# Vector Store

## Description

Vector store providers register through `IFlexVectorStore`. Your application code stays provider-agnostic while the selected provider handles storage, indexing, and similarity search.

## Infrastructure arrangement

When you add a vector store provider through **Flex Studio**, Flex generates the provider infrastructure and configuration:

1. Provider files under `Infrastructure/Providers/{YourApplication}.DataStoreProviders/Vector/{Provider}/`
2. Default configuration under `Application/EndPoints/{YourApplication}.EndPoint.CommonConfigs/AppSettings/DataStores/Vector/{Provider}.json`

Flex auto-wires generated Queries/Handlers/Plugins that use `IFlexVectorStore`—you only register the provider.

## Configuration in DI

Register only the provider vector store you need.

```csharp
// Example: Azure AI Search (options-bound provider)
var section = configuration.GetSection("FlexBase:DataStores:Vector:AzureAISearch");
services.AddFlexAzureAISearchVectorStore(options => section.Bind(options));

// Example: SQL Server / pgvector (parameter-based providers)
// var section = configuration.GetSection("FlexBase:DataStores:Vector:SqlServer");
// services.AddFlexSqlServerVectorStore(
// 	connectionString: section.GetValue<string>("ConnectionString")!,
// 	tableName: section.GetValue<string>("TableName")!,
// 	dimensions: section.GetValue<int?>("Dimensions") ?? 1536);
```

## appsettings.json

Vector store configuration is read from `FlexBase:DataStores:Vector:<Provider>`.

```json
{
  "FlexBase": {
    "DataStores": {
      "Vector": {
        "AzureAISearch": {
          "Endpoint": "https://your-search-service.search.windows.net",
          "ApiKey": "<store-in-secrets>",
          "IndexName": "your-index-name",
          "Dimensions": 1536
        }
      }
    }
  }
}
```

## Sample usage ({YourApplication})

```csharp
public class KnowledgeIndex
{
	private readonly IFlexVectorStore _vectorStore;

	public KnowledgeIndex(IFlexVectorStore vectorStore)
		=> _vectorStore = vectorStore;

	public Task UpsertAsync(string id, float[] embedding, object? metadata, string content)
		=> _vectorStore.UpsertAsync(id, embedding, metadata, content);
}
```

## Provider pages

* SQL Server: `vector-store/sql-server-vector.md`
* Azure AI Search: `vector-store/azure-ai-search.md`
* pgvector (PostgreSQL): `vector-store/pg-vector-postgres.md`
* OpenSearch: `vector-store/open-search.md`

## Provider considerations

* `Dimensions` must match the embedding model you use.
* Keep secrets (API keys/connection strings) out of source control.
* For hybrid retrieval, combine Vector Store (semantic) + Full-Text Search (keyword/filtering).