im-andr.es
Full‑StackAI/MLData & AnalyticsDevOpsProduct & UX

AgentFlow

Instant workflow creation via chat

Home-grown Langflow alternative with natural language workflow generation and n8n automation integration.

ReactNode.jsn8n APIOpenAIFlow Builder

Problem

Creating complex automation workflows requires technical expertise and time-consuming manual configuration.

Solution

Built an intuitive visual flow builder with AI-powered natural language processing that generates n8n workflows instantly from chat descriptions.

Results

  • Generate complete automation workflows in seconds
  • Natural language interface for non-technical users
  • Seamless n8n integration for enterprise automation
  • Visual flow builder with drag-and-drop simplicity
  • AI-powered workflow optimization suggestions

Tech stack

Frontend
  • React 18 + TypeScript (Vite)
  • React Flow / @xyflow/react (graph editor)
  • Zustand (state)
  • Zod (schema validation)
  • Tailwind CSS + Headless UI
  • dnd-kit (drag & drop)
  • React Router
  • Vitest + Testing Library
Backend
  • Python FastAPI (API)
  • Uvicorn
  • PostgreSQL + SQLAlchemy 2
  • Redis + RQ (workers/queues)
  • httpx + Tenacity (resilient I/O)
  • Pydantic v2
  • OpenAI integrations
  • n8n API client
Infrastructure
  • Docker Compose (local stack)
  • Services: Postgres, Redis, n8n, MinIO, Backend, Worker, Frontend
  • Object storage via MinIO

Key challenges

  • Graph-to-workflow translation with strict typing and safe execution mapping to n8n
  • Realtime streaming across nodes with backpressure, retries, and reconnection handling
  • Scalable state persistence and template versioning for large graphs
  • Performance in large canvases: virtualization, zoom/pan smoothness, and memoized renders
  • Secure integration with external APIs (n8n/OpenAI) and key management

Future plans

  • Multi-tenant workspaces with RBAC, audit logs, and org policies
  • Observability: Sentry + tracing/metrics for node execution and API hops
  • Queue scale-out: Redis cluster today; evaluate Celery/RabbitMQ or NATS/Kafka later
  • Plugin SDK and template marketplace for custom nodes
  • Vector-store adapters (Pinecone/Chroma) with managed embeddings pipelines
  • Realtime collaboration (CRDT/Yjs) with presence and conflict resolution

These upgrades harden reliability, enable enterprise governance, and open ecosystem growth via plugins and templates, while improving UX with realtime collaboration and deep observability.