The MongoDB team has announced the general availability release of Atlas Vector Search and Atlas Search Nodes.
MongoDB is a NoSQL document database that stores its documents in a JSON-like format with schema. MongoDB Atlas is the fully-managed cloud database from the MongoDB team.
Atlas Vector Search is a new tool that can be used to search unstructured data. It is accessed via the MongoDB Query API, and can be used by developers to build applications that use semantic search and generative AI, and can provide results when users are vague about what they're looking for.
It uses machine learning models to find results that are similar to the examples given by the user. Developers can create vector embeddings with machine learning models like OpenAI and Hugging Face, and store and index them in Atlas for retrieval augmented generation (RAG) and semantic search.
Semantic search is the process of searching and finding relevant results from unstructured data, based on semantic similarity, while Retrieval augmented generation (RAG) combines the use of LLMs with feeds of your own, real-time data.
The second component is a new Search Nodes dedicated infrastructure that underpins Atlas Search and Vector Search workloads, so developers can optimize compute resources. The MongoDB team says in some cases they've seen 60% faster query time for some users' workloads when they make use of concurrent querying in Search Nodes. The GA release includes a memory-optimized, low CPU option that is optimal for Vector Search in production.
Until now, Atlas Search has been co-located with Atlas on Atlas Nodes in the MongoDB cluster. Having both the standard and vector search co-located on Atlas Nodes and clusters meant users needed to size their workload based on both Search and Database requirements. Search nodes are located separately in their own cluster, so their workload is isolated.
Vector Search and Search Nodes are generally available now.
IBM has released new Granite models that it says provide state-of-the-art performance relative to model size. The Granite 3.0 collection includes a new, instruction-tuned, dense decoder-only LLM.
After early indications that AI-powered coding assistants increase developer productivity, findings are emerging that challenge this and suggest code quality may be reduced by their use. At the same t [ ... ]
