[Feature]: Export AI execution steps as reusable Playwright scripts

[surebye]

English Translation

What problem does this feature solve?

🚀 Feature Request

Allow Midscene to export AI execution steps as reusable Playwright scripts.

💡 Background

Midscene currently provides powerful AI testing capabilities:

• Execute tests via natural language (aiAct)
• Automatic page understanding and interaction
• Execution logs and reports
• Can generate Playwright-like test case examples

🎯 Problem

I need:

• Repeatable, deterministic scripts
• Stable selectors (avoiding flakiness from visual recognition)
• CI/CD execution (no AI dependency, reducing cost and uncertainty)

Currently missing:

After AI executes once successfully, can the corresponding Playwright script be exported for independent future execution?

✅ Expected Behavior

Provide an export capability that converts successfully executed test cases into actual Playwright scripts.

🧠 Desired behavior:

• Convert AI execution steps into Playwright code
• Generate stable selectors where possible, e.g.: getByRole, getByText, with fallback to CSS / XPath when necessary
• Preserve execution order and semantics
• Allow users to manually optimize after export

What does the proposed API look like?

N/A

---

原始中文描述 / Original Chinese

What problem does this feature solve?

🚀 功能请求

希望 Midscene 能支持：将 AI 执行过程导出为可复用的 Playwright 脚本。

---

💡 背景

当前 Midscene 提供了非常强大的 AI 测试能力：

• 通过自然语言执行测试（aiAct）
• 自动完成页面理解与操作
• 提供执行日志和报告
• 可以生成类似 Playwright 的测试用例示例

🎯 问题

我需要：

• 可重复执行的确定性脚本（deterministic）
• 稳定的 selector（避免视觉识别带来的波动）
• 在 CI/CD 中运行（不依赖 AI，降低成本和不确定性）

目前缺少一种能力：

能否在 AI 执行一次后，导出对应的 Playwright 脚本，并在后续独立执行？

✅ 期望

提供一种导出能力，将本次执行成功的用例导出为实际playwright的脚本

🧠 期望行为
将 AI 执行步骤转换为 Playwright 代码
尽可能生成稳定 selector，例如：
getByRole
getByText
必要时 fallback 到 CSS / XPath
保留执行顺序和语义
允许用户在导出后手动优化

What does the proposed API look like?