A transactional, relational-graph-vector database that uses Datalog for query, designed as the hippocampus for AI.
CozoDB is a transactional, multi-model database that integrates relational, graph, and vector data, using Datalog as its query language. It solves the problem of managing interconnected and high-dimensional data in a single system, enabling complex queries like recursive graph traversals and vector similarity searches seamlessly. It is designed to be embeddable across environments—from browsers to servers—while scaling for performance and concurrency.
Developers building AI applications, graph analytics, or data-intensive systems that require unified relational, graph, and vector operations in an embeddable or scalable database.
Developers choose CozoDB for its unique combination of Datalog's expressiveness, multi-model flexibility, and embeddability—offering graph algorithms, vector search, and time travel in a single transactional database without sacrificing performance.
A transactional, relational-graph-vector database that uses Datalog for query. The hippocampus for AI!
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Uses a powerful Datalog dialect for recursion and composability, enabling complex graph operations like reachability and shortest paths naturally, as demonstrated in the tutorial examples.
Unifies relational, graph, and vector data in a single database with integrated HNSW vector search, allowing seamless joins and searches across data types without separate systems.
Runs embedded in browsers via WASM, on mobile devices, and servers with multiple storage engines like RocksDB and SQLite, supporting diverse environments without setup.
Offers optional immutable history tracking per relation for querying past data states, useful for auditing or analysis, as highlighted in the feature list.
Versions before 1.0 do not promise syntax/API stability or storage compatibility, risking breaking changes and making it unsuitable for production-critical applications without thorough testing.
Requires learning Datalog, a niche query language compared to SQL, which can increase onboarding time and limit adoption in teams familiar with traditional databases.
Has a smaller community and fewer third-party tools, libraries, and documentation compared to established databases, potentially increasing development effort for integrations.
Supports multiple backends like RocksDB that require tuning via options files, and misconfiguration can lead to performance issues or corruption, as noted in the tuning section.