Add PostgreSQL Index Profile Support

[MikeAlhayek]

Now we can create IndexProfile using PostgreSQL in additional to Elasticsearch and AzureAISearch!

This pull request introduces first-class support for PostgreSQL (with pgvector) as a vector search backend for both core and AI-powered features in the CrestApps.Core ecosystem. It adds new primitives, updates dependency management, and provides comprehensive documentation and configuration guidance. The PostgreSQL provider is now available as an alternative to Elasticsearch and Azure AI Search for RAG (retrieval-augmented generation) and indexing scenarios.

Key changes include:

PostgreSQL Provider Integration:

• Added new projects: CrestApps.Core.PostgreSQL and CrestApps.Core.AI.PostgreSQL, enabling PostgreSQL + pgvector support for vector search, AI documents, memory, and data sources, with all required service registrations under the "PostgreSQL" provider name. [1] [2] [3]
• Registered PostgreSQL-specific services (ISearchIndexManager, ISearchDocumentManager, IDataSourceContentManager, IDataSourceDocumentReader, IODataFilterTranslator) and AI extensions (AddAIDocuments(), AddAIDataSources(), AddAIMemory()) for seamless integration into MVC and Blazor hosts. [1] [2]

Dependency and Package Management:

• Added Npgsql and Pgvector NuGet packages for PostgreSQL and vector support, and included Aspire.Hosting.PostgreSQL for streamlined local development and deployment using .NET Aspire. [1] [2]

Documentation and Developer Experience:

• Introduced a detailed PostgreSQL provider documentation page, including setup instructions, configuration samples, Docker Compose for local development, and troubleshooting. The new doc is linked in the sidebar navigation. [1] [2]
• Updated the v1.0.0 changelog to reflect the addition of PostgreSQL/pgvector as a supported backend for vector search and AI features.

[MikeAlhayek]

@copilot fix build

[Copilot]

@copilot fix build

Build is fixed in commit a39db47 by replacing the non-static NpgsqlCommandBuilder.QuoteIdentifier(...) call with PostgreSQLHelpers.QuoteIdentifier(...) in PostgreSQLSearchIndexManager.DeleteAsync.

[vlad2kgithub]

@MikeAlhayek, are Elasticsearch and AzureAISearch not transactional databases like PostgreSQL? What am I missing?

[MikeAlhayek]

@vlad2kgithub Yes — PostgreSQL is typically the transactional system of record, while the other two are primarily search/indexing engines used for retrieval.

In our case, we’re using PostgreSQL to also store vector data for AI features. It’s not replacing a search engine; it’s simply serving as an additional persistence layer for AI-related embeddings alongside the rest of the application data.