Navi — Your AI Co-Navigator

Navigate your code with AI. Free. Local. No limits.

Publisher: amanblaze
GitHub: amangautamm/navi
Support: buymeacoffee.com/amanblaze

What is Navi?

Navi is a free AI coding assistant for VS Code, powered by Ollama running on your own machine. It works like GitHub Copilot and Cursor — but it costs nothing, has no usage limits, and your code never leaves your computer.

It works in two ways:

Project mode — Navi scans your entire project, finds errors across all files, and fixes them
File mode — Navi focuses on the file you have open and applies changes directly into the editor

Requirements

VS Code 1.74 or higher
Ollama installed and running
At least 8GB RAM
5–10GB free disk space for models

Quick Setup

Step 1 — Install Ollama

Go to ollama.com/download and install for your OS.

Step 2 — Download AI Models

Open your terminal and run:

# Main model — uncensored and smart
ollama pull dolphin-llama3:8b

# Best coding quality
ollama pull qwen2.5-coder:7b

# Fast ghost text completions (only 1GB)
ollama pull qwen2.5-coder:1.5b

Step 3 — Start Ollama

ollama serve

Keep this running in the background while using Navi.

Step 4 — Install the Extension

Download navi-1.0.0.vsix
Open VS Code
Press Ctrl+Shift+P → type Install from VSIX
Select the downloaded file
Reload VS Code when prompted

How to Use

Opening the Panel

Click the rocket icon (⚙) in the VS Code activity bar on the left. The Navi panel opens as a sidebar — always visible while you code.

Two Modes

At the top of the panel you'll see two buttons:

Button	What it does
📁 Project	Navi works on your entire codebase
📄 This File	Navi works only on the file currently open

Switch between them any time.

Sending a Prompt

Type anything in the text box at the bottom and press Enter (or Ctrl+Enter). You can write in English, Hindi, or Hinglish — Navi understands all of them.

Examples:

Fix all errors in the project
Explain what this file does
Add authentication to the Express app
Write unit tests for this file
Refactor the payment module
Security audit the whole project

Three Buttons

Button	Action
Send	Send your typed prompt
Fix Project	Instantly fix all errors across all files
Fix File	Fix errors in the currently open file

Applying Changes

When Navi suggests file changes, a preview box appears with an Apply button. You review the changes and click Apply — nothing changes without your confirmation (unless you turn on Auto Apply in settings).

Ghost Text

While typing code, Navi suggests completions automatically. Press Tab to accept, Esc to dismiss.

Keyboard Shortcuts

Shortcut	Action
`Ctrl+Shift+A`	Open agent prompt
`Ctrl+Shift+F`	Fix current file
`Ctrl+Shift+P`	Fix all project errors
`Ctrl+Shift+E`	Explain current file
`Tab`	Accept ghost text completion

Right-click on any file in the editor to access:

Fix This File
Explain This File
Refactor This File
Write Tests
Optimize
Add Documentation
Security Check
Fix All Project Errors

Tabs in the Panel

Tab	Use for
Chat	General questions about code
Agent	Complex tasks — Navi plans and executes
Debug	Finding and fixing bugs
Refactor	Improving code structure
Build	Creating new features or apps
Tests	Writing unit and integration tests
Security	Finding vulnerabilities

Settings

Open VS Code settings (Ctrl+,) and search for Navi to configure:

Setting	Default	Description
`navi.model`	`dolphin-llama3:8b`	Main model for chat and agent tasks
`navi.codingModel`	`qwen2.5-coder:7b`	Model for code tasks
`navi.fastModel`	`qwen2.5-coder:1.5b`	Fast model for ghost text
`navi.host`	`http://localhost:11434`	Ollama server URL
`navi.uncensored`	`true`	No content restrictions
`navi.autoComplete`	`true`	Enable ghost text
`navi.autoDebug`	`true`	Suggest fix when errors appear
`navi.autoApply`	`false`	Apply changes without confirmation
`navi.maxTokens`	`2048`	Max length of AI response

Recommended Models for 8GB RAM

Model	Size	Best for
`dolphin-llama3:8b`	4.7GB	Main chat, agent, uncensored
`qwen2.5-coder:7b`	4.7GB	Code generation and fixing
`qwen2.5-coder:1.5b`	1.0GB	Fast ghost text completions
`dolphin-mistral:7b`	4.1GB	Uncensored + fast responses
`mistral:7b`	4.1GB	General purpose, fast
`codellama:7b`	3.8GB	Lightweight coding model

Troubleshooting

Navi says it cannot connect to Ollama:

# Make sure Ollama is running
ollama serve

Ghost text is not appearing:

Check that navi.autoComplete is true in settings
Use qwen2.5-coder:1.5b as your fast model for quicker suggestions

Responses are slow:

Use a smaller model: qwen2.5-coder:1.5b
Reduce navi.maxTokens to 512 for shorter responses
Close other heavy applications to free up RAM

Model not found:

# Check what models you have
ollama list

# Download a model
ollama pull qwen2.5-coder:7b

Privacy

Everything runs on your machine. Your code is never sent to any server. Ollama runs 100% locally.

Support the Project

If Navi saves you time, consider buying a coffee:
buymeacoffee.com/amanblaze

GitHub: github.com/amangautamm/navi
Issues: github.com/amangautamm/navi/issues

Navi Ai

Amanblaze