LocalPilot Logo LocalPilot v0.2.5
🌐 العربية Mohamed Fakhruldeen
📖 Beginner Walkthrough

Your First Conversation with LocalPilot

A calm, plain-English guide explaining what local AI is, how your computer's memory works, and how to chat with your first model without technical frustration.

1

What is "Local AI"?

When you use online chatbots like ChatGPT or Claude, your words travel across the internet to giant data centers owned by large tech companies. Your conversations may be stored, used to train future systems, or processed on shared servers.

☁️ Cloud Chatbots

  • • Sends your words over the internet to remote servers
  • • Requires internet connection and often monthly fees
  • • Can experience queues during peak hours
  • • Your data lives on external infrastructure

💻 Local AI (LocalPilot)

  • • Runs inside your computer's memory and processor
  • • Chats run locally without sending conversations to cloud APIs
  • • Completely free and open-source software (GPLv3)
  • • Internet is needed only for downloading and updates

Local AI simply means downloading an AI model once onto your laptop's hard drive and running it directly, just like a local word processor or media player.

2

The No-Jargon AI Dictionary

When exploring local AI, you will encounter terms that sound like technical code. Here is what they actually mean:

🧠

Model

The downloaded knowledge file containing patterns learned from language. Think of it like a specialized digital book. Popular models include Llama 3.2, Qwen 2.5, Gemma 2, and Mistral.

💻

Terminal

A text window on your computer where you paste commands. LocalPilot opens a visual, full-screen dashboard inside this window so you do not have to memorize complex flags.

⚙️

Supporting Runtime

The software engine (like llama.cpp or Ollama) that reads the model file and generates replies word by word. LocalPilot configures this automatically.

🪑

RAM (Short-Term Memory)

When you chat with a model, its file must sit in your computer's RAM. LocalPilot measures your memory first and only recommends models that leave a safe buffer for your operating system and browser.

📜

Context

How much text the model can remember at once during a conversation. A larger context allows longer articles, but uses slightly more memory.

🎭

Hallucination

When an AI states an incorrect fact with total confidence. Always review important facts and test generated code before relying on it.

3

Step-by-Step Setup Walkthrough

A Open Your Terminal

On macOS: Press ⌘ Command + Space, type Terminal, and press Return.
On Linux: Press Ctrl + Alt + T, or search for "Terminal" in your applications menu.

B Install LocalPilot

Paste this single command into your terminal window and press Enter:

curl -fsSL https://raw.githubusercontent.com/fakhruldeen/LocalPilot/main/install-localpilot.sh | bash

The installer places LocalPilot into an isolated application folder. It does not touch your personal files and never requests root or administrator passwords.

C Open the Dashboard

Type the launch command and press Enter:

localpilot

A full-screen dashboard will open with the Guided Setup Wizard ready to help you.

D The 4 Guided Steps Inside the App

Step 1: Computer Capability Check

LocalPilot inspects your processor, graphics, and RAM, reserving a safe buffer for your operating system. Click Continue → when ready.

Step 2: What would you like help with?

Choose between Everyday questions and writing, Coding assistance, or Browse all compatible models. Click Find Best Models →.

Step 3: Recommended Models

Select a recommended model that fits your computer's safe budget. Click Review Setup →.

Step 4: Review & Approve Setup

Check the download size and folder location. Nothing downloads until you click Set Up Model. LocalPilot downloads the weights, sets up supporting software, and tests a quick query to confirm the model works on your hardware.

E Send Your First Message

Click Start Chatting (or run localpilot chat in your terminal). Here are practical prompt ideas:

"Help me write a warm, professional thank-you note to a colleague who helped me finish a project on time."
"Summarize the following meeting notes into 3 clear bullet points: [paste your text]"
"Explain what an API is in simple terms, using an analogy like ordering at a restaurant."
4

Everyday Tasks, Offline Search & Voice

Beyond open-ended conversation, LocalPilot gives you focused offline tools that run directly against your files without uploading anything to cloud servers:

⚡ 18 Offline AI Task Tools

Run targeted actions from your terminal or dashboard: summarize articles, polish drafts, translate between languages, explain code, analyze tables, and transcribe images.

localpilot tools summarize report.txt
🔍 Offline Document Search (RAG)

Index your local folders, research papers, or manuals. Ask natural questions and receive answers with exact file and line number citations.

localpilot search ask "how do I configure backup?"
🎙️ Voice & Push-to-Talk Speech

Talk directly to your local model using voice input, transcribe recordings into text, and hear responses read aloud using your computer's built-in speech engine.

localpilot voice chat
📊 Performance Benchmarking & Health

Measure actual streaming speed (tokens/sec) and time-to-first-token. Run self-healing diagnostics to automatically resolve broken connections or stale files.

localpilot doctor --repair
5

Leaving Chat vs. Stopping a Model

Understanding how model memory works gives you total control over your computer's speed:

Leaving Chat (/exit)

Closes your conversation window, but keeps the model loaded in memory in the background. If you open chat again later, replies begin immediately without reload delay.

Stopping the Model (localpilot stop)

Terminates the model process cleanly, releasing every megabyte of RAM and GPU memory back to your computer for other demanding applications or games.

Removing a Model to Free Disk Space

When you no longer need a model, remove it cleanly:

localpilot remove "Llama 3.2 3B"

This unregisters the model and deletes its weight files from your hard drive to recover storage space.

6

If Something Goes Wrong

Download stopped or was interrupted:

Simply run localpilot again. It verifies already downloaded files and resumes safely.

Computer feels sluggish:

Run localpilot stop to free memory immediately. For daily use, select a smaller model using localpilot discover.

"command not found" error:

Add export PATH="$PATH:$HOME/.local/bin" to your shell profile (~/.bashrc or ~/.zshrc).

Need diagnostics for help?

Run localpilot support to generate a safe, redacted diagnostic report you can review and share.