The Grimoire Library

Welcome to the Grimoire library — a curated set of tutorials and background reading about large language models and the systems built around them. Everything here is self-contained: no external images, no CDN assets, no accounts required to read. Each document is meant to be genuinely useful to a practitioner, not marketing copy.

The library is vendor-neutral. Where a concept is illustrated with a specific model or tool (OpenAI, Anthropic, Ollama, or an open-weights model), it is only as an example — the ideas apply across providers. Where a topic is fast-moving, the text says so rather than inventing precise numbers that will be stale by the time you read them.

How the library is organized

Each top-level folder is a browsable "space." Read them in roughly this order if you are new, or jump straight to what you need.

content/
├── foundations/   what an LLM is, tokens, embeddings, attention, sampling
├── rag/           retrieval-augmented generation, end to end
├── prompting/     prompt patterns, structured output, few-shot, chain-of-thought
├── agents/        tool use, agent architectures, memory, guardrails
├── evaluation/    how to measure model and system quality
├── safety/        prompt injection, secrets, jailbreaks, red-teaming
└── reference/     glossary, model-family cheat sheet, further reading

Sections

• Foundations — the mental model. What a language model actually is, how text becomes tokens and vectors, why attention matters, and how sampling turns probabilities into words.
• Retrieval-Augmented Generation — how to ground a model in your own documents: chunking, embeddings, vector search, reranking, and evaluation.
• Prompting — practical patterns for getting reliable behavior, including structured output and the real limits of chain-of-thought.
• Agents — letting a model take actions through tools, and the architecture and guardrails that keep that safe and debuggable.
• Evaluation — measuring quality: offline harnesses, LLM-as-judge, and regression gates so quality does not silently drift.
• Safety — the adversarial side: prompt injection, secret handling, jailbreaks, and structured red-teaming.
• Reference — a glossary, a model-family cheat sheet, and a curated further-reading list.

Suggested reading paths

Total beginner: Foundations → Prompting → Reference glossary.

Building a RAG app: Foundations (embeddings, sampling) → the whole RAG section → Evaluation → Safety (prompt injection).

Building an agent: Foundations → Prompting (tool schemas) → Agents → Safety → Evaluation.

Happy reading.