VecPost AI Search for Posts

VecPost AI Search for Posts replaces WordPress’s default SQL LIKE search with vector-based semantic search. Instead of matching exact words, it understands the meaning of a search query.

Example: A user searching “heart workouts” will find your post titled “Best cardiovascular exercises” – even though no words overlap – because the meanings are similar.

How It Works

1. When you publish a post, the plugin sends its content to your chosen AI provider (OpenAI or Google Gemini) to generate a vector embedding – a list of numbers that represents the meaning of the text.
2. These numbers are stored in your database.
3. When a user searches, their query is also converted to numbers, and the plugin finds posts whose numbers are closest – meaning most semantically similar.

Features

• Semantic search powered by OpenAI (text-embedding-3-small or text-embedding-3-large) or Google Gemini (gemini-embedding-001)
• Hybrid re-ranking: combines semantic similarity with keyword matching for best results
• Gutenberg block and shortcode [vecpost_semantic_search] for easy placement
• Bulk indexer with progress bar for existing posts
• WP-CLI support: wp vecpost-semantic-search index, wp vecpost-semantic-search status, wp vecpost-semantic-search search "query"
• Configurable scoring thresholds via Settings -> VecPost – AI Semantic Search for Posts
• Automatic re-indexing when you switch embedding models
• Results cached via WordPress object cache (Redis/Memcached compatible)

Third-Party Services

This plugin sends post content to external AI APIs to generate embeddings. By using this plugin, you agree to the terms of service and privacy policies of your chosen provider:

• OpenAI: https://openai.com/policies/privacy-policy | https://openai.com/policies/terms-of-use
• Google Gemini: https://policies.google.com/privacy | https://ai.google.dev/terms

No data is sent without your API key being configured. Data is only transmitted when posts are published or during bulk indexing.

Performance Note

Semantic search requires loading all embeddings into PHP memory for comparison. This works well for sites with up to approximately 1,500 posts. For larger sites, a dedicated vector database (pgvector, Qdrant, or Pinecone) is recommended.