Description

In tools/web_tools.py, four asyncio.gather() calls do not use return_exceptions=True. If any single LLM summarization task raises an exception, the entire gather fails and all other successfully completed results are lost.

Affected Locations

Impact

When extracting content from multiple web pages in parallel, a transient failure on one page (e.g., rate limit, timeout) causes the agent to lose all results, even from pages that were successfully fetched and summarized.

Fix

Add return_exceptions=True to all four asyncio.gather() calls and filter out BaseException instances from results with appropriate logging.