What changed

• Client::markdown() — new method returning search results as Markdown, alongside the existing html(). Requests /search with output=md.
• Version 0.4.1 → 0.5.0 in meson.build and the VERSION source tag (minor bump for a new public method).
• Doxygen comments on every method in serpapi.cpp (@brief/@param/@return), plus a Doxyfile so doxygen Doxyfile generates HTML docs into docs/ (gitignored).
• README — usage example, changelog entry, class diagram, download command bumped to v0.5.0, and a coverage-results section.
• Test — client.markdown in test/test_serpapi.cpp, mirroring the existing client.html test.

Why output=md

SerpApi documents three equivalent ways to request Markdown. I verified all three against the live API — each returns HTTP 200, text/markdown, and byte-identical bodies (34,885 bytes for the same query):

output=md is the one that fits the existing get(uri, output, parameter) plumbing — the other two would need path special-casing or new header support for no functional difference.

The response is YAML frontmatter (search_metadata) followed by ## sections (Organic Results, Knowledge Graph, Ai Overview, …), formatted for LLM/agent consumption.

Verification

Built and tested against current main with -Dtests=true, with SERPAPI_KEY set so the live-API tests actually execute:

• 23/23 tests pass, including client.markdown (865 ms, green)
• Coverage: lines 78.0%, functions 100% (15/15), branches 45.9%

Notes for reviewers

• This supersedes the markdown commits I previously pushed to docs/download-command. That branch's PR (docs: add curl download command to README #5) was already merged, so those commits were stranded and never reached main. This branch is a clean rebase of the same work onto current main. docs/download-command can be deleted.
• Follow-up not included here: the Markdown response advertises a markdown_endpoint, and /searches/<id>.md works (verified, 200) — but search_archive() hardcodes .json and returns a rapidjson::Document, so archived searches can't be fetched as Markdown. Worth a follow-up; it needs an API-naming decision (markdown_archive(id) vs. an output param on search_archive).
• The README download command now follows main's newer archive/refs/tags/ tarball format, just bumped to v0.5.0.

🤖 Generated with Claude Code