▣ Chapter 1: Why RAG (Research Augmentation Generation) is Needed

1.1 What is RAG?
1.2 Simple, advanced, and modular RAG configurations
1.3 RAG vs. Fine Tuning
1.4 RAG Ecosystem
__1.4.1 Search Engine (D)
__1.4.2 Generator (G)
__1.4.3 Evaluator (E)
__1.4.4 Training (T)
1.5 Simple, Advanced, and Modular Python Implementations of RAG
__1.5.1 Part 1: Basics and Basic Implementation
__1.5.2 Part 2: Advanced Techniques and Evaluation

▣ Chapter 2: RAG Embedding Vector Repository Using Deep Lake and OpenAI

2.1 From raw data to embedding in vector storage
2.2 Configuring the RAG system as a single pipeline
2.3 RAG-based generative AI pipeline
2.4 Building the RAG Pipeline
__2.4.1 Preferences
__2.4.2 Component 1: Data Collection and Preparation
__2.4.3 Component 2: Data Embedding and Storage
__2.4.4 Vector storage information
__2.4.5 Component 3: Input Augmentation and Response Generation
2.5 Output Evaluation Using Cosine Similarity

▣ Chapter 3: Building an Index-Based RAG Using RamaIndex, Deep Lake, and OpenAI

3.1 Why Use Index-Based RAG?
__3.1.1 Architecture
3.2 Building a Semantic Search Engine and Generative Agent for Drone Technology Information
__3.2.1 Environment Installation
__3.2.2 Pipeline 1: Document Collection and Preparation
__3.2.3 Pipeline 2: Preparing the Vector Store
__3.2.4 Pipeline 3: Index-based RAG
3.3 Vector Storage Index and Query Engine
__3.3.1 Query Response and Source Verification
__3.3.2 Optimized chunking
__3.3.3 Performance Metrics
3.4 Tree Index Query Engine
__3.4.1 Performance Metrics
3.5 List Index Query Engine
__3.5.1 Performance Metrics
3.6 Keyword Index Query Engine
__3.6.1 Performance Metrics

▣ Chapter 4: Multimodal Modular RAG for Drone Technology

4.1 What is a multi-modal modular RAG?
4.2 Building a Multimodal Modular RAG Program for Drone Technology
__4.2.1 Loading the LLM dataset
__4.2.2 Loading and Visualizing Multimodal Datasets
__4.2.3 Multimodal Dataset Structure
__4.2.4 Building a Multi-Modal Query Engine
__4.2.5 Multi-modal modular query result summary
__4.2.6 Performance Metrics

▣ Chapter 5: Improving RAG Performance Using Expert Feedback

5.1 Adaptive RAG
5.2 Building a Hybrid Adaptive RAG System Using Python
__5.2.1 Finder (Section 1)
__5.2.2 Generator (Section 2)
__5.2.3 Evaluator (Section 3)

▣ Chapter 6: Expanding RAG Bank Customer Data Using Finecon

6.1 Extension using Pinecone
__6.1.1 Architecture
6.2 Pipeline 1: Dataset Collection and Preparation
__6.2.1 Dataset Collection and Processing (Section 1)
__6.2.2 Exploratory Data Analysis (Section 2)
__6.2.3 ML Model Training (Section 3)
6.3 Pipeline 2: Expanding the Pinecone Index (Vector Storage)
__6.3.1 Challenges of Vector Storage Management
__6.3.2 Environment Installation
__6.3.3 Dataset Processing
__6.3.4 Dataset Chunking and Embedding
__6.3.5 Creating a Pinecone Index
__6.3.6 Upsert
__6.3.7 Pinecone Index Query
6.4 Pipeline 3: RAG Generative AI
__6.4.1 RAG using GPT-4o
__6.4.2 Extracting Related Text
__6.4.3 Input Augmentation and Prompt Engineering
__6.4.4 Augmented Generation

▣ Chapter 7: Building a Scalable Knowledge Graph-Based RAG Using the Wikipedia API and RamaIndex

7.1 RAG Architecture for Knowledge Graph-Based Semantic Retrieval
__7.1.1 Creating a tree graph with nodes
7.2 Pipeline 1: Document Collection and Preparation
__7.2.1 Searching Wikipedia data and metadata
__7.2.2 Preparing Data for Upsert
7.3 Pipeline 2: Creating and Populating the Deep Lake Vector Store
7.4 Pipeline 3: Knowledge Graph Index-Based RAG
__7.4.1 Creating a Knowledge Graph Index
__7.4.2 Graph display
__7.4.3 Interacting with the Knowledge Graph Index
__7.4.4 Installing the similarity score package and defining functions
__7.4.5 Re-ranking
__7.4.6 Example indicators

▣ Chapter 8: Dynamic RAG using chroma and hugging face llamas

8.1 Architecture of Dynamic RAG
8.2 Environment Installation
__8.2.1 Installing Hugging Face
__8.2.2 Installing Chroma
8.3 Enabling Session Time
8.4 Downloading and Preparing the Dataset
8.5 Data Embedding and Upsert in Chroma Collections
__8.5.1 Model Selection
__8.5.2 Document Embedding and Saving
__8.5.3 Embedding display
8.6 Executing queries on collections
8.7 Prompts and Searches
8.8 RAG using Rama
__8.8.1 Deleting a Collection
8.9 Total Session Time

▣ Chapter 9: Enhancing AI Model Capabilities - Fine-tuning RAG Data and Human Feedback

9.1 Architecture of static RAG data fine-tuning
__9.1.1 RAG Ecosystem
9.2 Environment Installation
9.3 Preparing the Dataset for Fine-Tuning (Section 1)
__9.3 1 Downloading and Visualizing the Dataset (Section 1.1)
__9.3 2 Preparing the Dataset for Fine-Tuning (Section 1.2)
9.4 Model Fine-Tuning (Section 2)
__9.4.1 Fine-tuning monitoring
9.5 Running the Fine-Tuned OpenAI Model (Section 3)
9.6 Indicator

▣ Chapter 10: RAG System for Video Stock Production Using Pinecone and OpenAI

10.1 RAG Architecture for Video Production
10.2 Environment of the video production ecosystem
__10.2.1 Importing modules and libraries
__10.2.2 GitHub
__10.2.3 OpenAI
__10.2.4 Pinecone
10.3 Pipeline 1: Generator and Commentary Writer
__10.3.1 AI-generated video dataset
__10.3.2 Generator and Commentary Writer
10.4 Pipeline 2: Vector Storage Manager
__10.4.1 Pinecone Index Query
10.5 Pipeline 3: Video Expert

▣ Appendix A: Practice Problem Answers

★ What this book covers ★

◎ Extend the RAG pipeline to efficiently process large datasets.
◎ Apply techniques to minimize hallucinations and ensure accurate responses.
◎ Implement an indexing technique that improves AI accuracy with traceable and transparent output.
◎ Customize and expand RAG-driven generative AI systems across various domains.
◎ Search data faster and more efficiently using Deep Lake and Pinecone.
◎ Build and control a robust generative AI system based on actual data.
◎ Combine text and image data for richer and more informative AI responses.