Botkibble

The plugin only serves posts and pages by default. To add a custom post type, use the botkibble_served_post_types filter in your theme or a custom plugin: add_filter( 'botkibble_served_post_types', function ( $types ) { $types[] = 'docs'; return $types; } ); Be careful. Only add post types that contain public content. Do not expose post types that may contain private or sensitive data (e.g. WooCommerce orders).

Every response starts with a YAML block containing:

• title — the post title
• date — publish date in ISO 8601 format
• type — post type (e.g. post, page)
• word_count — word count of the Markdown body
• char_count — character count of the Markdown body
• tokens — estimated token count (word_count × 1.3)
• categories — array of category names (posts only)
• tags — array of tag names (posts only, omitted if none)

Use the botkibble_frontmatter filter: add_filter( 'botkibble_frontmatter', function ( $data, $post ) { $data['excerpt'] = get_the_excerpt( $post ); return $data; }, 10, 2 );

Use the botkibble_content_signal filter: add_filter( 'botkibble_content_signal', function ( $signal, $post ) { return 'ai-train=no, search=yes, ai-input=yes'; }, 10, 2 ); Return an empty string to omit the header entirely.

Yes, use the botkibble_token_multiplier filter. The default multiplier of 1.3 (word count × 1.3) comes from Cloudflare's implementation: add_filter( 'botkibble_token_multiplier', function () { return 1.5; // Adjusted for a different model's tokenizer } );

Cache misses (when a post needs to be converted) are limited to 20 per minute globally. You can change this with the botkibble_regen_rate_limit filter: add_filter( 'botkibble_regen_rate_limit', function () { return 50; } );

Yes, use the botkibble_clean_html filter. This runs after the default cleanup and before conversion: add_filter( 'botkibble_clean_html', function ( $html ) { // Remove a plugin's wrapper divs $html = preg_replace( '/ (.*?)<\/div>/s', '$1', $html ); return $html; } );

Yes. Botkibble keeps converter node removal disabled by default (for backward compatibility), but you can opt in with botkibble_converter_remove_nodes: add_filter( 'botkibble_converter_remove_nodes', function ( $nodes ) { $nodes = is_array( $nodes ) ? $nodes : []; $nodes[] = 'script'; return $nodes; } ); If you also need application/ld+json, extract it in botkibble_clean_html first, then let converter-level script removal clean up any remaining script tags.

Use the botkibble_body filter. This is the best place to add content like ld+json that you want included in the word count and token estimation: add_filter( 'botkibble_body', function ( $body, $post ) { $json_ld = '...'; return $body . "\n\n" . $json_ld; }, 10, 2 );

Use the botkibble_output filter to append or modify the text after conversion: add_filter( 'botkibble_output', function ( $markdown, $post ) { return $markdown . "\n\n---\nServed by Botkibble"; }, 10, 2 );

Yes. Add ?botkibble_variant=slim when requesting Markdown to generate and serve a separate cached file. To ensure your variants are invalidated on post updates, return them from the botkibble_cache_variants filter: add_filter( 'botkibble_cache_variants', function ( $variants, $post ) { $variants[] = 'slim'; return $variants; }, 10, 2 );

Yes, if you only want to serve Markdown via explicit URLs (.md or ?format=markdown), use the botkibble_enable_accept_header filter: add_filter( 'botkibble_enable_accept_header', '__return_false' );

It works with the most common structures (post name, page hierarchy). Complex date-based permalink structures may require the query parameter or Accept header method instead.

They return a 403 Forbidden response. There's no point serving a password form to a bot.

• Content-Type: text/markdown; charset=utf-8
• Vary: Accept — tells caches that responses vary by Accept header
• X-Markdown-Tokens: <count> — estimated token count (word_count × 1.3)
• X-Robots-Tag: noindex — prevents search engines from indexing the Markdown version
• Link: <url>; rel="canonical" — points search engines to the original HTML post
• Link: <url>; rel="alternate" — advertises the Markdown version for discovery
• Content-Signal: ai-train=no, search=yes, ai-input=yes — see contentsignals.org