Configuration Methods

SmartRAG offers two configuration methods:

Quick Setup (Recommended)

Configure SmartRAG in your Program.cs or Startup.cs:

For Web API Applications:

using SmartRAG.Extensions;
using SmartRAG.Enums;

var builder = WebApplication.CreateBuilder(args);

// Simple one-line configuration
builder.Services.AddSmartRag(builder.Configuration, options =>
{
    options.StorageProvider = StorageProvider.InMemory;
    options.AIProvider = AIProvider.Gemini;
    options.DefaultLanguage = "tr";  // Optional: Default language for document processing
});

var app = builder.Build();
app.Run();

For Console Applications:

using SmartRAG.Extensions;
using SmartRAG.Enums;

var services = new ServiceCollection();
var configuration = new ConfigurationBuilder()
    .AddJsonFile("appsettings.json")
    .Build();

// UseSmartRag returns IServiceProvider with auto-started services
var serviceProvider = services.UseSmartRag(
    configuration,
    storageProvider: StorageProvider.InMemory,
    aiProvider: AIProvider.Gemini,
    defaultLanguage: "tr"  // Optional
);

// Use the service provider
var documentService = serviceProvider.GetRequiredService<IDocumentService>();

Advanced Setup

using SmartRAG.Extensions;
using SmartRAG.Enums;

var builder = WebApplication.CreateBuilder(args);

// Advanced configuration with options
builder.Services.AddSmartRag(builder.Configuration, options =>
{
    // AI Provider
    options.AIProvider = AIProvider.OpenAI;
    
    // Storage Provider
    options.StorageProvider = StorageProvider.Qdrant;
    
    // Chunking Configuration
    options.MaxChunkSize = 1000;
    options.MinChunkSize = 100;
    options.ChunkOverlap = 200;
    
    // Retry Configuration
    options.MaxRetryAttempts = 3;
    options.RetryDelayMs = 1000;
    options.RetryPolicy = RetryPolicy.ExponentialBackoff;
    
    // Fallback Providers
    options.EnableFallbackProviders = true;
    options.FallbackProviders = new List<AIProvider> 
    { 
        AIProvider.Anthropic, 
        AIProvider.Gemini 
    };
    
    // Default Language
    options.DefaultLanguage = "tr";  // Optional: Default language for document processing
});

var app = builder.Build();
app.Run();

SmartRagOptions - Core Options

Core configuration options available in SmartRagOptions:

Option	Type	Default	Description
`AIProvider`	`AIProvider`	`OpenAI`	AI provider for embeddings and text generation
`StorageProvider`	`StorageProvider`	`InMemory`	Storage backend for documents and vectors
`ConversationStorageProvider`	`ConversationStorageProvider?`	`null`	Separate storage for conversation history (optional)
`EnableAutoSchemaAnalysis`	`bool`	`true`	Automatically analyze database schemas on startup
`DefaultLanguage`	`string?`	`null`	Default language code for document processing (ISO 639-1 format, e.g., "tr", "en", "de"). Used when language is not specified in WatchedFolderConfig or document upload.

ConversationStorageProvider

Separate storage configuration for conversation history, independent from document storage:

builder.Services.AddSmartRag(configuration, options =>
{
    options.AIProvider = AIProvider.OpenAI;
    options.StorageProvider = StorageProvider.Qdrant;  // For documents
    options.ConversationStorageProvider = ConversationStorageProvider.Redis;  // For conversations
});

Available Options

Option	Description
`Redis`	Store conversations in Redis (high-performance cache)
`SQLite`	Store conversations in SQLite database (embedded, lightweight)
`FileSystem`	Store conversations in file system (simple, persistent)
`InMemory`	Store conversations in RAM (not persistent, development only)

Configuration Example

{
  "SmartRAG": {
    "ConversationStorageProvider": "Redis",
    "RedisConfig": {
      "ConnectionString": "localhost:6379"
    }
  }
}

Conversation Storage Tips

Redis: Best for production, high-performance caching
SQLite: Good for development and small deployments
FileSystem: Simple, human-readable storage
InMemory: Fast, but data lost on restart

Chunking Options

Option	Type	Default	Description
`MaxChunkSize`	`int`	`1000`	Maximum size of each document chunk in characters
`MinChunkSize`	`int`	`100`	Minimum size of each document chunk in characters
`ChunkOverlap`	`int`	`200`	Number of characters to overlap between adjacent chunks

Chunking Best Practices

MaxChunkSize: 500-1000 characters for optimal balance
ChunkOverlap: 15-20% of MaxChunkSize for context preservation
Larger chunks: Better context, but slower search
Smaller chunks: More precise results, but less context

Feature Toggles

Control which search capabilities are enabled globally:

builder.Services.AddSmartRag(configuration, options =>
{
    options.Features.EnableDatabaseSearch = true;
    options.Features.EnableDocumentSearch = true;
    options.Features.EnableAudioSearch = true;
    options.Features.EnableImageSearch = true;
    options.Features.EnableMcpSearch = false;
    options.Features.EnableFileWatcher = false;
});

