Ollama Configuration

2 Apr 20268 minutes to read

The Syncfusion^® Blazor Smart Rich Text Editor supports Ollama for running open-source models locally. This is ideal for privacy-conscious applications and development environments.

What is Ollama?

Ollama is a lightweight, open-source framework for running large language models locally on your machine. It provides:

• Privacy: All processing happens locally
• No API costs: No cloud service fees
• Offline capability: Works without internet connection
• Model variety: Access to many open-source models
• Easy setup: Simple installation and management

Prerequisites

• Windows 10/11, macOS, or Linux
• At least 8GB RAM (16GB+ recommended)
• GPU support optional but recommended for performance
• Docker (optional, for containerized deployment)

Installation

Step 1: Download and Install Ollama

Visit Ollama’s Official Website and download the installer for your operating system.

Windows

1. Download the Windows installer
2. Run the installer and follow the setup wizard
3. Accept default installation path or customize
4. Complete installation

macOS

1. Download the macOS installer
2. Open the .dmg file
3. Drag Ollama to Applications folder
4. Launch from Applications

Linux

Use the installation script:

curl https://ollama.ai/install.sh | sh

Step 2: Verify Installation

Open a terminal/command prompt and verify:

ollama --version

Step 3: Start Ollama Service

Windows

Ollama starts automatically. Access at http://localhost:11434

macOS/Linux

Start the Ollama service:

ollama serve

Installing Models

Available Models

Browse available models at Ollama Library

Popular models for text generation:

• Mistral: mistral - Fast, efficient
• Llama 2: llama2 - General purpose
• Neural Chat: neural-chat - Conversational
• Orca Mini: orca-mini - Lightweight
• Dolphin: dolphin-mixtral - Advanced

Install a Model

ollama pull mistral

This downloads and prepares the model (may take several minutes).

List Installed Models

ollama list

Run a Model Directly

Test the model:

ollama run mistral

Type your prompt and press Enter. Type /bye to exit.

Configuration in Blazor

Note: OllamaSharp is a community package for local model integration. Ensure compliance with your organization’s policy on third-party NuGet packages before using in production.

Installation

Install required NuGet packages:

Package Manager

Install-Package Syncfusion.Blazor.SmartRichTextEditor
Install-Package Syncfusion.Blazor.Themes
Install-Package Microsoft.Extensions.AI
Install-Package OllamaSharp

Setup in Program.cs

using Syncfusion.Blazor;
using Syncfusion.Blazor.AI;
using Microsoft.Extensions.AI;
using OllamaSharp;

var builder = WebApplication.CreateBuilder(args);

// Add services
builder.Services.AddRazorPages();
builder.Services.AddServerSideBlazor();
builder.Services.AddSyncfusionBlazor();

// Configure Ollama
string ollamaEndpoint = "http://localhost:11434";
string modelName = "mistral"; // or any other installed model

// Create Ollama client
IOllamaApiClient ollamaClient = new OllamaApiClient(ollamaEndpoint, modelName);

// Convert to IChatClient
IChatClient chatClient = ollamaClient;

builder.Services.AddChatClient(chatClient);

// Register Smart Rich Text Editor Components with Azure OpenAI
builder.Services.AddSingleton<IChatInferenceService, SyncfusionAIService>();

var app = builder.Build();

// ... rest of setup

Add to _Imports.razor

~/Components/_Imports.razor

@using Syncfusion.Blazor
@using Syncfusion.Blazor.SmartRichTextEditor

Use Ollama AI with Smart Rich Text Editor Component

RAZOR

@using Syncfusion.Blazor.SmartRichTextEditor

<SfSmartRichTextEditor>
    <AssistViewSettings Placeholder="Edit with local AI assistance..."/>
    <div>
        <h3>Ollama + Smart Rich Text Editor</h3>
        <p>Use AI Commands to improve selected text or change tone.</p>
        <ul>
            <li>Select text and request suggestions</li>
            <li>Press Alt+Enter for AI Query</li>
        </ul>
    </div>
</SfSmartRichTextEditor>

