Build a custom AI workflow builder for your product where users connect LLM nodes, HTTP calls, and conditional branches visually.
Create a node-graph editor for an automation platform without writing the canvas or variable-passing infrastructure from scratch.
Prototype a visual pipeline that loops through a list, fetches external data, and generates AI suggestions, all viewable in the browser.
Add a structured workflow UI to an existing app using FlowGram's pre-built node types and form system as npm packages.
Requires Node.js and npm, start from the CLI template then add custom nodes for your specific use case.
FlowGram is a developer framework for building visual workflow editors, the kind of drag-and-drop node graph interfaces popular in AI automation tools. It is not itself a finished workflow product. It is the set of building blocks you use to create your own workflow platform. The framework provides a canvas where nodes can be placed and connected, a form system for configuring each node's settings, a variable system so data can pass between nodes with proper scoping, and a library of pre-built node types including LLM (AI model calls), conditional branching, code execution, HTTP requests, and loop constructs. You assemble these pieces to build a custom workflow editor tailored to your product. The README demo shows a working example: a workflow that loops through a list of cities, fetches weather data over HTTP for each one, parses the result with a code node, generates a clothing suggestion using an AI model, filters by a condition, and produces a final ranked recommendation. This runs visually in the browser as a node graph. Starting a new project uses a single command that sets up a template, then you run it locally and open it in a browser. The framework is published as npm packages and is built with TypeScript. It comes from ByteDance, the company behind TikTok, and targets developers building AI workflow products who want to skip building the canvas and core infrastructure from scratch. More detail is available at the project's documentation site.
← bytedance on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.