flowgram.ai

FlowGram is a composable, visual, easy-to-integrate and extensible workflow development framework and toolkit from ByteDance, aimed at helping developers build AI workflow platforms faster and more simply. The README is clear that it is not a ready-made workflow platform: it provides the building blocks, namely a flow canvas, node configuration forms, a variable scope chain and ready-to-use materials such as LLM, Condition and Code Editor nodes.

The feature list covers a free layout canvas where nodes can be placed anywhere and joined with free-form lines, a fixed layout canvas with compound nodes such as branches and loops, a form engine that handles node data, rendering, validation and side effects, and a variable engine with scope constraints, structure inspection and type inference. The demo iterates over a list of cities, fetches weather over HTTP, parses temperatures with a Code node, generates outfit suggestions with an LLM and gates by a condition. To try it, run `npx @flowgram.ai/create-app@latest` (the Free Layout Demo template is recommended), then npm install and npm start. The README is available in seven languages and the code is MIT licensed.