How Xianyu Accelerated Delivery with a 2‑1‑1 Lean CI Pipeline

1. Requirements and Challenges

Business demands faster delivery; delayed releases diminish product value.

Team size growth increases communication and coordination costs across multiple platforms.

Parallel branch development adds packaging and manual testing overhead.

Frequent test submissions require faster test response.

2. Client Continuous Integration Difficulties

Linking requirement creation to code branches, packaging projects, and test execution.

Ensuring traceability between requirements and code.

Automatically triggering builds from code changes.

Associating code changes with test scopes.

Managing merge conflicts and maintaining trunk stability with many parallel branches.

Triggering tests immediately after code submission and moving testing left.

Reducing automated test costs while improving efficiency.

Lack of a unified tool platform for client‑side integration.

3. Implementing Client CI

To address the end‑to‑end workflow from requirement creation to release, Xianyu built a pipeline that automatically binds newly created requirement branches to the corresponding packaging projects, triggers builds on each commit, and launches automated test suites based on merge requests or push events. Test results are fed back instantly, highlighting issues on the project board to enable rapid defect resolution.

Automation and Unmanned Process

With the goal of unmanned, automated, and scalable CI, the team split work into two parts: lean development process support and automated test verification. The "fish CI" component handles demand binding, change listening, build triggering, and test execution, while "fish guard" schedules test jobs (UI traversal, UI recognition, monkey testing, unit tests) and reports results. Future plans include adding static code analysis, Weex automation, and server‑side tests.

Data Metrics

Following Peter Drucker’s principle, the pipeline tracks metrics in three dimensions: responsiveness (build time, feedback speed, integration frequency), efficiency (pipeline success rate), and quality (code changes, module distribution, bug counts). These indicators quantify continuous delivery capability and guide improvement.

4. Results

Since adopting the lean model in March, Xianyu reduced release cycles from one month to two weeks, increased build frequency, accelerated integration, and achieved higher quality, as shown by faster bug‑trend reduction and higher test coverage in the presented charts.

5. Conclusion and Future Plans

The team first unified the delivery process, then accelerated it, and finally achieved continuous, nightly, and on‑demand integration, enabling 24/7 releases with fully automated pipelines and testing.