fix(deep_crawling): allow single-label hostnames

[nightcityblade]

Summary

Fixes #2079.

Allow BFSDeepCrawlStrategy to process valid HTTP(S) URLs whose host is a single label, such as internal DNS names. The existing scheme and non-empty netloc validation continues to reject malformed URLs.

List of files changed and why

• crawl4ai/deep_crawling/bfs_strategy.py - Remove the overly strict requirement that every hostname contain a dot.
• tests/deep_crawling/test_deep_crawl_resume.py - Add a regression test proving a single-label hostname reaches and passes the configured domain filter.

How Has This Been Tested?

• .venv/bin/pytest -q tests/deep_crawling/test_deep_crawl_resume.py — 33 passed.
• .venv/bin/black --check crawl4ai/deep_crawling/bfs_strategy.py tests/deep_crawling/test_deep_crawl_resume.py — run; the command reports existing whole-file formatting drift also present on the upstream baseline. The changed lines were reviewed against Black's diff and introduce no new formatting changes.

Checklist:

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas — N/A; the validation removal is self-explanatory.
• I have made corresponding changes to the documentation — N/A; this is a narrow bug fix with no public API change.
• I have added/updated unit tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes

[keepitdigital]

k

[chuenchen309]

Reviewed against can_process_url — this looks correct and safe.

The removed "." not in parsed.netloc check was never the domain-restriction mechanism: right below it, if depth != 0 and not await self.filter_chain.apply(url) is what actually enforces DomainFilter/allowed-domains, and that is untouched. So the dot-check was purely a crude structural guard that wrongly rejected valid absolute URLs with a single-label host — not just intranet names but also http://localhost, http://localhost:8080, and Docker/K8s service hostnames, all of which have a non-empty netloc and a valid http(s) scheme.

Points that make this safe:

• No domain/isolation regression: domain scoping still runs through filter_chain.apply(); single-label hosts now merely reach that filter (exactly what the new DomainFilter(allowed_domains=["intranet"]) test asserts) instead of being dropped before it.
• Malformed input still rejected: relative paths / sch/netloc-less strings still fail the existing not parsed.scheme or not parsed.netloc guard (empty netloc → rejected), so nothing new slips through.
• Unfiltered crawls unchanged in spirit: with no DomainFilter configured, dotted hosts were already crawled unrestricted; single-label hosts just join them — the user opted into that by not setting a filter.

The regression test is well-targeted (proves the host both reaches and passes the configured filter). LGTM.