sharpbyte.dev
← Learning hub
AI ecosystem · RAG

Retrieval-Augmented Generation, grounded and live

Thirteen topics from the same reference PDF as LLMs, Prompt engineering, and Fine-tuning—vector stores, full RAG workflows, chunking, architecture patterns, HyDE, REFRAG, cache-augmented generation, and the path to agent memory.

Topic 1

What is RAG?

Retrieval, augmentation, and generation—grounding LLMs without retraining on every update.

Read topic →
Topic 2

Vector databases & embeddings

Why embeddings cluster by meaning—and how vector stores enable similarity search over unstructured data.

Read topic →
Topic 3

Vector databases in RAG

Static training corpora, private data, context windows—and where vector memory fits.

Read topic →
Topic 4

RAG system workflow

Chunk, embed, store, query, retrieve, re-rank, and generate—end-to-end pipeline anatomy.

Read topic →
Topic 5

Chunking strategies for RAG

Fixed-size, semantic, recursive, structure-aware, and LLM-driven chunking—trade-offs and when to test.

Read topic →
Topic 6

Prompting vs RAG vs fine-tuning

Choose among steering, retrieval, weight updates—or hybrid RAG + fine-tuning—using knowledge vs adaptation axes.

Read topic →
Topic 7

Eight RAG architectures

Naive, multimodal, HyDE, corrective, graph, hybrid, adaptive, and agentic patterns—at a glance.

Read topic →
Topic 8

Agentic RAG

Agents rewrite queries, choose sources, iterate, and self-check—beyond one-shot retrieve-and-read.

Read topic →
Topic 9

HyDE retrieval

Hypothetical document embeddings align queries with answer-like text for better dense retrieval.

Read topic →
Topic 10

Full fine-tuning vs LoRA vs RAG

Three ways to add knowledge: full weights, adapters, or retrieval—pros, costs, and limits.

Read topic →
Topic 11

REFRAG

Meta’s relevance-aware pipeline: compress chunks, RL-filter, selectively expand before the decoder.

Read topic →
Topic 12

RAG + cache-augmented generation (CAG)

Cache stable knowledge in KV memory; keep volatile facts on the retrieval path.

Read topic →
Topic 13

From RAG to agent memory

Read-only retrieval → tool-mediated retrieval → read/write memory for personalization and continual learning.

Read topic →
How to read this track: PDF pages 106–146 (section divider + body). Context Engineering follows the book’s TOC on pages 147–176; page 146 is the last agent-memory slide only in this RAG track. AI Agents starts at page 177. Figures are clipped per slide block; narrative follows the deck wording.