Installing and Running Ollama with Deepseek-R 1.5B

This guide walks through installing Ollama and running the Deepseek-R 1.5B model on your local machine.

Prerequisites

• A Linux, macOS, or Windows system with WSL2
• At least 8GB RAM (16GB recommended)
• 4GB free disk space
• NVIDIA GPU recommended but not required

Installing Ollama

Linux

curl -fsSL https://ollama.com/install.sh | sh

macOS

Download and install from https://ollama.com/download

Windows

1. Install WSL2 if not already installed
2. Follow Linux installation instructions within WSL2

Getting Started

1. Start the Ollama service:

systemctl start ollama    # Linux with systemd
ollama serve             # macOS or manual start

2. Pull the Deepseek-R 1.5B model:

ollama pull deepseek-coder:1.5b

Running the Model

Command Line Interface

ollama run deepseek-coder:1.5b

API Usage

Start making requests to the local API endpoint:

curl -X POST http://localhost:11434/api/generate -d '{
  "model": "deepseek-coder:1.5b",
  "prompt": "Write a Python function to calculate fibonacci numbers"
}'

Common Parameters

Adjust these parameters when running the model:

• Temperature (creativity): --temperature 0.7
• Context length: --ctx-size 2048
• System prompt: --system "You are a helpful coding assistant"

Example with parameters:

ollama run deepseek-coder:1.5b --temperature 0.7 --ctx-size 2048

Troubleshooting

1. If the model fails to load:

   • Check available system memory
   • Ensure GPU drivers are up to date (if using GPU)
   • Verify model was downloaded successfully

2. If you get API connection errors:

   • Confirm Ollama service is running
   • Check if port 11434 is available
   • Verify firewall settings

Resources

• Ollama documentation: https://ollama.com/docs
• Deepseek-R model card: https://ollama.com/library/deepseek-coder
• Community Discord: https://discord.gg/ollama

Additional Tips

• Use ctrl+c to exit the chat interface
• Use ollama list to see installed models
• Use ollama rm deepseek-coder:1.5b to remove the model

sadfsafd

private processPrompt(userMessage: string): string {
    const selectedCode = this.getSelectedCode();

    const systemPrompt = `You are an expert programming assistant. Format your responses using these rules:
        1. Use markdown formatting with proper code blocks
        2. Separate explanations and code with clear headings using markdown (##)
        3. Keep explanations concise and focused
        4. Structure complex responses in sections
        `;

    const prompt = `
        ${selectedCode ? `**Context (Selected Code):** \`\`\` ${selectedCode} \`\`\`\n` : ''}
        **User Query:** ${userMessage}
        **Task:** Provide a clear, well-structured response that directly addresses the query. Include relevant code examples where appropriate.
    `;

    return prompt;
    // const response = await this.getResponseText(prompt);
    // return response;
}