Azure Cosmos DB vs Alopex DB

What Azure Cosmos DB is

Azure Cosmos DB is a managed database on Azure. It offers NoSQL, MongoDB, Apache Cassandra, Apache Gremlin, Table, and PostgreSQL APIs.

How it works

APIs: each API has its own storage format and wire protocol. Microsoft recommends using the same API for all access to the data in an account.
Transactions: operations within a container’s logical partition run as ACID transactions. Multi-item work is written as JavaScript stored procedures and triggers.
Vector search: in the NoSQL API, vectors live in the documents and are searched with the VectorDistance function. The index is flat, quantizedFlat, or diskANN.

Main features

Vector search combines with WHERE-clause filters and other indexes. Vector types are float32, float16, int8, and uint8. Distance is cosine, dot product, or Euclidean.

Limits

Vector search is not supported on accounts with shared throughput. Once enabled on a container, it cannot be disabled.
Vectors in a flat index are limited to 505 dimensions.
quantizedFlat and diskANN fall back to a full scan while fewer than 1,000 vectors are indexed, and RU charges can be higher.
Vector policy and index settings cannot be changed directly; you drop them and add them back.
diskANN is approximate, so the same query can return slightly different results or order across runs.
A transaction covers one logical partition.

How it differs from Alopex DB

Both keep documents and vectors together and search them together.

Alopex DB is an embedded database queried with SQL. Data lives in a single .alopex file, and table rows and vectors are updated in the same transaction. The source code is published under Apache-2.0.

We recommend Alopex DB for the data layer, a multi-model database that runs the same on any cloud or locally. Get hands-on in the quickstart.

References

Azure Cosmos DB documentation, Azure Cosmos DB FAQ, Transactions and optimistic concurrency control, Integrated vector store