mirror of
https://github.com/jackwener/OpenCLI.git
synced 2026-09-14 18:25:42 +08:00
jakevin
68485cc54e
* feat(devto): surface article id + published_at on listings, add `read <id>` Agent-native gap: devto listings (`top`/`tag`/`user`) didn't include the article `id`, so an agent couldn't round-trip from a listing into a body read. They also dropped `reading_time` and `published_at`, which are cheap signals the API gives you for free. Changes: - `top` / `tag` / `user`: add `id`, `reading_time`, `published_at` columns alongside existing rank/title/etc. `user` keeps its no-author shape since it's already user-scoped. - New `devto read <id>`: hits `dev.to/api/articles/<id>` and returns one row with the article body (truncated by `--max-length`, default 20000, min 100). DEV.to's public API does not expose comments yet, so this is intentionally a single-row reader rather than a HN/lobsters-style threaded tree — if/when comments become public we can extend to POST + L0/L1. - Typed fail-fast: `ArgumentError` for non-numeric id and for `--max-length` below 100; `EmptyResultError` on 404; `CommandExecutionError` for other non-2xx HTTP statuses. No silent clamps. - Defensive tag normalization: the `/api/articles/<id>` endpoint returns `tag_list` as a comma-string and `tags` as an array (the opposite shape from listing endpoints). Caught this on live verification — both shapes now collapse to a comma-joined string. Tests: 12 vitest assertions covering listing column shape (all 3) + register/args/strategy + typed-error fail-fast paths + happy-path body extraction + truncation marker + alternate tag_list shape. Live verification: `devto top --limit 3` and `devto read 3602287` both return the expected agent-native shape. * fix(devto): harden article read contract