[bwiz]

Summary

CLI 1.1.0 follows the Databox API 2.1. Every type and request body was checked field by field against the API's contracts.

• metric data (new): reads a metric's values over a date range, with --granularity (no hourly), dimensions and filters. --json returns the API's response unchanged. syncInfo is {status, lastSyncedAt, message} with the API's statuses: synced, skipped, inProgress, neverSynced, failed. The timeout is 170 s, longer than the API's 150 s sync wait.
• dashboard topic (new):
  • dashboard list (refuses a blank --search), dashboard metrics <id> and dashboard rename <id> --name.
  • rename sends exactly {name}; name rules are left to the API.
  • databoard list / databoard metrics stay, on the new routes, and are deprecated: oclif prints a warning on stderr that points to dashboard …, and --json stdout stays clean.
• metric update --description sets a metric's description, and --description "" clears it. metric get shows a Description line. metric create gets no flag, because the API's create doesn't accept a description.
• dataset list --type dataset|mergedDataset, plus a Type column. Combining mergedDataset with --data-source-id is left to the API to refuse.
• 409 while a dataset modification is applied: the error ends with "Retry after N s", read from Retry-After. There is no automatic retry.
• Types:
  • DatasetListItem.type;
  • MetricListItem.description;
  • MetricDetail.logicalOperators (no longer optional);
  • new DataboardListItem.
• Docs (the npm page):
  • README: the Dashboards, Metric Descriptions and Reading Metric Values sections, Errors (409 / Retry-After), and the regenerated command reference.
  • CHANGELOG 1.1.0, written in place.
  • skills/databox-metrics and skills/databox-datasets.

The CLI makes no v1 calls, so the API's v1 admin-only change doesn't affect it. Its new error shapes and permission 403s are shown the way they come.

Also in 1.1.0: --api-url validation (#9 plus the follow-up fixes)

• fix: reject an --api-url that is not an http(s) URL, with exit 2 #9 (merged into main, brought in here): --api-url, DATABOX_API_URL and a stored apiUrl must be an http(s) URL. A bad flag or env value exits 2 before any prompt or request, a bad stored value exits 1, and auth login refuses a bad --api-url before reading the key.
• Follow-up from the review of fix: reject an --api-url that is not an http(s) URL, with exit 2 #9:
  • Messages say what is wrong and never repeat the value, so a key pasted as the URL stays out of CI logs.
  • Surrounding whitespace is ignored: requests are built from one normalised base, and auth login saves that.
  • auth login without --api-url refuses a bad stored URL before reading the key.
  • A username, password, query string or fragment is refused.
  • A stored null falls back to the default.
• Tests: 734 unit tests passing.

Verified

• Clean build; test typecheck and lint pass.
• npm test: 672 passing.
• Unit tests:
  • request bodies asserted with lastBody for dashboard rename, metric update --description (including "") and metric data;
  • flag validation for --type, --granularity, --search and the IDs.
• e2e on a 2.1 dynamic environment: the dashboard, databoard, dataset and metric suites gave 64 passing, 0 failing, 0 skipped.
  • Dashboard rename: it runs under withRestore(), and the name is restored.
  • Description: it is set, read back and cleared with "" on a metric the suite created.
  • --type mergedDataset: the API refuses it when combined with --data-source-id.
  • Unknown dashboard id: the API answers "Dashboard not found" on id.
• npm pack --dry-run: 1.1.0, with the dashboard commands included.

Noticed, not changed

• About 15 existing comments, including in .claude/rules/commands.md and e2e-testing.md, claim @oclif/test's runCommand refuses an empty flag value. It doesn't; update.test.ts now passes "".
• The other list commands still drop a blank --search silently; only dashboard list refuses it, as commands.md requires.

🤖 Generated with Claude Code