[googio]

Summary

Adds serply_scholar, an optional /research source that gets Google Scholar results through the Serply API. It sits beside the scholarly-based google_scholar source and gives the same Scholar citation counts without scraping, so a query takes about 1-2s and does not hit CAPTCHAs.

It follows the Brave and Tavily sources: tier optional, SERPLY_API_KEY (or /config serply_api_key=...), and it raises SourceSkipped when no key is set, so nothing changes for users without one.

Changes

• cheetahclaws/research/sources/serply_scholar.py: the new source, registered for the academic domain.
  • Scholar's "Cited by N" goes to engagement_raw and engagement_label.
  • The year comes from the byline and goes to published.
  • The first three author names and the PDF and "cited by" links go to extra.
  • --range/--since/--until map to Scholar's own year filter (as_ylo/as_yhi), so the date filter runs on the server instead of after the fetch.
  • Results whose link is not an http(s) URL are dropped.
• cheetahclaws/research/sources/__init__.py: one import line.
• cheetahclaws/research/ranker.py: calibration of 100, the same as semantic_scholar and the default google_scholar falls back to.
• docs/guides/research.md: a row in the optional-sources table and a row in the time-range mapping table.
• tests/test_research.py: a skip-without-key test like the Brave and Tavily ones, plus a parse test with a mocked get. The parse test covers the citation count, year, author truncation, the year filter params, the key header and dropping a row with no link.

How to test

export SERPLY_API_KEY=...     # or /config serply_api_key=...
/research --sources serply_scholar --range 2y retrieval augmented generation

Validation

• python -m pytest tests/test_research.py -q: 90 passed.
• Full suite: 2915 passed, 25 skipped (Python 3.13, Linux).
• Live call through SOURCES["serply_scholar"].search(...) with a real key and parse_range("2y"): 5 results, all dated 2024-2026, each with a citation count, authors and a publisher URL.

Disclosure: I work on Serply. Happy to rename the source or adjust anything to match project conventions.