July 3, 2024 in Product

8.7 Release Updates for SingleStore Kai™

Spandana Rachamadugu

Product Manager

Manish Kumar

Principal Product Manager

We are thrilled to announce the latest advancements for SingleStore Kai with the 8.7 release. Since the initial GA release with version 8.5, SingleStore Kai has continuously evolved, introducing powerful new capabilities designed to deliver ultra-fast performance for your applications built on MongoDB®.

8.7 Release Updates for SingleStore Kai™

With this release, we've added support for building intelligent applications through vector search, increased compatibility with the MongoDB command set and further optimized query performance.

Vector search with SingleStore Kai

SingleStore Kai enables NoSQL/MongoDB apps to harness the robust vector search capabilities of the SingleStore database seamlessly through any MongoDB client or driver. With the commands $vectorSearch and createIndexes, you can easily build gen AI capabilities into your applications by performing an Approximate Nearest Neighbor (ANN) vector search with a variety of index building options like IVF_PQ and HNSW. Follow this sample notebook to try out $vectorSearch.

'$vectorSearch': {
               "index": "vector_index_name",
               "path": "emb_field",
               "queryVector": query_vector,
               "limit": 3,
           }

SingleStore Kai works seamlessly with integrations like LangChain and LlamaIndex without any code changes. You can now build powerful RAG applications by integrating analytics, full-text search and joins with multi-model data to provide richer context to LLM models — without any performance trade-offs.

Query performance improvements

The latest updates make processing massive data volumes faster and more efficient, especially for transactional operations like document lookups and updates. Optimizations such as applying column group indexing on SingleStore Kai tables now accelerate full-document retrievals.

The BSON storage type — along with BSON function built-ins — are specifically optimized for SingleStore’s columnstore, enabling analytical queries on billions of rows with minimal latency and high concurrency. Moreover, functions previously supported as WASM modules are now seamlessly integrated as native engine functions.

Arrays in MongoDB documents are widely used to store one-many mappings. SingleStore Kai's new multi-key index feature accelerates queries filtering individual array values. Selective queries involving arrays are now nearly 2x faster. With the introduction of vectorized execution on array aggregations, you can achieve performance on aggregation queries over array fields on par with non array fields.

CDC out

Change Data Capture (CDC) in SingleStore allows you to monitor and capture changes made to a database in real time. It allows downstream consumers to receive a near real-time, immutable stream of changes made to a SingleStore database. You can enable change streams using your Mongo clients, and read changes through watch command.

const changeStreamOptions = { fullDocument: 'updateLookup' };
const cursor = col.watch([], changeStreamOptions)
col.insertOne({_id: 1, a: 1});

Seamless user experience

Our native CDC capabilities make data migration and replication from MongoDB effortless. You can set up CDC directly through the MongoAPI, SQL commands or now intuitively via our portal UI with just a few clicks. The Kai shell allows you to interact with your workspace directly, without needing external clients. It has been updated to improve usability, featuring hotkeys for easier command navigation.

As we continue to enhance compatibility and power for applications built on MongoDB, we are excited to see how you leverage SingleStore Kai to unlock innovative use cases. Experience the power of it today by activating your Free Starter Workspace, and start building your intelligent applications..

Product Engineering