Show HN: Giselle – open-source visual editor for building AI workflows
github.comGitHub: https://github.com/giselles-ai/giselle
Giselle is an open source tool for designing and running AI workflows using a visual node-based editor. Apache 2.0 licensed.
Key features:
- Drag-and-drop workflow builder for chaining AI apps
- Multi-model support (OpenAI, Anthropic, Google Gemini) in one canvas
- GitHub integration for automating issues, PRs, and code reviews
- Built-in knowledge store with vector search
- Self-hostable with Docker or run locally
We built this because we needed a way to compose multiple AI models for complex tasks without writing glue code. The visual editor makes it easy to see the flow and debug when things go wrong.
Tech stack:
- Frontend: Next.js, TypeScript, Tailwind CSS
- Visual Node Editor: React Flow, Zustand, Tiptap
- AI: Vercel AI SDK, AI Gateway
- Database: PostgreSQL with Drizzle ORM
- Auth & Storage: Supabase
- Background Jobs: Trigger.dev (for long-running agent tasks)
- Analytics: PostHog
Happy to answer questions about the architecture or implementation. How does this compare to n8n? n8n is more powerful as a general-purpose workflow automation platform.
Giselle is more focused on AI workflows — simpler playground UI for team use, easier to iterate on content creation and knowledge analysis agents. Also has native GitHub integration (events as triggers, codebase vector store), designed for use cases like code review agents.