Building Data Driven Applications with LlamaIndex

Cover
Title Page
Copyright and Credits
Dedicated
Contributors
Table of Contents
Preface
Part 1:Introduction to Generative AI and LlamaIndex
Chapter 1: Understanding Large Language Models
	Introducing GenAI and LLMs
		What is GenAI?
		What is an LLM?
	Understanding the role of LLMs in modern technology
	Exploring challenges with LLMs
	Augmenting LLMs with RAG
	Summary
Chapter 2: LlamaIndex: The Hidden Jewel - An Introduction to the LlamaIndex Ecosystem
	Technical requirements
	Optimizing language models – the symbiosis of fine-tuning, RAG, and LlamaIndex
		Is RAG the only possible solution?
		What LlamaIndex does
	Discovering the advantages of progressively disclosing complexity
		An important aspect to consider
	Introducing PITS – our LlamaIndex hands-on project
		Here’s how it will work
	Preparing our coding environment
		Installing Python
		Installing Git
		Installing LlamaIndex
		Signing up for an OpenAI API key
		Discovering Streamlit – the perfect tool for rapid building and deployment!
		Installing Streamlit
		Finishing up
		One final check
	Familiarizing ourselves with the structure of the LlamaIndex code repository
	Summary
Part 2: Starting Your First LlamaIndex Project
Chapter 3: Kickstarting your Journey with LlamaIndex
	Technical requirements
	Uncovering the essential building blocks of LlamaIndex – documents, nodes, and indexes
		Documents
		Nodes
		Manually creating the Node objects
		Automatically extracting Nodes from Documents using splitters
		Nodes don’t like to be alone – they crave relationships
		Why are relationships important?
		Indexes
		Are we there yet?
		How does this actually work under the hood?
		A quick recap of the key concepts
	Building our first interactive, augmented LLM application
		Using the logging features of LlamaIndex to understand the logic and debug our applications
		Customizing the LLM used by LlamaIndex
		Easy as 1-2-3
		The temperature parameter
		Understanding how Settings can be used for customization
	Starting our PITS project – hands-on exercise
		Let’s have a look at the source code
	Summary
Chapter 4: Ingesting Data into Our RAG Workflow
	Technical requirements
	Ingesting data via LlamaHub
	An overview of LlamaHub
	Using the LlamaHub data loaders to ingest content
		Ingesting data from a web page
		Ingesting data from a database
		Bulk-ingesting data from sources with multiple file formats
	Parsing the documents into nodes
		Understanding the simple text splitters
		Using more advanced node parsers
		Using relational parsers
		Confused about node parsers and text splitters?
		Understanding chunk_size and chunk_overlap
		Including relationships with include_prev_next_rel
		Practical ways of using these node creation models
	Working with metadata to improve the context
		SummaryExtractor
		QuestionsAnsweredExtractor
		TitleExtractor
		EntityExtractor
		KeywordExtractor
		PydanticProgramExtractor
		MarvinMetadataExtractor
		Defining your custom extractor
		Is having all that metadata always a good thing?
	Estimating the potential cost of using metadata extractors
		Follow these simple best practices to minimize your costs
		Estimate your maximal costs before running the actual extractors
	Preserving privacy with metadata extractors, and not only
		Scrubbing personal data and other sensitive information
	Using the ingestion pipeline to increase efficiency
	Handling documents that contain a mix of text and tabular data
	Hands-on – ingesting study materials into our PITS
	Summary
Chapter 5: Indexing with LlamaIndex
	Technical requirements
	Indexing data – a bird’s-eye view
		Common features of all Index types
	Understanding the VectorStoreIndex
		A simple usage example for the VectorStoreIndex
		Understanding embeddings
		Understanding similarity search
		OK, but how does LlamaIndex generate these embeddings?
		How do I decide which embedding model I should use?
