External reach — CDSL's scholarly footprint

Who builds on, clones, ships, and cites the Cologne Digital Sanskrit Lexicon (CDSL) — framed as scholarly reach, not funder-facing vanity metrics (Workstream G6). Measured signals (stars/forks, GitHub traffic, downstream dependents) carry a URL and/or fetch date; the citation tier is a systematic OpenAlex sweep whose recall bounds are stated rather than hedged. Source: reports/external_reach.md, generated by scripts/external_reach.py; citation method and full result set in reports/citation_sweep.md.

The stars-vs-clones gap is the finding. The org collects roughly stars in total, yet its core repositories are cloned times in a single 14-day window. CDSL is consumed as infrastructure — cloned, mirrored, and served through the Cologne website — not favourited on GitHub. Star count badly understates reach.

GitHub traffic — the real usage signal

How to read: Clones over a rolling 14-day window (GitHub only exposes two weeks, and only to accounts with push access). This is a sample of core infrastructure repos, not all 76, and the window slides — read it as a spot measurement, not a cumulative total. Bars are clone counts; the tooltip notes unique cloners. Example: csl-orig — the master dictionary source — draws the heaviest clone traffic because every downstream build pulls it.

Conclusion: Two weeks of clones ( across the sampled core) dwarf thirteen years of stars (). The audience is builders and mirrors pulling data programmatically, not GitHub stargazers — exactly the profile of a piece of scholarly infrastructure.

Downstream dependents — who ships CDSL

Third-party projects that ship, wrap, or serve CDSL data. Each is a URL-checkable project that chose the Cologne lexicon as its lexical backbone — the strongest scholar-facing reach evidence.

Plus further external repositories surfaced by GitHub code search referencing org raw-URLs or the Cologne site (a floor — code search indexes only a subset of public code):

Scholarly citations — systematic sweep, stated bounds

How to read: works were found by a systematic OpenAlex sweep — citation-graph anchors (cites:<id>, identifier-exact) plus name-unique phrase probes, deduplicated, domain-gated, with the project's own Zenodo records removed. This is a documented lower bound, not a census: works citing the dictionary only by siglum (MW s.v. …) are invisible by design, because enumerating "MW" would drag in millions of unrelated works. Method and every excluded probe's measured collision count: reports/citation_sweep.md.

The distinction that must not blur. A further works attest the print dictionaries CDSL digitises (Monier-Williams, Böhtlingk, Apte). Citing Monier-Williams 1899 says nothing about whether the Cologne digital edition was used, so that envelope is reported separately and is never added to the above. Summing them would be the easiest way to make this tier dishonest.

Zenodo OBS-T stats — blocked (DOI mismatch)

Once the correct OBS-T record exists, update ZENODO_RECORD_ID in scripts/external_reach.py and re-run --fetch to populate this tier.

Last API fetch: . API tiers are cached (committed) under reports/external_reach_cache/ so this page regenerates offline. Object of analysis: repository metadata, GitHub traffic, third-party code references, and publication citations — in scope per docs/BOUNDARY_RULES.md. Roadmap: Workstream G6.