⚡ TL;DR Embeddings let users find content by meaning. Gemini Embedding 2 (default) is multimodal; text-embedding-3-small suits volume and -large suits accuracy.
Search that understands what people meant, not just the words they typed.
🔎 Uses
Semantic search over documents, FAQ bots answering from help content, and internal knowledge bases.
🧬 Embedding Models
Options:
- ✅ Gemini Embedding 2: multimodal default
- ✅ Gemini Embedding 001: legacy text
- ✅ text-embedding-3-small: cost-efficient
- ✅ text-embedding-3-large: most accurate
🛠️ Building It
Describe the feature in chat; Lovable writes the indexing and query backend, then publish.
🤝 Pairing With Chat
Combine search with a chat model for answers grounded in your own content.
💡 Pro tip Do not switch embedding models later without re-indexing; vectors are not compatible.
Need a knowledge base built? Lord of the Wix can help. ✨
❓ Lovable AI search FAQ
What is the default embedding model?
Gemini Embedding 2.
Is it multimodal?
Yes, Gemini Embedding 2 is.
Which is most accurate?
text-embedding-3-large.
Can it power RAG?
Yes.
When should Embedding 001 be used?
Only for existing data.
🙏 Credit: Based on public information from the Lovable documentation. Lovable is a product of Lovable. Source: docs.lovable.dev. Lord of the Wix is an independent agency and is not affiliated with Lovable.