fix(confluence): drop scripts, styles, and unresolved Jira placeholders from indexed page text

[waleedlatif1]

Summary

• Confluence rendered page bodies now drop <script>/<style> before text extraction. App macros (draw.io, table filter, Markdown) emit a Connect bootstrap script and colored text / table of contents / roadmap emit inline CSS, which were being indexed as page text
• Unresolved Jira issue macros collapse to their issue key. The REST view body only carries the browser placeholder (KEY - Getting issue details... STATUS, localized), detected by the issue-placeholder class; macros Confluence resolved keep their summary and status
• Jira issues-table placeholder shells (header row, loading spinner, refresh settings) are dropped; tables with real rows are untouched
• Callout labeling is unchanged — the view path now runs one cheerio pass (confluenceViewToPlainText) instead of re-serializing and stripping twice
• Hydrated page metadata no longer overwrites the listing's spaceKey/contentType with undefined, and getDocument reports page vs blogpost from the endpoint that answered
• Bumps the view representation marker (view-callouts → view-text-v2) so existing non-scoped Confluence documents rehydrate once with the cleaned text; a page whose only content is an app macro now skips as having no extractable text instead of indexing its bootstrap JS

Type of Change

• Bug fix

Testing

• Captured live Confluence Cloud v2 body-format=view bodies from 502 public pages across 18 sites and diffed old vs new output: callout labels identical on every page; every removed segment is CSS, Connect bootstrap JS, Jira issue placeholder text, or a Jira table placeholder shell; no text added and every Jira key preserved
• Confirmed export_view returns no Jira summary/status either, so reading a different body format would not recover more
• New unit tests built from real markup (localized placeholder, resolved macro, table shell vs real table, macro inside a callout, word boundary around a dropped style, endpoint content type / metadata); each fails with its fix reverted
• Confluence + connector sync suites, lint, type-check, check:audits, docs manifest check

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
docs	Skipped		Sep 14, 2026 11:41pm UTC

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge; no concrete correctness, security, or repository-rule violations remain.

Summary

This PR consolidates rendered Confluence body cleanup into one text-extraction pass and improves hydrated metadata accuracy.

• Removes scripts, styles, unresolved Jira placeholder prose, and Jira loading-table shells from indexed text.
• Preserves resolved Jira details and existing callout labels.
• Retains listing metadata when hydrated fields are absent and records the content type of the endpoint that resolved the document.
• Advances the rendered-content representation marker so existing documents receive the corrected extraction behavior.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  API[Confluence view body] --> Parse[Parse with Cheerio]
  Parse --> Noise[Replace scripts and styles]
  Noise --> Jira[Collapse unresolved Jira placeholders]
  Jira --> Callouts[Rewrite callouts with labels]
  Callouts --> Text[Convert to normalized plain text]
  Text --> Document[Hydrated knowledge document]

Loading

_{Reviews (1) · Last reviewed commit: "fix(confluence): drop scripts, styles, a..."}