Rewritten in place as work lands. One section per live thread; each open item names its next step and an issue or the reason it has none.
Released 01/11/2026: tag v2026.11.0 at 083566a, on PyPI, wheel 2026.11.1103324 on the household index, mcp-servers' lock pinned to it (mcp-servers 18d7660).
change runs on main were cancelled three times in a row on 01/11 by newer landings (41258347187, 41258661170, 41258879192), so the image carrying the new lock has not deployed. Next step: once a change run on mcp-servers main completes green, call supacrawl_health and expect version.supacrawl_lib == 2026.11.1103324, then one scrape through it; tracked on #180.foreman/wt-5ea830aa carries search image rows with image_url + width/height and scrape image_details. Not released or live. Next step: merge, release, bump the mcp-servers lock, redeploy supacrawl-mcp, then repeat the MQ-28 image search and the Commons scrape on the gate; tracked on #180.quotes.toscrape.com/scroll returns 34 characters from every reader except Crawl4AI's scan_full_page. Next step: decide whether a scroll-until-stable pass belongs in BrowserManager.fetch_page; [no-track: a synthetic page; an issue is filed when a real page needs it].USE_DB_AUTHENTICATION=false and a database URL to start; docs/benchmark.md records the outcome of the attempt on 01/11/2026. Re-run with supacrawl bench run --provider firecrawl once a Firecrawl API answers on FIRECRAWL_API_URL; [no-track: a hosted account is out of scope; the self-hosted row is in the table].docs/benchmark.md was measured on a 20-core host at load average about 30. Next step: re-run supacrawl bench run --provider <each> --no-reference on a quiet host and refresh the two tables; [no-track: one command per reader on a quiet host].