wtf is tantu-flow?

Tantu flow is a visual programming tool for building data pipelines by dragging and connecting nodes on a canvas instead of writing a program from scratch. Every connection you draw between nodes is really a NATS publish and subscribe channel underneath, which means a node can run inside the main program or as a completely separate process without you needing to change how it is wired up. The project is made of two main pieces: a single self contained HTML file that is the visual editor, with no build step required, and a Python backend that manages running flows and the NATS connections between nodes. You start a NATS server, either with Docker or a local install, then run the Python backend, and open your browser to the address it prints to reach the editor. A built in example flow shows a stock screener that fetches quotes, filters them by criteria, ranks the results, and displays the top picks. Nodes come in several types: sources that inject data on a schedule or manual trigger, transform nodes that run plain Python code or route messages based on rules, and sink nodes that display debug output or send data to an external NATS subject. Regular function nodes run inside the main backend process, while function process nodes run as fully separate operating system processes, which protects the rest of the flow if that node crashes and is suited to heavy tasks such as data processing. You can also add entirely new node types by writing a small class in the backend and a matching entry in the editor's configuration. The backend exposes a REST API for listing, loading, deploying, and running flows, plus a WebSocket connection that streams live debug output and animates the wires in the editor as data moves through them. The last deployed flow is saved to disk and automatically redeploys when the backend restarts and NATS is available. The README is direct about one limitation: function nodes run using Python's exec, so you should only deploy code you trust, and wire connections do not validate that the data shape a node expects actually matches what it receives. Tantu flow is released under the MIT license.