Generative AI with LangChain

Table of contents

1. Preface
   1. Who this book is for
   2. What this book covers
   3. To get the most out of this book
      1. Work with notebooks and projects
   4. Get in touch
2. What Is Generative AI?
   1. Introducing generative AI
      1. What are generative models?
      2. Why now?
   2. Understanding LLMs
      1. How do GPT models work?
         1. Transformers
         2. Pre-training
         3. Tokenization
         4. Conditioning
      2. How have GPT models evolved?
         1. Model size
         2. The GPT model series
         3. PaLM and Gemini
         4. Llama and Llama 2
         5. Claude 1–3
         6. Mixture of Experts (MoE)
      3. How to use LLMs
   3. What are text-to-image models?
   4. What can AI do in other domains?
   5. Summary
   6. Questions
3. LangChain for LLM Apps
   1. Going beyond stochastic parrots
      1. What are the limitations of LLMs?
      2. How can we mitigate LLM limitations?
   2. What is an LLM app?
   3. What is LangChain?
   4. Exploring key components of LangChain
      1. What are chains?
      2. What are agents?
      3. What is memory?
      4. What are tools?
   5. How does LangChain work?
   6. LangChain package structure
   7. Comparing LangChain with other frameworks
   8. Summary
   9. Questions
4. Getting Started with LangChain
   1. How to set up the dependencies for this book
   2. Exploring cloud integrations
      1. Environment setup and API keys
         1. OpenAI
         2. Hugging Face
         3. Google
   3. Building blocks for LLM interaction
      1. LLMs
         1. Fake LLM
      2. Chat models
      3. Prompts
      4. Chains
      5. LangChain Expression Language
      6. Text-to-Image
         1. Dall-E
         2. Replicate
      7. Image understanding
   4. Running local models
      1. Hugging Face Transformers
      2. llama.cpp
      3. GPT4All
   5. Prototyping an application for customer service
      1. Sentiment analysis
      2. Text classification
      3. Document summarization
   6. Applying map-reduce
   7. Monitoring token usage
   8. Summary
   9. Questions
5. Building Capable Assistants
   1. Answering questions with tools
      1. Tool use
      2. Defining custom tools
         1. Tool decorator
         2. Subclassing BaseTool
         3. StructuredTool dataclass
      3. Error handling
   2. Implementing a research assistant with tools
      1. Building a visual interface
      2. Exploring agent architectures
   3. Extracting structured information from documents
   4. Mitigating hallucinations through fact-checking
   5. Summary
   6. Questions
6. Building a Chatbot Like ChatGPT
   1. What is a chatbot?
   2. From vectors to RAG
      1. Vector embeddings
         1. Embeddings in LangChain
      2. Vector storage
         1. Vector indexing
         2. Vector libraries
         3. Vector databases
      3. Document loaders
      4. Retrievers in LangChain
         1. kNN retriever
         2. PubMed retriever
         3. Custom retrievers
   3. Implementing a chatbot with a retriever
      1. Document loaders
      2. Vector storage
      3. Conversation Memory: Preserving Context
         1. ConversationBufferMemory
         2. ConversationBufferWindowMemory
         3. ConversationSummaryMemory
         4. ConversationKGMemory
         5. CombinedMemory
         6. Long-term persistence
   4. Moderating responses
      1. Guardrails
   5. Summary
   6. Questions
7. Developing Software with Generative AI
   1. Software development and AI
      1. Code LLMs
   2. Writing code with LLMs
      1. Vertex AI
      2. StarCoder
      3. StarChat
      4. Llama 2
      5. Small local model
   3. Automating software development
      1. Implementing a feedback loop
      2. Tool use
         1. Error handling
      3. Finishing touches to our developer
   4. Summary
   5. Questions
8. LLMs for Data Science
   1. The impact of generative models on data science
   2. Automated data science
      1. Data collection
      2. Visualization and EDA
      3. Preprocessing and feature extraction
      4. AutoML
   3. Using agents to answer data science questions
   4. Data exploration with LLMs
   5. Summary
   6. Questions
9. Customizing LLMs and Their Output
   1. Conditioning LLMs
   2. Methods for conditioning
      1. Reinforcement learning with human feedback
      2. Low-rank adaptation
      3. Inference-time conditioning
   3. Fine-tuning
      1. Setup for fine-tuning
      2. Open-source models
      3. Commercial models
   4. Prompt engineering
      1. Prompt techniques
         1. Zero-shot prompting
         2. Few-shot learning
         3. CoT prompting
         4. Self-consistency
         5. ToT
   5. Summary
   6. Questions
10. Generative AI in Production
    1. How to get LLM apps ready for production
    2. How to evaluate LLM apps
    3. Comparing two outputs
       1. Comparing against criteria
       2. String and semantic comparisons
       3. Running evaluations against datasets
    4. How to deploy LLM apps
       1. FastAPI web server
       2. Ray
    5. How to observe LLM apps
       1. Tracking responses
       2. Observability tools
       3. LangSmith
       4. PromptWatch
    6. Summary
    7. Questions
11. The Future of Generative Models
    1. The current state of generative AI
       1. Challenges
       2. Trends in model development
       3. Big Tech vs. small enterprises
       4. Artificial General Intelligence
    2. Economic consequences
       1. Creative industries
       2. Education
       3. Law
       4. Manufacturing
       5. Medicine
       6. Military
    3. Societal implications
       1. Misinformation and cybersecurity
       2. Regulations and implementation challenges
    4. The road ahead
12. Other Books You May Enjoy
13. Index