Marqo Open Source
A self-hosted vector search engine for text and images, with built-in embedding generation, Docker deployment, and an Apache 2.0 license.
Marqo Open Source is a self-hosted search engine for developers building semantic document search, image search, or retrieval for AI applications. It handles embedding generation alongside storage and retrieval, so applications can submit documents without maintaining a separate embedding service. The open-source project is deprecated and no longer receives updates.
It runs in Docker under the Apache 2.0 license and supports CPU or GPU inference. The stated Docker requirements are at least 8 GB of memory and 50 GB of storage. Self-hosting puts inference and search storage on your own infrastructure; Marqo Cloud is a separate managed offering.
Marqo supports PyTorch and Hugging Face models, including CLIP for searching images with text or other images. It also accepts custom models and supports OpenAI embeddings. Local models run on your infrastructure, while the OpenAI option uses an external service.
Search can use semantic similarity or keywords. Weighted queries let applications favor some concepts and reduce matches to others, while metadata filters narrow results. Documents can hold text, images, and structured metadata together. Combined text and image fields let both contribute to a document's relevance score.
Integrations with Haystack and LangChain connect the search engine to question answering and retrieval pipelines. Griptape and Hamilton integrations support agent and LLM applications. For larger collections, Marqo supports horizontal index sharding and asynchronous uploads and searches.
