MongoDB Launches Public Preview Embeddings, GA Memory Store and 45% Faster Reads
MongoDB introduced Automated Voyage AI Embeddings in public preview and GA LangGraph.js Long-Term Memory Store, adding native embeddings generation and persistent agent memory to its AI data platform. It released MongoDB 8.3 with up to 45% faster reads, 35% writes and added AWS PrivateLink cross-region support.
1. Unified AI Data Platform Expansion
MongoDB highlighted its vision to deliver a unified AI data platform at its London event, combining real-time database, full-text and vector search, memory and embedding models into a single offering. This approach aims to simplify production AI workloads and reduce reliance on disparate systems.
2. Automated Voyage AI Embeddings Preview
Automated Voyage AI Embeddings entered public preview, automatically generating embeddings upon data write or update to enhance retrieval accuracy. These Voyage AI models top the Retrieval Embedding Benchmark, enabling enterprises to deploy semantic search in minutes instead of weeks.
3. LangGraph.js Long-Term Memory Store GA
The LangGraph.js Long-Term Memory Store reached general availability, offering JavaScript and TypeScript developers persistent, cross-conversation agent memory powered by MongoDB Atlas. This integration removes the need for separate databases or synchronization, letting developers focus on AI outcomes.
4. MongoDB 8.3 Performance and AWS PrivateLink
MongoDB 8.3 launched with performance improvements—up to 45% faster reads, 35% writes, 15% more ACID transactions and 30% more complex operations—without code changes. Cross-region connectivity via AWS PrivateLink also became generally available, securing private database traffic across regions for compliance-sensitive deployments.