Configuration Options

Custom Endpoint

If Ollama runs on different host/port:

string ollamaEndpoint = "http://192.168.1.100:11434";
IOllamaApiClient ollamaClient = new OllamaApiClient(ollamaEndpoint, "mistral");

Model Parameters

Configure model behavior:

// Create client with parameters
IOllamaApiClient ollamaClient = new OllamaApiClient("http://localhost:11434", "mistral");

// Set parameters (varies by OllamaSharp version)
var request = new GenerateRequest
{
    Model = "mistral",
    Prompt = "Your prompt",
    Temperature = 0.7f,
    TopK = 40,
    TopP = 0.9f,
};

Dynamic Model Selection

// Read model name from configuration
string modelName = builder.Configuration["Ollama:ModelName"] ?? "mistral";
string endpoint = builder.Configuration["Ollama:Endpoint"] ?? "http://localhost:11434";

IOllamaApiClient ollamaClient = new OllamaApiClient(endpoint, modelName);

Docker Deployment

Docker Compose Setup

version: '3.8'
services:
  ollama:
    image: ollama/ollama:latest
    ports:
      - "11434:11434"
    volumes:
      - ollama_data:/root/.ollama
    environment:
      - OLLAMA_HOST=0.0.0.0:11434

volumes:
  ollama_data:

Start with Docker

docker-compose up -d

Pull Models in Container

docker exec -it <container-id> ollama pull mistral

Advanced Configuration

Multiple Models

Run different models for different purposes:

// For general editing
var generalClient = new OllamaApiClient("http://localhost:11434", "mistral");

// For code-related tasks
var codeClient = new OllamaApiClient("http://localhost:11434", "neural-chat");

// Conditionally use based on context
IChatClient chatClient = useForCode ? codeClient : generalClient;

Performance Optimization

Tips for Better Performance

1. Use GPU Acceleration
   • NVIDIA GPUs: Install CUDA toolkit
   • AMD GPUs: Install ROCm
   • Intel GPUs: Install Intel oneAPI
2. Increase Memory Allocation

   # Linux/macOS
   export OLLAMA_NUM_PARALLEL=1
   ollama serve

3. Choose Efficient Models
   • Mistral is fastest
   • Orca-mini for lightweight deployment
   • Dolphin-mixtral for best quality
4. Implement Caching
   • Cache common prompts
   • Reuse model responses
5. Optimize Request Size
   • Keep prompts concise
   • Limit response length with max tokens

Troubleshooting

Connection Issues

Error: Unable to connect to Ollama

1. Verify Ollama is running:

   curl http://localhost:11434/api/tags

2. Check endpoint configuration in Program.cs

3. If running on different machine, update endpoint URL

Model Issues

Model not found

# List available models
ollama list

# Pull the model
ollama pull mistral

Out of memory errors

• Use smaller models (orca-mini, mistral)
• Close other applications
• Restart Ollama service

Performance Issues

Slow response times

• Check CPU/GPU usage
• Enable GPU acceleration if available
• Use faster models
• Increase system RAM

Model Recommendations

Use Case	Model	Pros	Cons
General editing	Mistral	Fast, good quality	Less context
Content creation	Llama 2	Balanced	Larger model
Code assistance	Neural-chat	Good reasoning	Slower
Lightweight	Orca-mini	Very fast	Limited capability
Best quality	Dolphin-mixtral	Excellent	Resource heavy

Security Considerations

Local Processing Benefits

• All data stays on your machine
• No external API calls
• Compliance with data regulations
• Full control over data retention

Best Practices

1. Restrict network access to Ollama
2. Use firewall rules if networked
3. Keep Ollama updated
4. Monitor resource usage

See also

• Getting Started with Smart Rich Text Editor
• OpenAI Configuration
• Azure OpenAI Configuration
• AI Features and Customization
• Ollama Official Documentation