Trigix
bj-qizhi/trigix
Engine de workflows para IA com 140+ nós e orquestração visual
Ver no GitHubPlataforma de automação de workflows nativa para IA, construída em Rust com engine de execução DAG, 140+ tipos de nós e editor visual em tela. Alternativa self-hosted ao n8n focada em automação com IA, integrando modelos LLM, APIs, bancos de dados e ferramentas SaaS.
Instalação
npx convez-ai bj-qizhi/trigix1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Automação de processos empresariais envolvendo múltiplos modelos de IA
- Integração de sistemas legados com agentes IA modernos
- Pipelines de RAG e processamento de dados com LLMs
- Orquestração de workflows complexos com paralelização
Veredito

Trigix é uma solução madura e robusta para automação enterprise com IA, oferecendo controle total via self-hosting, performance comprovada em Rust e flexibilidade para workflows complexos. Ideal para equipes que precisam orquestrar múltiplos modelos LLM e integrações SaaS em ambiente corporativo.
README
A tradução para o português deste README ainda está na fila.
Trigix
AI-Native Workflow Automation Platform
Open-source, AI-native workflow automation — a Rust execution engine, 140+ node types, RAG, agents, MCP, and a visual canvas. A self-hosted n8n alternative built for AI-first automation.
© 2026 北京祺智科技有限公司 · managecode@gmail.com
What is Trigix?
Trigix is an enterprise-grade, AI-native workflow automation platform. Build, run, and monitor complex workflows visually — connecting AI models, APIs, databases, and SaaS tools with a drag-and-drop canvas editor.
Key differentiators:
- 140 node types — AI models (Claude, GPT-4, Gemini, Groq, Mistral…), SaaS integrations (Slack, Jira, Notion, Salesforce…), data transforms, control flow — plus your own via the node SDK
- Rust execution engine — DAG scheduling, parallel fan-out, retries, timeouts, cancellation
- AI-native — built-in LLM nodes, RAG over pgvector, an agent tool-use loop, MCP protocol support
- Enterprise-ready — SSO (OIDC), JWT + RBAC, multi-tenant, encrypted secrets, audit log, webhook signatures, Kubernetes Helm chart
Features
| Category | Highlights |
|---|---|
| Canvas Editor | Drag-and-drop, React Flow, minimap, snap-to-grid, undo/redo, keyboard shortcuts |
| Execution Engine | Async DAG, parallel branches (Fan-out/Fan-in), sub-workflows, ForEach, loop |
| AI Nodes | Claude, OpenAI, Gemini, Groq, Mistral, Cohere, Replicate, Perplexity + 7 Chinese LLMs |
| Integrations | 100+ nodes: GitHub, Jira, Notion, Slack, Stripe, Salesforce, Airtable, Linear… |
| Transforms | Filter, Map, Aggregate, Sort, Merge, Extract, Dedupe, Regex, CSV, XML, YAML… |
| Triggers | Webhook (HMAC-SHA256), Cron, Interval, Manual, Form submit |
| Auth & Security | JWT, RBAC (Viewer/Editor/Admin), API Keys, bcrypt passwords, email verification |
| Observability | Audit log, execution timeline, Prometheus metrics, OpenTelemetry tracing |
| Infrastructure | PostgreSQL, Redis Streams, Docker, Kubernetes Helm Chart |
Quick Start
# 1. Start local infrastructure
docker compose up -d
# 2. Run the platform backend
DATABASE_URL=postgres://trigix:trigix@localhost:35432/trigix \
PLATFORM_HTTP_ADDR=127.0.0.1:38080 \
cargo run -p trigix-platform
# 3. Run the execution engine
EXECUTOR_HTTP_ADDR=127.0.0.1:38090 \
cargo run -p trigix-executor
# 4. Run the web console
cd apps/web && npm install && npm run dev
# Open http://localhost:3100
Default dev API key: dev
Deploy
Kubernetes (Helm chart, published to GHCR):
helm install trigix oci://ghcr.io/bj-qizhi/charts/trigix --version 0.3.2 \
--namespace trigix --create-namespace
Deploys the platform, AI runtime, PostgreSQL/pgvector, and Redis. Configure via
charts/trigix/values.yaml. The chart .tgz is also attached to each
chart-v* release.
Docker Compose (single host):
docker compose -f docker-compose.prod.yml up -d --build
Repository Structure
apps/web React web console (Vite + React Flow)
services/platform-rs Rust platform API (Axum, JWT, multi-tenant)
services/executor Rust execution engine (DAG, parallel, retries)
services/ai-runtime Python AI runtime (FastAPI) — RAG (pgvector) + agent tool-use
sdk/python Custom node SDK (Python) — published as trigix-node-sdk
sdk/typescript Custom node SDK (TypeScript/JS) — published as trigix-node-sdk
crates/workflow-core Shared WorkflowGraph model + DAG validation
crates/execution-core Shared ExecutionStatus types
infra/postgres PostgreSQL migrations
charts/trigix Kubernetes Helm chart
docs/ Architecture, ADRs, dev guides
Documentation
| Document | Description |
|---|---|
| Architecture Design | System architecture, service boundaries, domain model |
| Workflow Graph JSON | Node/Edge schema reference |
| Dev Bootstrap Guide | Local setup with PostgreSQL |
| Port Reference | All local service ports |
| ADR-0001: Layered Architecture | Architecture decision record |
| Demo: Feedback Triage | Annotated end-to-end run of a custom-node pipeline |
Extend with custom nodes (Node SDK)
Write your own workflow node as a small HTTP service — in Python or TypeScript/JavaScript — and use it like any built-in node. No changes to the Trigix executor required.
pip install trigix-node-sdk # Python
npm install trigix-node-sdk # TypeScript / JavaScript
from trigix_node_sdk import node, create_app
@node(slug="greet", label="Greeter",
config_schema={"type": "object", "properties": {"name": {"type": "string"}}})
def greet(config, input, node_outputs):
return {"greeting": f"Hello, {config.get('name') or input.get('name', 'world')}!"}
app = create_app() # uvicorn module:app --port 9000
Then in the web console → Custom Nodes, paste your service URL and click
Import All — every node from GET /manifest is registered at once and shows
up in the workflow editor.
Environment Variables
| Variable | Service | Purpose |
|---|---|---|
DATABASE_URL | Platform | PostgreSQL connection (default: in-memory) |
EXECUTOR_BASE_URL | Platform | Standalone executor URL (default: inline) |
AI_RUNTIME_BASE_URL | Executor | Python AI runtime URL |
ANTHROPIC_API_KEY | AI Runtime | Claude API key |
AUTH_REQUIRED | Platform | Enforce JWT on all routes (true/false) |
DEV_API_KEY | Platform | Dev API key (default: dev) |
License
MIT License — see LICENSE
Copyright © 2026 北京祺智科技有限公司 · https://www.qzso.com/