Skip to content

Metadata & Indexing Configurations

Hibiscus automatically extracts, tracks, and manages metadata from your workspace files to build a rich, connected knowledge base. This includes handling relationships, tags, and explicitly ignoring specific files.

Automated Metadata Extraction

During the indexing pipeline, Hibiscus parses your Markdown files to build a persistent, localized metadata store. This metadata feeds directly into the graph visualization and search systems.

Tags

Hibiscus automatically detects and indexes #tags within your content. Tags must use alphanumeric characters, hyphens, or underscores. These are aggregated to support tag-based filtering in the search and graph interfaces.

Bidirectional relationships are established via [[wiki-link]] syntax. The indexer captures these links and resolves them against file paths to build the underlying knowledge graph.

Aliases (Frontmatter)

If your Markdown files utilize YAML frontmatter, the indexer parses the aliases array.

---
aliases: [AI, Machine Learning]
---

These aliases are extracted and associated with the file's chunk data, ensuring search queries match the document even if the primary filename is different.

As wiki-links are processed, the system automatically builds a persistent backlinks.json mapping. This allows the Right Panel UI to display a list of all documents that reference the currently active file in O(1) time.

Workspace Exclusions (.hibiscusignore)

You can prevent Hibiscus from indexing specific files or directories by creating a .hibiscusignore file in the root of your workspace.

Behavior

  • The .hibiscusignore file follows simple substring matching logic.
  • Adding a directory name (e.g., node_modules/ or secrets/) will prevent the indexer from scanning any files within those paths.
  • Adding a specific filename (e.g., private_notes.md) will exclude that exact file.
  • Ignored files are tracked and reported in the Knowledge Dashboard under the "Skipped" count.

Note: The .hibiscus internal state directory is always ignored by default to prevent recursive indexing loops.

Large File Management

To preserve memory efficiency and prevent UI stuttering, the background worker enforces a strict large file threshold.

  • Threshold: Any file exceeding 10MB is automatically bypassed.
  • Reporting: Bypassed files are logged to skipped_files.json and visibly reported in the Knowledge Dashboard.
  • Handling: These files will not appear in search results, nor will their contents be processed for wiki-links or tags.