Run Powerful AI Coding Locally on a Normal Laptop
A Developer-Friendly Guide to Setting Up ROO Code + Ollama + Qwen (8GB/16GB RAM)

If you are a developer who wants to use AI coding assistants locally without paying for cloud APIs or owning a high-end GPU, this guide is for you.

In this article, we will set up:

ROO Code inside Visual Studio Code
Ollama for running local AI models
Qwen2.5-Coder model locally
Optimized for:
8GB RAM laptops
16GB RAM laptops
No dedicated GPU / No VRAM

By the end, you’ll have your own private AI coding assistant running fully offline.
Why Run AI Locally?

Running AI locally gives developers:

✅ No API cost
✅ Better privacy
✅ Faster experimentation
✅ Offline development
✅ Full control over models
✅ No dependency on cloud providers

Recommended Hardware
Configuration Recommended Model
8GB RAM qwen2.5-coder:1.5b
16GB RAM qwen2.5-coder:7b
16GB+ RAM qwen2.5-coder:14b (slow but possible)

If you have no GPU, don’t worry. Ollama can run models entirely on CPU.

Step 1 — Install Visual Studio Code

Download and install:
Visual Studio Code
Use the official website:

After installation:

code --version

Verify VS Code is properly installed.

Step 2 — Install Ollama

Install:

Ollama

Windows

Download installer from the official Ollama website.

Verify installation:

ollama --version

Step 3 — Start Ollama

Run:

ollama serve

This starts the local AI server at:

http://localhost:11434

Keep this terminal running.

Step 4 — Install Qwen Coding Model
For 8GB RAM Systems

Recommended:

ollama run qwen2.5-coder:1.5b
Why?

Lightweight
Fast on CPU
Good enough for:
Code generation
Refactoring
Unit tests
Small automation tasks

For 16GB RAM Systems

Recommended:
ollama run qwen2.5-coder:7b

This gives much better:

Reasoning
Architecture suggestions
Refactoring quality
Multi-file understanding

Step 5 — Test the Model

Try:

ollama run qwen2.5-coder:7b

Then ask:

Who are you and create a hello world example in python

If the model responds, you’re ready.

Step 6 — Install ROO Code Extension

Inside VS Code:

Open Extensions
Search:

Roo Code
Install the extension

ROO Code converts VS Code into an AI-powered development environment.

Step 7 — Configure ROO Code for Ollama

Open ROO Code settings.

Set:

Provider: Ollama

API Endpoint:

http://localhost:11434

Model:

For 8GB RAM:

qwen2.5-coder:1.5b

For 16GB RAM:

qwen2.5-coder:7b

Save settings.

Step 8 — First AI Coding Test

Open a project and ask ROO Code:

Create a Java Spring Boot CRUD API with Controller, Service, Repository

Or:

Generate Cypress automation for login page

You now have a local AI coding assistant.

Best Practices for Low-RAM Systems
For 8GB RAM Machines
Recommended Settings
Setting Value
Context Window Small
Concurrent Apps Minimal
Model 1.5B
Browser Tabs Limited
Avoid

❌ Running Docker + AI together
❌ Opening large IDE projects
❌ Using 7B models continuously

Best Practices for 16GB RAM Machines

You can comfortably use:

qwen2.5-coder:7b
Medium-size repositories
Spring Boot projects
React applications
Cypress automation generation

Recommended:

OLLAMA_NUM_PARALLEL=1

This prevents RAM spikes.

Performance Optimization Tips
Reduce Model Temperature

Better coding consistency:

temperature = 0.2
Keep Context Smaller

Instead of entire repositories:

✅ Open only relevant folders

This improves response quality and speed.

Restart Ollama Occasionally

Long sessions can consume memory.

Restart:

ollama stop
ollama serve
Recommended Models by Use Case
Use Case Recommended Model
Basic coding qwen2.5-coder:1.5b
Java development qwen2.5-coder:7b
Test automation qwen2.5-coder:7b
Architecture discussion qwen2.5-coder:7b
Large enterprise code DeepSeek-Coder 14B (16GB+)
What Works Surprisingly Well Locally?

Even without a GPU, local models perform very well for:

✅ Boilerplate generation
✅ Refactoring
✅ Unit tests
✅ Cypress automation
✅ SQL generation
✅ Spring Boot scaffolding
✅ API creation
✅ Debugging suggestions
✅ Documentation generation

Limitations

Be realistic about CPU-only setups.

You may experience:

Slower response time
Limited context handling
Occasional hallucinations
Reduced multi-file reasoning

But for day-to-day development, the experience is still highly productive.

My Recommended Setup
For Most Developers
8GB RAM
Ollama + qwen2.5-coder:1.5b + Roo Code
16GB RAM
Ollama + qwen2.5-coder:7b + Roo Code

This provides the best balance between:

Performance
Memory usage
Coding quality
Stability
Final Thoughts

Local AI development is no longer limited to expensive GPUs.

Today, even a normal laptop can run surprisingly capable coding assistants using:

Ollama
Qwen2.5-Coder
Visual Studio Code
ROO Code

For developers working in Java, Spring Boot, React, Cypress, AI automation, and system design — this setup is an excellent starting point into the world of local AI engineering.

Useful Commands Cheat Sheet