Get the eBook free when you register your print book at Manning.

Forget steep AI subscription fees, privacy concerns, and a hard dependency on cloud servers! Build Applications with Local AI Models on a Mac shows you exactly how to build and run a ChatGPT-style assistant entirely on your own Mac—free forever, with zero data leaving your machine. In this practical book, AI authority Keiji Kamigusa guides you step by step using readily available open source models, Ollama, Python, and Streamlit.

Every chapter builds toward a single goal: your own ChatGPT-like app, running 100% offline. Along the way, you’ll install and manage multiple models (Llama, Gemma, Qwen, and Mistral) through Ollama, set up a professional Python environment from scratch with Homebrew and VS Code, and master Streamlit’s elegant Python-only UI framework. Architectural diagrams, flowcharts, sequence diagrams, and side-by-side comparison tables bring abstract concepts to life.

You’ll customize AI behavior through system prompts, temperature tuning, and Ollama Modelfiles to create your own custom LLM personalities. You’ll verify true privacy by running the famous "airplane mode test" by cutting your network and watching your chatbot keep working. Next, add streaming responses, conversation memory, and model-selection sidebars. Finally, you’ll extend your app with RAG (retrieval-augmented generation) to read your own documents, and LangChain agents that use tools autonomously.

What's inside

• Install and run models like Llama, Gemma, Qwen, and Mistral on macOS
• Build a complete ChatGPT-like chatbot with Streamlit in 70 lines of code
• Add streaming responses, conversation memory, and model selection
• Run AI fully offline with verifiable privacy
• Troubleshoot memory errors, port conflicts, and other local LLM issues

About the reader

No programming experience required! Anyone with basic computer literacy can set up an offline AI model with this guide.

About the author

Keiji Kamigusa is an authority in AI and machine learning with over 16 years of industry experience. A former CTO of a text mining company, he has deployed local language models for 50+ companies, taught 42,000+ students across 42 Udemy courses, and authored more than 50 books on AI. He holds three patents in text analytics and is the founder of AIGYM Inc., which focuses on AI-driven corporate training and agent management platforms.