TangoFlow: A Low-Code Flow-Based Programming Platform for Backend Service Orchestration at NetEase Cloud Music

NetEase Cloud Music has adopted a GraphQL-based BFF development model, which improves frontend‑backend collaboration but fails to solve core pain points of backend developers and introduces new issues.

The main problems are: low logic orchestration ability causing excessive use of unregulated Groovy scripts; high resource costs due to many idle service instances; complex interactions and steep learning curve that hinder interface debugging and troubleshooting; and unclear role division between frontend and backend teams, weakening service governance.

To address these challenges, the team seeks a way to assemble service assets, express service dependencies and business logic, reuse service assets, and integrate with the existing Tango low‑code UI platform to reduce development cost and improve delivery quality.

The answer lies in Flow‑Based Programming (FBP), a data‑flow paradigm where programs are represented as directed graphs with input/output nodes, intermediate processing nodes, and connections that automatically execute when inputs are provided.

TangoFlow implements this idea by providing a visual flow‑assembly platform built on the Ant Group‑opened X6 graph‑editing engine. Users drag‑and‑drop nodes (triggers, logic controls, basic services, data results) onto a canvas, connect them, and the platform serializes the graph into a DSL that a backend engine executes, exposing the results through gateway APIs or a unified SDK.

The product design includes a top navigation bar for app/branch info and core actions, a materials panel showing four node categories, a canvas for node layout and connection, a dynamic property panel for node configuration, a console for request/response and engine debugging, built‑in mock capabilities for service nodes, and a strict release checkpoint process to ensure stability.

TangoFlow enables online debugging (flow, node, and remote), mock‑driven testing, and visual edge‑labeling, connection validation, and node‑embedding features. Future work will integrate Tango’s model‑driven UI, further consolidate backend tech stacks, and explore AIGC/AI‑Agent capabilities for natural‑language‑driven workflow automation.