What do I use for search over text chunks and images together?

Use a multimodal search / vector database setup.

For text chunks + images together, the most common approach is:

1. Embed both modalities into the same vector space

   • Text: use a text embedding model
   • Images: use a multimodal/image-text embedding model like CLIP or a similar model

2. Store the embeddings in a vector database

   • Examples: Pinecone, Weaviate, Milvus, Qdrant, Elasticsearch/OpenSearch with vector search

3. Search with a text query or image query

   • Convert the query to an embedding
   • Retrieve the nearest neighbors across both text and images

Good options depending on your goal

• Best general answer:
  CLIP + vector database

• If you want one system for text, images, and metadata:
  Weaviate or Qdrant with multimodal embeddings

• If you need enterprise search with hybrid keyword + vector:
  Elasticsearch/OpenSearch

• If you want simple developer-managed vector search:
  Pinecone or Qdrant

Important note

If you want cross-modal search like:

• “find images related to this paragraph”
• “find text describing this image”

then you need a shared embedding model such as CLIP.
If you only embed text and images separately with unrelated models, they won’t be directly comparable.

If you want, I can recommend a specific stack based on whether you’re building in Python, JavaScript, or a cloud platform.