经 AI Skill Hub 精选评估，星云节点 获评「强烈推荐」。这款Agent工作流在功能完整性、社区活跃度和易用性方面表现出色，AI 评分 8.0 分，适合有一定技术背景的用户使用。
星云节点 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
星云节点 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
# 方式一：pip 安装（推荐）
pip install nebula-nodes
# 方式二：虚拟环境安装（推荐生产环境）
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install nebula-nodes
# 方式三：从源码安装（获取最新功能）
git clone https://github.com/JustinPerea/nebula-nodes
cd nebula-nodes
pip install -e .
# 验证安装
python -c "import nebula_nodes; print('安装成功')"
# 命令行使用
nebula-nodes --help
# 基本用法
nebula-nodes input_file -o output_file
# Python 代码中调用
import nebula_nodes
# 示例
result = nebula_nodes.process("input")
print(result)
# nebula-nodes 配置文件示例（config.yml） app: name: "nebula-nodes" debug: false log_level: "INFO" # 运行时指定配置文件 nebula-nodes --config config.yml # 或通过环境变量配置 export NEBULA_NODES_API_KEY="your-key" export NEBULA_NODES_OUTPUT_DIR="./output"
--- name: Nebula Nodes slug: nebula-nodes status: active tagline: A multi-surface AI creation studio — node graph, Create view, Cinema, Character, and Moodboard — running locally on your own keys. description: AI creation studio built around a visual node graph. 138 built-in nodes across 15 provider families, four universal nodes that reach 300+ more (OpenRouter, Nous Portal, Replicate, FAL), seven specialized workspaces (Canvas, Create, Cinema Studio, Character Studio, Moodboard Studio, Video Editor, Remotion Editor), smart subgraph caching, real-time streaming, and an optional chat agent that builds graphs from natural language. stack: - Python 3.12+ - FastAPI - React 19 - TypeScript - Vite - React Flow (@xyflow/react) - Zustand - WebSockets - Hermes Agent (optional) features: - 138-node BYOK catalog across 15 provider families - Universal nodes for OpenRouter / Nous Portal / Replicate / FAL - Seven workspaces: Canvas, Create, Cinema Studio, Character Studio, Moodboard Studio, Video Editor, Remotion Editor - Smart subgraph caching with topological execution - Real-time streaming (text tokens, video / audio / SVG previews) - Create view with model picker, presets/styles library, results gallery, and per-result actions - Save / load graphs as JSON; configurable output directory; Reveal in Finder - Daedalus / Claude / Codex agent chat (builds and edits the graph from natural language) - 1,320+ tests (980 backend + 347 frontend) hero: docs/assets/banner.svg links: github: https://github.com/JustinPerea/nebula-nodes lab: https://justinperea.com/lab/nebula-quiver visibility: public
SPEC // MULTI-SURFACE STUDIO| **138-NODE CATALOG** | Every major provider as a first-class node. Image, video, 3D, audio, and text generation across 15 provider families — see [CATALOG](#-catalog) for the breakdown. |
| **7 WORKSPACES** | Canvas (node graph), Create, Cinema Studio, Character Studio, Moodboard Studio, Video Editor, Remotion Editor. See [WORKSPACES](#-workspaces). |
| **CREATE VIEW** | Prompt → generate with a model picker, parameter pills, reference image slots, quantity/variations control, a results gallery (Session + Canvas tabs), a Presets/Styles library, and per-result actions: download, open-in-canvas, use-as-input, Reveal in Finder, Save to folder, delete. Generations author real nodes onto the canvas and persist. |
| **UNIVERSAL NODES** | One node each for OpenRouter, Nous Portal, Replicate, and FAL. The universal node pattern means a single node reaches every model on that platform — 300+ models total, no per-model wrapper needed. |
| **SMART EXECUTION** | Topological graph sort, parallel where independent, sequential where dependent. Unchanged subgraphs skip re-computation automatically — if you change a prompt downstream, only the affected nodes re-run. |
| **STREAMING OUTPUTS** | Token-by-token text via WebSocket, live frame previews for video and audio nodes, SSE streaming for Quiver SVG generation. The canvas updates in real time as outputs arrive. |
| **PARTIAL EXECUTION** | Run the full graph, or just the upstream subgraph of any single node. Fast iteration on the bit you care about. |
| **TYPED PORTS** | Every port carries a type (Image, Video, Text, Audio, Mesh, SVG, Array, Any). Wrong wires fail at design time, not runtime. Color-coded for visual rhythm. |
| **UNDO THAT STICKS** | 50-step history. Outputs survive undo, so experiment freely. |
| **SAVE / LOAD** | Graphs serialize to JSON. Outputs written to disk and served via /api/outputs. Configurable output directory. Reveal in Finder. |
| **AGENT CHAT** | Daedalus (Hermes Agent), Claude, or Codex can build and edit graphs from natural language. Sees the live canvas; iterates with you. See [AGENT](#-agent--daedalus). |
| **AUDIT DISCIPLINE** | Every API-backed node is verified against canonical provider docs. 1,320+ tests (980 backend + 347 frontend). Live-smoke gate for high-risk handler paths. See [QUALITY](#-quality--audit-discipline). |
```bash
curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
cp .hermes/profiles/daedalus/SOUL.md ~/.hermes/profiles/daedalus/SOUL.md mkdir -p ~/.hermes/profiles/daedalus/skills/creative cp -R .hermes/skills/daedalus-core ~/.hermes/profiles/daedalus/skills/creative/
SPEC // REQUIRES Python 3.12+, Node.js 18+```bash
node scripts/check-node-contracts.mjs ```
If you are adding a new model, the smallest useful contribution is a single handler in backend/handlers/ plus a node definition in backend/data/node_definitions.json and its mirror in frontend/src/constants/nodeDefinitions.ts. Add a structural audit note under docs/model-providers/<provider>/. Existing nodes are good templates — copy the closest match and adjust.
If you are extending Daedalus, the playbook lives at .hermes/skills/daedalus-core/SKILL.md. The persona contract is .hermes/profiles/daedalus/SOUL.md. Both are copied into the user's Hermes profile during setup.
高质量的开源AI工作流项目，支持多媒体处理
该工具使用 AGPL-3.0 协议，商用场景请仔细阅读协议条款，必要时咨询法律意见。
AI Skill Hub 为第三方内容聚合平台，本页面信息基于公开数据整理，不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后，再部署至生产环境，并做好必要的安全评估。
⚠️ AGPL 3.0 — 最严格的 Copyleft，网络服务端使用也需开源，SaaS 使用受限。
AI Skill Hub 点评：星云节点 的核心功能完整，质量优秀。对于自动化工程师和运维人员来说，这是一个值得纳入个人工具库的选择。建议先在非生产环境试用，再逐步推广。
| 原始名称 | nebula-nodes |
| Topics | ai-agentsbyokcreative-aigenerative-aipython |
| GitHub | https://github.com/JustinPerea/nebula-nodes |
| License | AGPL-3.0 |
| 语言 | Python |
收录时间：2026-06-22 · 更新时间：2026-06-22 · License：AGPL-3.0 · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型，复制安装指令后粘贴到对应客户端