Modern applications are becoming increasingly complex.
A typical AI application often relies on multiple systems:
- PostgreSQL for relational data
- Redis for caching
- Kafka for event streaming
- Elasticsearch for keyword search
- A vector database for semantic search
While each tool excels at its job, integrating and operating them together introduces complexity—multiple deployments, data synchronization, increased latency, and higher infrastructure costs.
The Idea
I started building CogniDB to explore a different approach.
Instead of combining several specialized systems, my goal is to create a unified engine that supports:
- Relational SQL
- Vector search
- Full-text search
- Event streaming
- In-memory caching
- AI-native functions
all within a single database.
Technology Stack
The project is being built using:
- Java 21
- Netty
- LSM-based storage engine
- HNSW vector indexing
- Raft consensus
- Spring Boot integration
Why Build Another Database?
Not because existing databases are bad.
PostgreSQL, Redis, Kafka, Elasticsearch, and modern vector databases are excellent products.
CogniDB is an experiment in combining the strengths of these technologies into a unified architecture optimized for AI-powered applications.
Current Focus
The roadmap currently includes:
- SQL engine
- Transaction support
- Vector indexing
- Distributed clustering
- AI-native query capabilities
- Spring Boot starter
Looking for Contributors
This is an open-source project, and feedback is always welcome.
If you're interested in storage engines, distributed systems, Java, or AI infrastructure, I'd love to collaborate.
Top comments (0)