Question: Is there a database that allows me to perform hybrid search and metadata filtering on large collections of vectors?

If you need a database that can handle hybrid search and metadata filtering on a big vector database, Pinecone is worth a look. It's designed for low-latency vector search, metadata filtering and real-time updates, with a serverless design that means you don't have to worry about database scaling. Pinecone supports keyword boosting in hybrid search, and it has a variety of pricing plans including a free starter tier.

Another contender is Milvus, an open-source vector database that's optimized for high-dimensional vector search. It's got metadata filtering, hybrid search and support for multiple vectors. It's designed to scale up to tens of billions of vectors, and it's good for image search, recommender systems and anomaly detection.

Vespa is another contender, a unified search engine and vector database that can handle both vector and lexical search. It's designed to let you combine the two searches in a single query, which is useful for AI applications like recommendation systems and generative AI. The service has auto-elastic data management and scalable machine-learned model inference for high performance and low latency.

If you want a managed vector database service, Zilliz is worth a look. Based on Milvus, Zilliz is tuned for large-scale vector data and has fast vector retrieval speeds. It's got high availability, scalability and support for multiple cloud platforms, so it's easy to run and manage complex vector search applications without worrying about lots of infrastructure.