

3) 平台架构图（Platform Architecture）

flowchart LR
    U[Frontweb 用户端] --> G[Nginx / API Gateway]
    A[Lookfront 运营中台] --> G

    G --> API[Backend API & Auth]

    API --> ORCH[Workflow Orchestration]
    ORCH <--> R[(Redis / Asynq)]
    API <--> DB[(PostgreSQL)]

    ORCH --> W1[GIF Worker]
    ORCH --> W2[Image Worker<br/>PNG/JPG/WebP]
    ORCH --> W3[MP4/Live Worker]

    W1 --> QN[(Qiniu Object Storage)]
    W2 --> QN
    W3 --> QN

    W1 --> QA[Evaluation / Review]
    W2 --> QA
    W3 --> QA
    QA --> DB

Loading

---

4) 模块关系图（Backend Modules）

flowchart TD
    Router[Router] --> Handler[HTTP Handlers]
    Handler --> Service[Domain Services]
    Service --> Repo[Storage Layer]

    Service --> Job[Video Job Orchestrator]
    Job --> AI1[AI1 Briefing]
    Job --> AI2[AI2 Planning/Scoring]
    Job --> Worker[Format Workers]
    Job --> AI3[AI3 Review]

    Repo --> PG[(PostgreSQL)]
    Job --> Redis[(Redis/Asynq)]
    Worker --> OSS[(Qiniu)]

Loading

---

5) 路线图目录（Roadmap）

阶段	目标	状态
Phase 1	视频→视觉资产主链路稳定化（多格式、可观测、可回溯）	✅ In Progress
Phase 2	引入 GPU 微服务（Real-ESRGAN / SeedVR2）提升重建能力	🚧 Planned
Phase 3	团队化与平台化能力（API/SaaS/协作）	🗓️ Planned

---

6) Tech Stack

• Go 1.25+
• Gin + GORM
• PostgreSQL
• Redis + Asynq
• ffmpeg / ffprobe
• Qiniu Object Storage

---

7) Quick Start

cp .env.example .env
go run ./cmd/api
# new terminal
go run ./cmd/worker

• API default: :5050
• Health: GET /healthz

---

8) Database Migration

for f in migrations/*.sql; do
  psql "$DATABASE_URL" -v ON_ERROR_STOP=1 -f "$f"
done

---

9) Deployment

See: docs/DEPLOYMENT.md

---

10) Open-source Safety

• Do not commit real secrets (.env is ignored)
• Do not commit private model weights / private prompts / private datasets

---

11) License

See LICENSE.