Skip to content

QUICKSTART.md

Latest commit

 

History

History
119 lines (83 loc) · 2.84 KB

QUICKSTART.md

File metadata and controls

119 lines (83 loc) · 2.84 KB
Raw

Quick Start Guide

Get up and running with AI Agents from Scratch in 10 minutes.

Prerequisites

  • Python 3.10 or higher
  • 8GB+ RAM (for running local models)
  • ~5-10GB free disk space (for model files)

Step 1: Install Dependencies

pip install llama-cpp-python

Optional but recommended:

# Create a virtual environment first
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install llama-cpp-python

Step 2: Download a Model

You need a GGUF model file. Here's the easiest way:

  1. Go to https://huggingface.co/bartowski/Meta-Llama-3-8B-Instruct-GGUF
  2. Download Meta-Llama-3-8B-Instruct-Q4_K_M.gguf (~5GB)
  3. Place it in the models/ directory
  4. Rename it to llama-3-8b-instruct.gguf (optional, for simplicity)

Alternative models:

Step 3: Verify Setup

python setup_check.py

This checks:

  • Python version
  • Dependencies
  • Models directory
  • Repository structure

Step 4: Run Examples

python complete_example.py

This will run examples from all 10 lessons. You can also open complete_example.py and modify the model path or comment out lessons you want to skip.

Step 6: Start Learning

Now read the lessons in order:

  1. lessons/01_basic_llm_chat.md - Understanding the basics
  2. lessons/02_system_prompt.md - Adding behavior
  3. lessons/03_structured_output.md - Making it reliable
  4. ... and so on through lesson 10

Each lesson builds on the previous one.

Troubleshooting

"Module 'llama_cpp' not found"

pip install llama-cpp-python

"Model file not found"

Check that:

  1. The model file is in models/ directory
  2. The path in complete_example.py matches the actual filename
  3. The file has a .gguf extension

"Out of memory" error

Try a smaller model or smaller quantization:

  • Q4_K_M: ~5GB RAM
  • Q5_K_M: ~6GB RAM
  • Q8_0: ~8GB RAM

Slow responses

This is normal for CPU inference. Each response takes 10-30 seconds depending on:

  • Your CPU speed
  • Model size
  • Response length

Next Steps

  • Read philosophy.md to understand the approach
  • Work through lessons one at a time
  • Modify the agent to experiment
  • Check examples/ for complete code examples

Getting Help

Tips for Success

  1. Don't skip lessons - they build on each other
  2. Run the code - reading isn't enough
  3. Experiment - modify examples and see what happens
  4. Be patient - local inference is slow but worth it
  5. Read comments - the code explains the "why"

Happy learning! 🚀