Example Projects

Each official example is a standalone project directory hosted as its own repository under the CodefyUI organization. Clone one, start the server on it, and you have a working service on your canvas -- then publish it as a real HTTP API and push the repo anywhere you like. They are also honest templates: fork one and swap in your own graph.

Every repo follows the same three commands:

git clone https://github.com/CodefyUI/<example-name>
cd <example-name>
cdui start --project .

Open the editor at the URL the server prints, open the graph from the sidebar's Graphs tab, and press Run. If the welcome screen shows instead, no tab is open and there is no sidebar; click New blank graph first (see Tabs & Persistence). Each README walks through the full publish flow (commit, cdui project publish, mint an API key, invoke with curl) and lists anything the example needs beyond CodefyUI itself.

The examples

Repository	What the service does	Needs
example-word-analogy	Word-vector analogy lookup: three words in, nearest analogy words out.	Nothing -- fully offline.
example-tabular-predictor	Tabular classifier: feature rows in, class predictions out.	Nothing -- fully offline.
example-llm-document	Document summarizer: document text in, summary out.	A local Ollama install.
example-mnist-train-serve	One project, two graphs: train a small CNN on MNIST on the canvas, then serve YOUR trained weights as a digit-recognition API.	MNIST download on first training run (about 12 MB; about 66 MB on disk once unpacked).

Why they are separate repositories

A published service is code you own: it deserves its own history, its own remote, its own CI. Keeping each example as a real repository (instead of files bundled inside CodefyUI) means the clone-to-published-API path you practice on an example is exactly the path you will use for your own services. Each example repo runs cdui project validate . cleanly, so any of them also works as a template for CI validation of your own project.