Ollama for Windows

# Open Windows Terminal (or Command Prompt) and check version
ollama --version
# You should see something like: ollama version 0.16.3

# Check if the service is running
ollama list
# Empty list means Ollama is running but no models downloaded yet

# Download Llama 3.2 (Meta's latest small model, ~2GB)
ollama pull llama3.2

# You'll see download progress:
pulling manifest
pulling 8934d96d3f08... 100% |████████████████████| 2.0 GB
verifying sha256 digest
writing manifest
success

# Start an interactive chat session
ollama run llama3.2

# Type your question and press Enter
>>> What is vibe coding?

# The model responds in real-time, generated on YOUR hardware
# Type /bye to exit the chat

# List all downloaded models with sizes
ollama list

# Show model details (architecture, parameters, license)
ollama show llama3.2

# See which models are loaded in GPU/CPU memory
ollama ps

# Remove a model to free disk space
ollama rm llama3.2

# Update a model to latest version
ollama pull llama3.2

# Launch integrated apps (new in 2026)
ollama launch openclaw    # AI assistant

# Install Open WebUI with Docker (GPU support)
docker run -d -p 3000:8080 --gpus=all \
  -v ollama:/root/.ollama \
  -v open-webui:/app/backend/data \
  --name open-webui --restart always \
  ghcr.io/open-webui/open-webui:ollama

# CPU-only version (no --gpus flag)
docker run -d -p 3000:8080 \
  -v ollama:/root/.ollama \
  -v open-webui:/app/backend/data \
  --name open-webui --restart always \
  ghcr.io/open-webui/open-webui:ollama

# Access at http://localhost:3000

# Save this as "Modelfile" (no extension)
FROM mistral-nemo

SYSTEM """You are a senior Shopify developer specializing in
Liquid templates, custom sections, and theme development.
Always write production-ready code with proper error
handling. Use vanilla JS only — no jQuery. Scope all CSS
under unique wrapper classes."""

PARAMETER temperature 0.3
PARAMETER num_ctx 8192

# Create and run your custom model:
# ollama create shopify-dev -f Modelfile
# ollama run shopify-dev

from ollama import chat

response = chat(model='mistral-nemo', messages=[
    {'role': 'user', 'content': 'Write a Shopify section schema for a product grid'}
])
print(response.message.content)

# Set in System Properties → Environment Variables
# Or run in PowerShell before starting Ollama:

# Reduce KV cache memory (fit more context in less VRAM)
$env:OLLAMA_KV_CACHE_TYPE = "q8_0"

# Change model storage location (useful if C: is small)
$env:OLLAMA_MODELS = "D:\ollama\models"

# Force CPU-only mode (if GPU causes issues)
$env:OLLAMA_NO_GPU = "1"

# Use specific GPUs in multi-GPU setups
$env:CUDA_VISIBLE_DEVICES = "0,1"

┌─────────────────────────────────────────────────────────┐
│  YOUR APPS & SYSTEMS                                    │
│                                                         │
│  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐  │
│  │ Synthetic    │  │ Custom       │  │ VS Code +    │  │
│  │ Director     │  │ Python/Node  │  │ Continue     │  │
│  │ v10.0        │  │ Scripts      │  │ Extension    │  │
│  └──────┬───────┘  └──────┬───────┘  └──────┬───────┘  │
│         │                 │                 │          │
│         ▼                 ▼                 ▼          │
│  ┌─────────────────────────────────────────────────┐    │
│  │        http://localhost:11434/api/chat          │    │
│  │        Ollama REST API (always running)         │    │
│  └──────────────────────┬──────────────────────────┘    │
│                         │                              │
│                         ▼                              │
│  ┌─────────────────────────────────────────────────┐    │
│  │  LOCAL MODELS  (Mistral-Nemo / Llama / Gemma)  │    │
│  │  Running on YOUR GPU (RTX 3060) + YOUR RAM     │    │
│  └─────────────────────────────────────────────────┘    │
│                                                         │
│  🔒 Everything stays on your machine. Zero cloud calls. │
└─────────────────────────────────────────────────────────┘

# Check if Ollama's API server is responding
curl http://localhost:11434
# Should return: "Ollama is running"

# Or in PowerShell:
Invoke-WebRequest -Uri http://localhost:11434 | Select-Object -ExpandProperty Content

# List available models via API
curl http://localhost:11434/api/tags

curl http://localhost:11434/api/chat -d '{
  "model": "mistral-nemo",
  "messages": [
    {
      "role": "system",
      "content": "You are a Shopify content writer for a sustainable fashion brand."
    },
    {
      "role": "user",
      "content": "Write an Instagram caption for our new organic cotton hoodie."
    }
  ],
  "stream": false
}'

# pip install ollama
from ollama import chat

# Simple chat — works exactly like calling a cloud API
response = chat(
    model='mistral-nemo',
    messages=[
        {'role': 'system', 'content': 'You are a senior developer.'},
        {'role': 'user', 'content': 'Review this code for bugs and security issues.'}
    ]
)
print(response.message.content)

import requests, json

# Call Ollama the same way you'd call OpenAI — just change the URL
response = requests.post(
    'http://localhost:11434/api/chat',
    json={
        'model': 'mistral-nemo',
        'messages': [
            {'role': 'user', 'content': 'Generate 5 product descriptions.'}
        ],
        'stream': False
    }
)

result = response.json()
print(result['message']['content'])

// npm install ollama
import ollama from 'ollama';

const response = await ollama.chat({
  model: 'mistral-nemo',
  messages: [
    { role: 'system', content: 'You are an AI content strategist.' },
    { role: 'user',   content: 'Plan a week of social media posts.' }
  ]
});

console.log(response.message.content);

// Works in Node.js 18+, Deno, Bun, or any modern runtime
const response = await fetch('http://localhost:11434/api/chat', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({
    model: 'mistral-nemo',
    messages: [{ role: 'user', content: 'Your prompt here' }],
    stream: false
  })
});

const data = await response.json();
console.log(data.message.content);

# pip install openai
from openai import OpenAI

# Point the OpenAI client at your local Ollama
client = OpenAI(
    base_url='http://localhost:11434/v1',
    api_key='ollama'  # Required by the library, but Ollama ignores it
)

response = client.chat.completions.create(
    model='mistral-nemo',
    messages=[
        {'role': 'system', 'content': 'You are a helpful assistant.'},
        {'role': 'user', 'content': 'Explain how Ollama works.'}
    ]
)

print(response.choices[0].message.content)

# That's it. Same OpenAI library, same code pattern.
# Just change the base_url to localhost:11434/v1
# Works with LangChain, LlamaIndex, CrewAI, and more.