Weaviate 1.37
#12 en Embeddings & bases vectoriellesWeaviate · v1.37 · depuis April 2026 (laut Blog-Post-Metadaten; GitHub-Patch-Reihe v1.37.x) · 2× · vu le 30 juin 2026
Weaviate 1.37 is a minor release of the open-source vector database Weaviate, published in April 2026. The version expands integration with AI agents and tools: a new native MCP server (preview) enables direct communication with LLMs/IDEs, and Diversity Search via Maximum Marginal Relevance (MMR) was introduced as a preview feature to reduce redundancy in vector search results. Additional new features include extensible tokenizers, query profiling, incremental backups, Gemini audio support in the multi2vec-google module, and the new BlobHash property type. Weaviate remains fully open source (BSD-3-Clause) and is available both self-hosted and as a managed cloud service (Weaviate Cloud).
Fonctionnalités
| Deployment (Self-host/Cloud) | Beides: Self-hosted via Docker/Kubernetes (kostenlos) oder Weaviate Cloud als Managed Service inkl. BYOC |
| Durchsatz/Latenz | Semantische Suche über Milliarden Vektoren im Millisekundenbereich (laut GitHub-Projektbeschreibung) |
| Lizenz | BSD 3-Clause License (Open Source) |
| Plattform | Cloud-native Vektordatenbank, verfügbar als Docker/Kubernetes-Self-Host oder Weaviate Cloud (AWS, GCP, Azure) |
| Preis | Self-hosted: kostenlos (BSD-3-Clause). Weaviate Cloud: Flex ab $45/Monat, Plus ab $280/Monat (Jahresvertrag), Premium auf Anfrage |
| Protokoll-Kompatibilität | REST API, gRPC API und GraphQL API; neu in 1.37: nativer MCP-Server (Model Context Protocol, Preview) |
| Release-Datum | Weaviate v1.37 veröffentlicht (Open Source & Weaviate Cloud), Blog-Beitrag datiert 23. April 2026 |
| Unterstützte Modelle/Provider | Integrierte Vectorizer/Reranker/Generative-Module für OpenAI, Cohere, Google/Gemini, HuggingFace, VoyageAI, NVIDIA u.a.; neu: Gemini-Audio in multi2vec-google |