lychee.toml

# Config for the lychee link checker (https://lychee.cli.rs/usage/config/)
# Run in CI via .github/workflows/pr-checks.yml

timeout = 20
retry_wait_time = 5
max_retries = 3

# Some sites (Cornell Merlin, Crunchbase) block the default lychee/curl
# user agent as a bot-filtering heuristic; a browser-like UA resolves it.
user_agent = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0.0.0 Safari/537.36"

# Example/placeholder hosts used in API documentation samples — not real endpoints.
# The next few entries are real, live sites that reject automated/CI
# requests regardless of user agent (verified manually) and are excluded
# rather than accepted outright, so a genuine break elsewhere still fails:
#   - linkedin.com returns a nonstandard 999 to any non-browser client
#   - merlin.allaboutbirds.org 403s CI IP ranges
#   - crunchbase.com intermittently 403s automated requests
#   - the GCP console has a known HTTP/2 handshake issue with automated clients
# The last entry excludes the site's own canonical/self-referencing URLs
# (e.g. mkdocs-material's <link rel="canonical">), since a page added or
# moved in a PR won't exist on the live deployment until the PR merges —
# checking those against production would always 404. Everything else on
# the page is still checked normally, either as a relative path against
# the local build or as a real external link.
exclude = [
  "^https?://localhost",
  "^https?://<.*>",
  "^https?://(www\\.)?linkedin\\.com",
  "^https?://merlin\\.allaboutbirds\\.org",
  "^https?://(www\\.)?crunchbase\\.com",
  "^https?://console\\.cloud\\.google\\.com",
  "^https?://brstuder\\.github\\.io/portfolio/",
]

# mkdocs-material's generated 404 page intentionally uses site_url-rooted
# absolute paths (it can be served from any depth on the real domain), which
# only resolve against the live deployment, not a local build.
exclude_path = [
  "404\\.html$",
]

accept = [200, 201, 202, 203, 204, 206, 301, 302, 303, 307, 308, 429]