0.6.2 (aafed61)

browser-cli 0.6.2

0.6.2 是一次以全页阅读模式和参数验证加固为主的版本更新，汇总了自 0.6.1 以来的新功能与质量改进。

亮点

• 新增 page --all 全页阅读模式
  • browser-cli page <session-id> --all 会自动滚动完整页，等待懒加载渲染，并把所有逻辑分页聚合成一个阅读视图。
  • 新增 --settle <ms> 参数控制每次滚动后的等待时间，默认 500ms；适合需要更多时间渲染的懒加载页面。
  • 聚合结果统一归一化为 current=1 total=1 truncated=false，方便 AI 一次性拿到完整页面内容。
  • 文档明确说明 --all 输出更偏阅读视图，如需后续交互建议回到普通 page 重新获取元素 ID。
• CLI 参数约束与验证更严格
  • --all 与 -p / --next / --prev 互斥。
  • --settle 只能配合 --all 使用。
  • 新增 CLI 层参数解析测试用例，确保边界条件正确报错。

兼容性与行为说明

• 这是一次向前兼容的版本更新；新增参数是增量能力，不会影响现有 open / page / click / type / wait 工作流。
• 聚合页面中的元素 ID 由系统重新编号，与原始快照中的 ID 不同；这是设计预期，不是 bug。
• 聚合去重策略只跳过“后续页前缀中与前序页完全重复”的节点（主要是 sticky header），正文中的重复内容不会被误删。

English

0.6.2 is a feature release focused on full-page reading mode and tighter CLI parameter validation, covering new functionality and quality improvements since 0.6.1.

Highlights

• New page --all full-page reading mode
  • browser-cli page <session-id> --all auto-scrolls through every logical page, waits for lazy content to render, and aggregates the result into a single reading view.
  • A new --settle <ms> argument controls the fixed delay after each scroll, defaulting to 500ms; useful for pages with lazy-loaded content that needs extra time.
  • The aggregated result is normalized to current=1 total=1 truncated=false, so AI agents can ingest the full page in one shot.
  • Documentation clarifies that --all output is optimized for reading; follow-up interactions should use a regular page call to get current element IDs.
• CLI parameter constraints and validation are now stricter
  • --all is mutually exclusive with -p, --next, and --prev.
  • --settle only works in combination with --all.
  • New CLI-level argument parsing tests ensure boundary conditions raise the correct errors.

Compatibility and behavior notes

• This is a forward-compatible release. New arguments are additive and do not affect existing open, page, click, type, or wait workflows.
• Element IDs in aggregated output are reassigned by the system and differ from the original snapshot IDs; this is by design, not a bug.
• The deduplication strategy only skips nodes that are entirely repeated at the start of later pages (primarily sticky headers); repeated content within the main body is preserved.

Full Changelog: 0.6.1...0.6.2