Kush Agrawal explains why Duolingo built an agent-assisted workflow to improve unit-test coverage in its iOS codebase.
He describes selecting files, generating test pull requests, monitoring continuous integrationWorkflow automation uses software or AI to complete a connected sequence of routine steps with less manual effort. and routing passing changes to human reviewersCode review examines proposed source changes to assess their correctness, clarity and suitability before or around integration.. The team added rules to prevent production edits and deleted or disabled tests.
Agrawal reports that the pipeline merged more than 700 human-approved test pull requests and that overall unit-test coverage increased from 11% to 24%Test coverage measures how much of a program is exercised when its tests run.. He discusses lint failures, flaky tests and applying the workflow to other migrations.
Watch on YouTube