Feature Toggle Options

Option	Type	Default	Description
`Features.EnableDatabaseSearch`	`bool`	`true`	Enable multi-database query capabilities (SQL Server, MySQL, PostgreSQL, SQLite)
`Features.EnableDocumentSearch`	`bool`	`true`	Enable document text search (PDF, Word, Excel, etc.)
`Features.EnableAudioSearch`	`bool`	`true`	Enable audio file transcription and search (MP3, WAV, etc.)
`Features.EnableImageSearch`	`bool`	`true`	Enable image OCR and search (PNG, JPG, etc.)
`Features.EnableMcpSearch`	`bool`	`false`	Enable MCP (Model Context Protocol) server integration for external tools
`Features.EnableFileWatcher`	`bool`	`false`	Enable automatic document indexing from watched folders

Feature Toggle Tips

Per-Request Control: Use SearchOptions for per-request feature control
Global Control: Use Features for global feature enablement
Performance: Disable unused features to improve performance
Resource Management: Disable audio/image search if not needed to save processing resources

DocumentType Property

The DocumentType property in DocumentChunk allows filtering chunks by content type:

“Document”: Regular text documents (PDF, Word, Excel, TXT)
“Audio”: Audio file transcriptions (MP3, WAV, M4A)
“Image”: Image OCR results (PNG, JPG, etc.)

Automatic Detection

DocumentType is automatically determined based on file extension and content type:

// Uploaded files are automatically categorized
var document = await _documentService.UploadDocumentAsync(
    fileStream, 
    "invoice.pdf",      // → DocumentType: "Document"
    "application/pdf",
    "user-123"
);

var audio = await _documentService.UploadDocumentAsync(
    audioStream,
    "meeting.mp3",      // → DocumentType: "Audio"
    "audio/mpeg",
    "user-123"
);

var image = await _documentService.UploadDocumentAsync(
    imageStream,
    "receipt.jpg",      // → DocumentType: "Image"
    "image/jpeg",
    "user-123"
);

Filtering by DocumentType

Use SearchOptions to filter by document type:

// Search only in text documents
var options = new SearchOptions
{
    EnableDocumentSearch = true,
    EnableAudioSearch = false,  // Exclude audio chunks
    EnableImageSearch = false  // Exclude image chunks
};

var response = await _searchService.QueryIntelligenceAsync(
    "Find invoice details",
    maxResults: 10,
    options: options
);

Retry & Resilience Options

Option	Type	Default	Description
`MaxRetryAttempts`	`int`	`3`	Maximum number of retry attempts for AI provider requests
`RetryDelayMs`	`int`	`1000`	Delay between retry attempts in milliseconds
`RetryPolicy`	`RetryPolicy`	`ExponentialBackoff`	Retry policy for failed requests
`EnableFallbackProviders`	`bool`	`false`	Enable fallback to alternative AI providers on failure
`FallbackProviders`	`List<AIProvider>`	`[]`	List of fallback AI providers to try sequentially

RetryPolicy Enum Values:

RetryPolicy.None - No retries
RetryPolicy.FixedDelay - Fixed delay between retries
RetryPolicy.LinearBackoff - Linearly increasing delay
RetryPolicy.ExponentialBackoff - Exponentially increasing delay (recommended)

Example Configurations

Development Environment

builder.Services.AddSmartRag(configuration, options =>
{
    // For fast development
    options.AIProvider = AIProvider.Gemini;
    options.StorageProvider = StorageProvider.InMemory;
    options.MaxChunkSize = 500;
    options.ChunkOverlap = 100;
});

Production Environment

builder.Services.AddSmartRag(configuration, options =>
{
    // For reliable production
    options.AIProvider = AIProvider.OpenAI;
    options.StorageProvider = StorageProvider.Qdrant;
    options.MaxChunkSize = 1000;
    options.ChunkOverlap = 200;
    options.MaxRetryAttempts = 5;
    options.RetryPolicy = RetryPolicy.ExponentialBackoff;
    options.EnableFallbackProviders = true;
    options.FallbackProviders = new List<AIProvider> { AIProvider.Anthropic };
});

Next Steps

AI Providers

OpenAI, Anthropic, Google Gemini and custom providers

AI Providers

Storage Providers

Qdrant, Redis, SQLite and other storage options

Storage Providers

Basic Configuration

Configuration Methods

Quick Setup (Recommended)

Advanced Setup

SmartRagOptions - Core Options

ConversationStorageProvider

Available Options

Configuration Example

Conversation Storage Tips

Chunking Options

Chunking Best Practices

Feature Toggles

Feature Toggle Options

Feature Toggle Tips

DocumentType Property

Automatic Detection

Filtering by DocumentType

Retry & Resilience Options

Example Configurations

Development Environment

Production Environment

Next Steps

AI Providers

Storage Providers