	Persisting and reusing Indexes
		Understanding the StorageContext
		The difference between vector stores and vector databases
	Exploring other index types in LlamaIndex
		The SummaryIndex
		The DocumentSummaryIndex
		The KeywordTableIndex
		The TreeIndex
		The KnowledgeGraphIndex
	Building Indexes on top of other Indexes with ComposableGraph
		How to use the ComposableGraph
		A more detailed description of this concept
	Estimating the potential cost of building and querying Indexes
	Indexing our PITS study materials – hands-on
	Summary
Part 3: Retrieving and Working with Indexed Data
Chapter 6: Querying Our Data, Part 1 – Context Retrieval
	Technical requirements
	Learning about query mechanics – an overview
	Understanding the basic retrievers
		The VectorStoreIndex retrievers
		The DocumentSummaryIndex retrievers
		The TreeIndex retrievers
		The KnowledgeGraphIndex retrievers
		Common characteristics shared by all retrievers
		Efficient use of retrieval mechanisms – asynchronous operation
	Building more advanced retrieval mechanisms
		The naive retrieval method
		Implementing metadata filters
		Using selectors for more advanced decision logic
		Understanding tools
		Transforming and rewriting queries
		Creating more specific sub-queries
	Understanding the concepts of dense and sparse retrieval
		Dense retrieval
		Sparse retrieval
		Implementing sparse retrieval in LlamaIndex
		Discovering other advanced retrieval methods
	Summary
Chapter 7: Querying Our Data, Part 2 – Postprocessing and Response Synthesis
	Technical requirements
	Re-ranking, transforming, and filtering nodes using postprocessors
		Exploring how postprocessors filter, transform, and re-rank nodes
		SimilarityPostprocessor
		KeywordNodePostprocessor
		PrevNextNodePostprocessor
		LongContextReorder
		PIINodePostprocessor and NERPIINodePostprocessor
		MetadataReplacementPostprocessor
		SentenceEmbeddingOptimizer
		Time-based postprocessors
		Re-ranking postprocessors
		Final thoughts about node postprocessors
	Understanding response synthesizers
	Implementing output parsing techniques
		Extracting structured outputs using output parsers
		Extracting structured outputs using Pydantic programs
	Building and using query engines
		Exploring different methods of building query engines
		Advanced uses of the QueryEngine interface
	Hands-on – building quizzes in PITS
	Summary
Chapter 8: Building Chatbots and Agents with LlamaIndex
	Technical requirements
	Understanding chatbots and agents
		Discovering ChatEngine
		Understanding the different chat modes
	Implementing agentic strategies in our apps
		Building tools and ToolSpec classes for our agents
		Understanding reasoning loops
		OpenAIAgent
		ReActAgent
		How do we interact with agents?
		Enhancing our agents with the help of utility tools
		Using the LLMCompiler agent for more advanced scenarios
		Using the low-level Agent Protocol API
	Hands-on – implementing conversation tracking for PITS
	Summary
Part 4: Customization, Prompt Engineering, and Final Words
Chapter 9: Customizing and Deploying Our LlamaIndex Project
	Technical requirements
	Customizing our RAG components
		How LLaMA and LLaMA 2 changed the open source landscape
		Running a local LLM using LM Studio
		Routing between LLMs using services such as Neutrino or OpenRouter
		What about customizing embedding models?
		Leveraging the Plug and Play convenience of using Llama Packs
		Using the Llama CLI
	Using advanced tracing and evaluation techniques
		Tracing our RAG workflows using Phoenix
		Evaluating our RAG system
	Introduction to deployment with Streamlit
	HANDS-ON – a step-by-step deployment guide
		Deploying our PITS project on Streamlit Community Cloud
	Summary
Chapter 10: Prompt Engineering Guidelines and Best Practices
	Technical requirements
	Why prompts are your secret weapon
	Understanding how LlamaIndex uses prompts
	Customizing default prompts
		Using advanced prompting techniques in LlamaIndex
	The golden rules of prompt engineering
		Accuracy and clarity in expression
		Directiveness
		Context quality
		Context quantity
		Required output format
		Inference cost
		Overall system latency
		Choosing the right LLM for the task
		Common methods used for creating effective prompts
	Summary
Chapter 11: Conclusion and Additional Resources
	Other projects and further learning
		The LlamaIndex examples collection
		Moving forward – Replit bounties
		The power of many – the LlamaIndex community
	Key takeaways, final words, and encouragement
		On the future of RAG in the larger context of generative AI
		A small philosophical nugget for you to consider
	Summary
Index
Other Books You May Enjoy