Major upgrade: MyScale now features text search functionality!
Major upgrade: MyScale now features
text search functionality!
Major upgrade: MyScale now features text search functionality!
Major upgrade: MyScale now features
text search functionality!

AI Character Chat

Using MyScale to build an AI chatbot and provide users with personalized chat services

Your Requirements

  • Isolate multi-tenant data, ensuring data privacy and preventing unauthorized access or leakage between tenants.
  • Handle a large number of concurrent requests and scale flexibly to accommodate growth as user numbers increase.

Our Solution

MyScale is a database service fully hosted on the AWS cloud platform. Its architecture deeply integrates multiple AWS services, including Amazon EC2, Amazon EKS, Amazon S3 and AWS Elastic Load Balancer.

The architecture design of the MyScale cloud service includes three levels: the global control plane, the regional control plane, and the regional data plane, each corresponding to a Kubernetes cluster. All MyScale services are deployed on Amazon EKS. Amazon EKS provides a highly available, secure, and scalable Kubernetes environment, enabling MyScale to fully leverage Kubernetes's powerful features, such as service discovery, load balancing, auto-scaling, and security isolation. With the Cluster Autoscaler on Amazon EKS, MyScale can rapidly launch, stop, and scale instances according to user workload demands, scaling the Amazon EKS node pools up or down.

MyScale has chosen Amazon EC2 instances equipped with NVMe-based local SSD disks to deploy the MyScale database. Unlike most vector databases that choose pure in-memory HNSW vector index algorithms, MyScale's self-developed MSTG algorithm allows vector data to be cached on local NVMe SSD disks. Therefore, while providing high-performance vector search for users, MyScale significantly reduces memory usage.

Moreover, MyScale has achieved deep integration with Amazon SageMaker. MyScale leverages the open-source embedding model Jina from SageMaker, enabling efficient vector embedding and similarity computation. Additionally, MyScale has adopted the open-source llama3 model from SageMaker, which understands and processes user natural language input, converting it into SQL queries, achieving seamless integration of semantic search. With its cost-effective data management support, exceptional linear scalability, and standard SQL support, MyScale eliminates the need for multiple expensive data warehouse products requiring different query languages.

MyScale SaaS service is deployed on AWS, it offers secure and scalable user data management features, allowing for the storage and management of user profiles, preferences, and other personal information. By integrating essential features such as filter search and RBAC (Role-Based Access Control), MyScale ensures efficient data retrieval and scalability while safeguarding user privacy.

MyScale is designed for high performance and scalability, capable of handling concurrent requests and scaling seamlessly as user demand grows. Its distributed architecture allows for horizontal scaling across multiple nodes, ensuring optimal performance even under heavy loads.

Case Study

Spicychat AI: Using an SQL Vector Database for Large-Scale Role-Playing Chat

SpicyChat AI offers personal users a chatbot service. Here, users can engage in free conversations with a variety of interesting characters such as anime characters, movie stars, and deceased scientists.


Spicychat has numerous tenants with corresponding personal user data, including their own conversation history or character settings. During conversations, it should be ensured that data between users cannot be queried.


For Spicychat AI, MyScale provides each user with a unique long-term memory, and support large-scale multi-tenancy through data partitions and primary keys. By isolating user data, implement proper permission management or skip queries at the database level, MyScale ensures users' personal privacy while saving query resources.

Service Used

Amazon EKS, Amazon EC2, Amazon S3


10x less memory consumption

10x more likely to cost savings