mirror of
https://github.com/browser-use/browser-use.git
synced 2026-09-14 19:59:47 +08:00
658d1ca623
When JS codegen extraction returns empty results ([], {}, {key: []}),
automatically fall back to markdown-based LLM extraction instead of
returning empty data that causes the agent to retry blindly.
Adds _is_empty_data() helper to detect effectively empty extraction
results including nested patterns like {"products": []}.
Addresses ~5 eval failures where extract_with_script returned [] on
pages with lazy-loaded or shadow DOM content that JS couldn't reach.