DocumentDB for AI: Open-source vector search for SQL and .NET developers

DocumentDB is back. The new open-source project is a document database built on PostgreSQL, with native BSON support and MongoDB-compatible APIs.

We’ll run DocumentDB locally with Docker and connect from C# using the MongoDB .NET driver. The code will cover CRUD operations, indexes, document queries, and aggregation pipelines.

SQL Server developers will see how tables, rows, relationships, and indexes map to collections, documents, embedded data, and MongoDB-style indexes. MongoDB developers will get a look at the PostgreSQL foundation underneath the API.

Then we’ll put the vector features to work. We’ll generate embeddings, store them beside the source text and metadata, and retrieve documents by semantic similarity. Those results will become the retrieval layer for a small RAG application written in .NET.

We’ll also compare the available vector indexes: HNSW and IVF in the open-source project, plus DiskANN in Azure DocumentDB. You’ll leave knowing where DocumentDB fits beside SQL Server, MongoDB, and a dedicated vector database, with enough C# to start testing it yourself.

Azure DocumentDB - https://azure.microsoft.com/en-us/products/documentdb