它通过将可视化 DAG 画布与动态 AI 规划相结合，平衡了确定性工作流工具（如 n8n）的僵化和自由循环智能体框架（如 AutoGen）的混乱，强制执行严格的循环安全与权限交集安全，同时支持本地 LLM 和丰富的内置智能体。
Build networks of AI agents that plan dynamically, coordinate via DAG, and never spin out of control.
git clone https://github.com/pauloberezini/hermes-synapse && cd hermes-synapse
cp .env.example .env # add your LLM API key
docker compose up -d
# Open → http://localhost:9119
Most multi-agent frameworks are either too rigid (n8n: hardwired workflows) or too chaotic (AutoGen: agents talking in circles forever).
Hermes sits in the middle: a visual drag-and-drop canvas where you wire up AI agents in a strict Directed Acyclic Graph (DAG). No infinite loops. No hardcoded pipelines. Just agents that actually coordinate.
| Hermes | n8n / Make | Flowise / LangFlow | AutoGen / CrewAI | |
|---|---|---|---|---|
| Execution | 🧠 Non-deterministic (AI plans dynamically) | 🔧 Deterministic (hardcoded steps) | 🔗 Visual LLM chains | 💬 Conversational loops |
| UI | 🎨 SVG canvas + isolated agent chats | Node editor | Visual chain designer | CLI / API only |
| Hierarchy | ✅ Strict DAG (cycle-safe) | ↪ Linear / conditional | Data-flow graphs | ⚠️ Free loops (cycle risk) |
| Security | 🔐 Permission Intersection | Hardcoded auth | Sandbox container | Local exec by default |
| Self-hosted | ✅ Docker, SQLite, local LLMs | ✅ | ✅ | ✅ |
👆 Click to watch — canvas node wiring, Telegram integration, and live DAG planning in 3 minutes.
Requirements: Docker + Docker Compose. That's it.
# 1. Clone
git clone https://github.com/pauloberezini/hermes-synapse
cd hermes-synapse
# 2. Configure (only OPENROUTER_API_KEY is required — everything else is optional)
cp .env.example .env
# 3. Launch
docker compose up -d --build
# 4. Open dashboard
open http://localhost:9119
Want to use a local model (Ollama)? Set in .env:
LLM_API_BASE=http://host.docker.internal:11434/v1
LLM_MODEL=llama3
OPENROUTER_API_KEY=local # any non-empty placeholder
User ──→ Web Dashboard (React Canvas)
│
▼
🤖 Root Orchestrator (Jarvis) ← DAG en
Visual multi-agent orchestrator with a drag-and-drop DAG canvas, dynamic LLM planning loops, permission-based tool execution, and local Qdrant RAG.
根据分类、Topic 和编程语言匹配的相似项目。
Harness Engineering is a methodology for improving coding agent outputs by carefully crafting the environment around them—providing curated context, tools, and executable constraints that encode an organization’s nonfunctional requirements and cumulative lessons.
A 28.9 million parameter language model runs on an $8 ESP32-S3 microcontroller entirely on-device, generating simple stories at about 9.5 tokens per second.
mimic captures traffic from any iOS or web app and automatically generates a Python client library that lets you call the app's API like a regular library.