Reindex Content

Rebuild the content index by scanning all content files, extracting frontmatter, and writing .content-ops/content-index.json. No MCP, no embeddings, no API calls.

When to use

• After initial plugin setup (first-time index)
• After bulk content edits or migrations
• After creating or modifying articles or glossary entries
• When the user explicitly asks to reindex

Procedure

1. Read config from .content-ops/config.md:

   • content_types — each type's path (e.g., article.path, glossary.path)
   • languages — list of language codes to scan
   • content_index_path

2. Scan content directories for each content type and language:

   • Use Glob to find all .md and .mdx files under {type.path}/{lang}/
   • For each file: Read it and extract frontmatter only (the YAML between --- delimiters). Do not include the body.

3. Build the index structure — use the stable minimal schema for every entry (all content types):

   Per entry:

   • slug: filename without extension
   • path: relative path to file (for traceability and linking edits)
   • type: content type key (e.g. article, glossary)
   • lang: language code from directory
   • title: articles → frontmatter title; glossary → frontmatter term
   • excerpt: articles → frontmatter excerpt; glossary → frontmatter definition (first 1–2 sentences)
   • tags: from frontmatter if present (default [])
   • translationKey: from frontmatter if present (omit if absent)

4. Write the JSON file to content_index_path from config:

{
  "generated_at": "<ISO 8601 timestamp>",
  "config_fingerprint": {
    "default_language": "<from config>",
    "languages": ["<from config>"],
    "content_types": { "article": { "path": "..." }, "glossary": { "path": "..." } }
  },
  "content": {
    "article": { "en": [...], "it": [...] },
    "glossary": { "en": [...], "it": [...] }
  }
}

5. Ensure the directory exists: If .content-ops/ does not exist, create it before writing.

6. Report results: List how many items were indexed per type and language.

Important notes

• This process only reads files and writes JSON. No external APIs or MCP tools.
• Items with missing or invalid frontmatter should be reported but should not stop the overall process.
• Re-indexing is idempotent — running again overwrites the index with current content.