Diagnostics

Installation checks

Run bin/rails open_blog:doctor to inspect configuration, assets, routes, storage, and publishing records. Errors return exit status 1; warnings identify setup still needed. The read-scoped API GET /blog/api/v1/doctor and MCP blog_doctor expose the same checks.

Surface reports

Inspect live reader pages and their publishing records with bin/rails open_blog:report SCOPE=post POST=article-slug, the read-scoped GET /blog/api/v1/report?scope=post&post=article-slug, or MCP blog_get_surface_report. The API returns JSON; use Accept: text/plain or format=text for text. Scope site checks site-wide pages, feeds, lists, and redirects. Scope post adds one public article; all checks public articles in batches of 50 using page (or task PAGE). Each result includes the inspected IDs and pagination totals.

Set config.public_base_url to the running host. Reports fetch the rendered pages and compare visible fields, metadata, dates, links, image bytes, FAQ content, approvals, and disclosures with stored records. Results distinguish pass, fail, not applicable, and not verified, and include evidence and explicit limits. They do not certify editorial quality. Missing configuration, unavailable evidence, and exhausted request budgets remain unverified. Browser layout, manual accessibility review, and field performance are listed separately as checks not run.

Set reach=true (task REACH=true) to also check crawler access, HTTP-to-HTTPS upgrades, and same-site link destinations. Requests identify themselves as a bot and do not add page views. Fetches are bounded by time and size, reject private network destinations outside the configured origin, and stop starting requests after 60 seconds, 500 distinct URLs, or 32 MiB of completed response bodies. Narrow the scope when a report reaches a limit. The configured origin is trusted so local and private hosts can inspect themselves.