The book
The Local LLM Bible — Beyond the Limits of Subscription AI (Korean title 『구독형 AI의 한계를 넘어선 로컬 LLM 구축 바이블』) was published by Silver Tree Books on 5 January 2026. It is Part 2 of the Silver Tree AI Trilogy. Where Part 1 taught how to talk to a language model, Part 2 teaches how to build and run one that an organization owns.
The book begins where most AI adoption stalls: the moment someone pastes company data into a subscription chatbot and hesitates. It answers three needs that hosted models cannot meet. Data sovereignty, so that internal documents never leave the building. A model that knows the organization’s own rules and vocabulary. And independence from a single vendor’s pricing, terms, and roadmap.
What it covers
The whole lifecycle of a local model is in one volume.
- Local infrastructure on Ollama: design, installation, model selection, and Modelfile basics.
- Prompt and role design: principles, system prompts, task patterns, and advanced technique.
- RAG, or retrieval-augmented generation, to connect the model to company documents: architecture, ingestion, chunking and embedding, vector search, and response quality.
- Fine-tuning to teach the model the house style: dataset design, quality control, LoRA and QLoRA, and deployment back to Ollama.
Two build-outs run end to end, from proof of concept to production: an internal knowledge assistant and a customer-support assistant. Each comes with runnable commands, Modelfiles, and architecture diagrams.
Contents
- Prologue. Why local: general-purpose LLMs versus a model that is yours.
- Chapter 1. LLMs and Ollama. Tokens, embeddings, and context; transformers; how prompting, RAG, and fine-tuning relate.
- Chapter 2. Building local infrastructure on Ollama.
- Chapter 3. Prompt and role design.
- Chapter 4. RAG: connecting company data.
- Chapter 5. Fine-tuning.
- Chapter 6. Two builds: an internal knowledge base and customer support, PoC to production.
- Afterword.
The stance
The book favors systems that can be stood up with the resources an organization already has over theory that cannot be deployed. It is written from the engineering side of the question, for readers who will be responsible for the result rather than for the demonstration.
For when the answer is not ChatGPT but a model your company owns.
Bibliographic details
- Publisher: Silver Tree Books
- Published: 5 January 2026
- ISBN: 979-11-993543-7-1
- Extent: 602 pages, 194 × 260 mm
- Language: Korean
- Series: Silver Tree AI Trilogy, Part 2 (Building AI)
The book page on this site carries the full description. The book is sold through YES24.
In the trilogy
Part 1, Understanding AI Through Prompts, was published in July 2025. Part 3, Introduction to Artificial Intelligence, a first course in how AI actually works with five core computations worked through by hand, was published on 3 August 2026 and completes the trilogy.