🎬 VerseFlow Studio

Next-Gen Enterprise Autonomous AI-Video Operating System & PaaS Infrastructure

Key Features • Quick Start • Architecture • API & SDK • License

---

📜 版权与作者声明 (Author Declaration & Notice)

• 作者：XiaoZhe (XiaoZhe AIGC工作室)
• 非商业用途：允许免费学习、研究和非商业化使用，但必须保留作者信息（“XiaoZhe创作”）。
• 商业化使用：未经授权不得擅自商业化使用。如需商业化部署、商用授权或项目定制，请联系作者：
  • Email: janejulius119@gmail.com
  • WeChat: julius119
  • QQ: 290187006

---

💡 项目简介

VerseFlow Studio 是一套企业级 AI 视频操作系统。对比传统生成工具，在底层计算内核、状态管理、导演编排及工业化流转上完成了全面跃迁：

---

🌟 Key Architectural Breakthroughs

1. 🧠 vLLM-Inspired PagedVideoLatent Memory Pool

• Implements page-based allocation for 3D spatial-temporal Latent Tensors.
• Completely eliminates memory fragmentation during long-sequence video generation.
• Enables continuous batching for high-concurrency enterprise video requests.

2. ⚡ ComfyUI-Style Incremental Latent DAG Caching

• Hashes sub-graph nodes based on parameters and upstream dependencies.
• Re-renders only modified shots while reusing cached Latents for unchanged shots.
• Slashes iteration and trial costs by up to 90%.

3. ⏸️ LangGraph-Driven Stateful HITL & Time-Travel Breakpoints

• Introduces Human-In-The-Loop (HITL) execution pause points at keyframes.
• Exposes mask restoration and canvas repainting controls to artists.
• Resumes execution seamlessly from saved state snapshots without restarting the DAG.

4. 🎯 Agentic Director & RLHF/DPO Self-Evolution Loop

• Automatically parses text scripts into camera movements, shot sizes, and lighting parameters.
• Ingests real-world video analytics (CTR, 5s retention, completion rate) to construct Direct Preference Optimization (DPO) datasets.
• Continuously fine-tunes director prompt strategies over time.

5. 🌐 Edge Heterogeneous Swarm & SaaS Infrastructure

• Features multi-tenant RBAC isolation and precise GPU-Second billing.
• Supports heterogeneous Edge rendering nodes (Vast.ai, RunPod, RTX 4090/Spot instances).
• Auto-handles SIGTERM preemptions with zero-data-loss checkpoint recovery.

---

🚀 快速启动

# 1. 克隆仓库
git clone [https://github.com/julius119/verseflow-studio.git](https://github.com/julius119/verseflow-studio.git)
cd verseflow-studio

# 2. Docker 容器化启动 (推荐)
docker compose -f docker/docker-compose.yml up -d

# 3. 运行端到端测试
python3 scripts/run_pipeline.py

# 4. 运行硬件性能基准测试
python3 scripts/benchmark.py