feat: add sentry schema command for API introspection

[BYK]

Phase 2 of #346: Add sentry schema command for runtime API introspection by AI agents.

What

Enables AI agents to discover Sentry API endpoints at runtime without pre-loaded documentation:

sentry schema                        # List all API resources with endpoint counts
sentry schema issues                 # Endpoints for 'issues' resource
sentry schema issues retrieve        # Detailed endpoint info (method, path, params, description)
sentry schema issues retrieve --json # Machine-readable output
sentry schema -q replay              # Search across all endpoints
sentry schema --all                  # Flat list of all 215 endpoints
sentry schema 'monitor*'             # Glob pattern matching on resources

Architecture

Build-time generation — script/generate-api-schema.ts fetches the dereferenced OpenAPI spec from getsentry/sentry-api-schema (version-tagged to match the installed @sentry/api package) and extracts:

• HTTP methods and URL templates from the OpenAPI paths
• Human-readable descriptions and deprecation status
• Query parameter names from OpenAPI parameter definitions
• SDK function name mapping from the @sentry/api JS bundle
• Resource names derived from URL paths (last non-parameter segment)

Output: src/generated/api-schema.json (215 endpoints, ~117KB) — gitignored, generated at build/dev/test time via bun run generate:schema.

Runtime — src/lib/api-schema.ts provides query functions:

• getResourceSummaries() — grouped resource overview
• getEndpointsByResource() — filter by resource
• getEndpoint() — find specific endpoint (prefers operationId prefix match)
• searchEndpoints() — full-text search across all fields
• Glob pattern support via regex conversion for resource matching

The sentry schema command yields CommandOutput through the standard output system with mdKvTable + renderMarkdown for endpoint detail views.

Other changes

• Updated @sentry/api from 0.21.0 → 0.54.0
• CI workflow updated to run generate:schema before lint, typecheck, and tests
• src/generated/ added to .gitignore

Tests

19 unit tests for schema query functions covering all query modes, edge cases, and data invariants.

[github-actions]

Semver Impact of This PR

🟡 Minor (new features)

📋 Changelog Preview

This is how your changes will appear in the changelog.
Entries from this PR are highlighted with a left border (blockquote style).

---

New Features ✨

• Add sentry schema command for API introspection by BYK in #437

---

_{🤖 This preview updates automatically when you update the PR.}

[github-actions]

Codecov Results 📊

✅ 111 passed | Total: 111 | Pass Rate: 100% | Execution Time: 0ms

📊 Comparison with Base Branch

Metric	Change
Total Tests	—
Passed Tests	—
Failed Tests	—
Skipped Tests	—

✨ No test changes detected

All tests are passing successfully.

✅ Patch coverage is 100.00%. Project has 1101 uncovered lines.
✅ Project coverage is 95.11%. Comparing base (base) to head (head).

Coverage diff

@@            Coverage Diff             @@
##          main       #PR       +/-##
==========================================
+ Coverage    95.09%    95.11%    +0.02%
==========================================
  Files          165       167        +2
  Lines        22424     22531      +107
  Branches         0         0         —
==========================================
+ Hits         21324     21430      +106
- Misses        1100      1101        +1
- Partials         0         0         —

---

Generated by Codecov Action

[BYK]

Addressing review comments

1. require.resolve — Now using require.resolve('@sentry/api/package.json') to locate the package instead of hardcoded node_modules path.

2. OpenAPI spec instead of JS parsing — Rewrote the generator to fetch the dereferenced OpenAPI spec from getsentry/sentry-api-schema. The version tag is derived from the installed @sentry/api package version (now 0.54.0). This gives us:

   • Proper operationId from the spec (e.g., "List an Organization's Issues")
   • Query parameters extracted from OpenAPI parameter definitions
   • Accurate deprecation flags from the spec
   • No fragile regex parsing of JS bundles

   The SDK function name mapping is still extracted from @sentry/api's bundled JS (to map operationIds to their TypeScript function names), but this is a secondary lookup — the primary data comes from the OpenAPI spec.

3. Updated @sentry/api from 0.21.0 → 0.54.0 to match the latest schema.

[BYK]

Round 3 review fixes

All six comments addressed:

1. mdKvTable + renderMarkdown for endpoint detail — Replaced manual string building with mdKvTable key-value rows (Resource, Operation, Function, Status, Path Params, Query Params) rendered through the markdown pipeline.

2. Renamed --list → --all — Flag now called --all for showing flat endpoint list.

3. Fixed padding before color formatting — Color is now applied to the raw method string, then spaces are appended outside the ANSI escape codes.

4. Empty array guard in formatResourceList — Returns early with "No resources found." before Math.max call on empty arrays.

5. Added -q alias for --search — Can now use sentry schema -q monitor as shorthand.

6. Glob-style search — Positional args containing * or ? are treated as glob patterns via Bun.Glob. Example: sentry schema monitor* matches resources starting with "monitor".

Re: fuzzy/levenshtein — there's a private levenshtein() in platforms.ts but it's not exported. Extracting it to a shared module is a good follow-up but out of scope for this PR. The substring + glob search covers the primary use cases.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}