EasyAppDev.Blazor.AutoComplete.AI.PostgreSql

PostgreSQL/pgvector integration for semantic search with the Blazor AutoComplete component.

Features

• pgvector Support: Native PostgreSQL vector similarity search
• 6 Distance Functions: Cosine, L2, DotProduct + 3 pgvector-only
• HNSW Index: High-performance approximate nearest neighbor search
• Self-Hosted: Full control over your data and infrastructure

Installation

dotnet add package EasyAppDev.Blazor.AutoComplete.AI.PostgreSql

Prerequisites

1. PostgreSQL 15+ with pgvector extension:

CREATE EXTENSION vector;

Quick Start

1. Configure Services

// Program.cs builder.Services.AddAutoCompletePostgreSql<Product>( connectionString: "Host=localhost;Database=myapp;Username=user;Password=pass", options => { options.TableName = "product_embeddings"; options.Dimensions = 1536; options.DistanceFunction = DistanceFunction.Cosine; }, textSelector: p => $"{p.Name} {p.Description}", idSelector: p => p.Id.ToString()); // Register embedding generator builder.Services.AddAutoCompleteVectorSearch<Product>( openAiApiKey: "sk-...");

2. Use the Component

@using EasyAppDev.Blazor.AutoComplete.AI <VectorAutoComplete TItem="Product" TextField="@(p => p.Name)" @bind-Value="@selectedProduct" Placeholder="Semantic search..." />

Configuration Options

Option	Description	Default
ConnectionString	PostgreSQL connection string	Required
TableName	Table for embeddings	{type}_embeddings
Dimensions	Vector dimensions	1536
DistanceFunction	Similarity metric	Cosine
CreateTableIfNotExists	Auto-create table	true
UseHnswIndex	Enable HNSW index	true

Distance Functions

Function	Use Case
Cosine	Normalized embeddings (default)
L2	Euclidean distance
DotProduct	Inner product similarity
L1	Manhattan distance
Hamming	Binary vectors
Jaccard	Set similarity

Indexing Data

@inject IVectorIndexer<Product> indexer // Index all products await indexer.IndexAsync(products); // Re-index on update await indexer.UpdateAsync(updatedProduct);

License

MIT