Marqo
Multimodal vector search engine with embedding inference built inside the database.
About
Marqo is an open-source, multimodal vector search engine that indexes text and images, generates embeddings on the fly, and serves tensor, lexical, and hybrid search through a single REST API. It bundles model inference (Sentence Transformers, OpenCLIP, ONNX) inside the engine, so one docker run produces a working semantic search stack without a separate embedding service.
Important caveat: the Apache 2.0 open-source engine has been marked deprecated by the maintainers as Marqo pivots to a hosted ecommerce search product. The project remains widely forked, downloaded, and self-hosted, but new adopters should evaluate the hosted offering or community forks for long-term support rather than assuming the OSS engine will keep evolving.
What it can do
- Server-side embedding inference (text, image, multimodal)
- Tensor, lexical, and hybrid search in one API
- Ecommerce-tuned embedding models
- Single-container deployment
Setup
docker run marqoai/marqo; or sign up for the hosted cloud product at cloud.marqo.ai.
Prerequisites
Docker for self-hosting; Python for the marqo client library.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 63; Trending score 68/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.