SlothQA: let an open model do the tapping, make evidence decide the result

Description and use

SlothQA uses a local Qwen model to automate Android UI testing, creating a task and verifying its presence without manual interaction.

Problem

Automating UI testing for Android apps is tedious; developers spend time opening the app, navigating screens, entering test data, and verifying results.

Outcome

In demonstration runs, SlothQA achieved independent assertion passes for normal and reordered control scenarios within 12.65–26.09 s, and correctly halted on a deliberately broken save action after 34.82 s.

Workflow

• Observe current UI state
• Propose next action from a constrained menu
• Validate proposed action with an independent checker
• Act on the chosen action via AI App Bridge
• Observe resulting UI state

Evidence

SlothQA is a new local-model regression agent that takes a goal instead of a recorded sequence of taps.

Source

At each step, a local Qwen model sees the current UI and chooses its next action.

Source

The first attempt was worse. A loose JSON response allowed missing fields and incorrect text values.

Source

Fifteen automated tests passed, including a test against the actual Bridge Intent implementation.

Source