Weaviate
Weaviate is an open-source, schema-based vector database that stores objects together with embeddings and supports search via vector similarity, keywords (BM25) and a hybrid of both.
Weaviate is an open-source vector database with hybrid search: it combines keyword and semantic search in a single query. It can be self-hosted or used as a service — which makes it the obvious choice when data sovereignty is up for negotiation.
Explanation
The practical difference from purely semantic search shows up with proper nouns, part numbers and technical terms: semantic search finds similar things, keyword search finds exactly the thing. Weaviate blends both with weighting in one query. Because the project is open source, the same database can run in your own data centre, in an EU cloud or as a managed service — the application stays the same, only operations change.
Marketing Relevance
For GDPR-driven projects Weaviate is often the only realistic option besides Qdrant, because self-hosting is demonstrably possible. For marketing knowledge bases hybrid search is the real gain: brand names and product codes are regularly lost in purely semantic search.
Example
A knowledge base holds campaign guidelines and product names. A search for a model designation hits exactly via the keyword portion, the question "what may we say about pricing?" via the vector portion. Both in one query, with configured weighting.
Common Pitfalls
Self-hosting is not a side project: backups, index sizes, memory requirements and upgrades stay your responsibility. And configuring vectorisation as a database module moves the model choice there — switching embedding models then forces a full reindex.
Origin & History
Weaviate started as an open-source project from 2019 and grew strongly with the 2023 RAG wave. With modules for vectorisation, reranking and generative answers, part of the application logic moved into the database — a pattern worth limiting deliberately.
Further Resources
Marketing Use Cases
Hybrid search across technical content with product names and codes.
Self-hosted RAG knowledge base in an EU environment.
Multilingual knowledge base with separate collections per language.
Frequently Asked Questions
How does Weaviate differ from Qdrant?
Both are open source and self-hostable. Weaviate leans more on schema, modules and hybrid search, Qdrant on a lean Rust core with fine-grained filtering. The decision usually comes down to operating model and team experience, not search quality.
Do I need two systems for hybrid search?
No — that is the point. Keyword and semantic search run in the same query instead of merging results from two systems afterwards.
Related Services
Go deeper: Measurement & attribution · Model comparison 2026