feat: add Bloomberg adapter

[KasumiChen]

Description

Adds a Bloomberg adapter with RSS-backed listing commands plus browser-backed article extraction for standard Bloomberg story/article pages.

This PR also includes the small manifest/compiler fix needed so helper/test TS files in src/clis/bloomberg/ do not get exposed as fake commands in opencli list.

Related issue:

Type of Change

• 🐛 Bug fix
• ✨ New feature
• 🌐 New site adapter
• 📝 Documentation
• ♻️ Refactor
• 🔧 CI / build / tooling

What this adds

New Bloomberg commands:

• opencli bloomberg main
• opencli bloomberg markets
• opencli bloomberg economics
• opencli bloomberg industries
• opencli bloomberg tech
• opencli bloomberg politics
• opencli bloomberg businessweek
• opencli bloomberg opinions
• opencli bloomberg feeds
• opencli bloomberg news <link>

What works now

• The listing commands above are RSS-backed via feeds.bloomberg.com and return structured headline data without requiring a browser.
• bloomberg news can read standard Bloomberg story/article pages and extract:
  • title
  • summary
  • canonical link
  • media links
  • full article text
• bloomberg news accepts either a full Bloomberg URL or a relative Bloomberg path.
• opencli list/manifest output only shows real Bloomberg commands (not helper files like utils or test files).

Out of scope / current limitations

• Audio pages and some other non-standard Bloomberg URLs are intentionally out of scope for this PR and may still fail.
• Pages that do not expose the expected __NEXT_DATA__ story payload may still fail.
• Bloomberg bot-protection / access-gated pages may still fail.
• This version is for data access/extraction only. It does not bypass Bloomberg paywall, login, entitlement, or other user-side access requirements. Whether an article is readable still depends on the user's own browser/session state.

Docs / wording updates

• Added a new adapter doc page for Bloomberg.
• Updated adapter docs index and VitePress sidebar.
• Added README notes that clarify:
  • which Bloomberg commands are RSS-backed
  • that bloomberg news is for standard story/article pages
  • that audio/non-standard pages may fail
  • that OpenCLI does not bypass Bloomberg access controls
• Updated bloomberg news command/help text and error messages to be explicit about current session/access requirements.

Tests / manual checks run

Automated

• npm run build
• npm run docs:build
• npx vitest run src/clis/bloomberg/utils.test.ts src/build-manifest.test.ts
• npx vitest run tests/e2e/public-commands.test.ts tests/e2e/browser-public.test.ts -t bloomberg

Manual

• node dist/main.js list -f json
  • verified Bloomberg commands are registered and helper/test files are not exposed as commands
• node dist/main.js bloomberg feeds -f json
• node dist/main.js bloomberg main --limit 1 -f json
• node dist/main.js bloomberg news <article-link> -f json
  • verified article extraction on an accessible standard /news/articles/... page

Checklist

• I ran the checks relevant to this PR
• I updated tests or docs if needed
• I included output or screenshots when useful

Documentation (if adding/modifying an adapter)

• Added doc page under docs/adapters/
• Updated docs/adapters/index.md table
• Updated sidebar in docs/.vitepress/config.mts

Screenshots / Output

Relevant manual smoke summary:

• bloomberg main --limit 1 -f json returned live RSS data successfully
• bloomberg news successfully extracted a standard article page with non-empty content

[ByteYue]

Review: Bloomberg Adapter PR

Nice work on this PR! Well-structured with good documentation and comprehensive tests. I've pushed a follow-up commit with a few improvements:

Changes made (c5a0bac):

1. news.ts: Reordered execution flow — page.goto(url) now comes before loadStory definition for clearer top-down reading. Increased wait times (5s initial + 4s retry) for Bloomberg's heavy Next.js pages.

2. utils.ts: Clarified the hostname regex in validateBloombergLink — replaced ([.]|^) with (?:\\.|^) (non-capturing group form is unambiguous).

3. build-manifest.ts: Added console.warn when scanTs() fails to parse a file, matching the existing scanYaml() pattern. The previous silent return null could hide real I/O errors.

4. public-commands.test.ts: Replaced the for-loop over 7 sections with it.each() for better test isolation — each section now gets its own test case with independent pass/fail reporting.

Considered but kept:

• Consolidating the 8 feed command files into a single factory — this would break the static manifest scanner which relies on regex to extract name: and description: from source files. Individual files are the correct pattern for this codebase.

LGTM overall — solid first adapter contribution! 🎉