RAG from Scratch

Lesson 12

RAG From Scratch Part 13 (RAPTOR)

This lesson demonstrates building a Retrieval Augmented Generation (RAG) system using LangChain's innovative RAPTOR method. RAPTOR creates a hierarchical index of document summaries, enabling efficient information retrieval from massive datasets and overcoming typical LLM token limitations.

7mFeb 16, 2025

Lesson 11

RAG from scratch Part 12 (Multi-Representation Indexing)

This lesson teaches how to build efficient Retrieval Augmented Generation (RAG) systems using multi-representation indexing. By creating and storing optimized document summaries alongside full documents, the lesson demonstrates faster retrieval and improved LLM performance for autonomous agents.

Lesson 10

RAG from scratch Part 11 (Query Structuring)

This lesson teaches you how to build effective query analyzers for retrieving information from diverse databases using LangChain. By leveraging LLMs and structured query schemas, you'll learn to translate natural language questions into optimized database queries for efficient information retrieval.

Lesson 9

RAG from scratch Part 10 (Routing)

This lesson teaches you how to build intelligent question routing systems using Langchain, leveraging LLMs for both logical and semantic routing. By combining structured outputs from LLMs with various databases and embedding techniques, you'll learn to efficiently direct queries to the most relevant data sources for accurate and fast responses.

7mFeb 16, 2025

All lessons

Lesson 1

RAG From Scratch Part 1 (Overview)

This lesson teaches you to build Retrieval Augmented Generation (RAG) pipelines using LangChain, covering everything from basic concepts to advanced techniques like query optimization and multi-representation indexing. Master practical RAG implementation with LangChain, improving LLM performance by connecting them to external data sources for enhanced context and accuracy.

Lesson 1

RAG From Scratch- Part 2 (Indexing)

This lesson demonstrates building a Retrieval Augmented Generation (RAG) system using Langchain, covering indexing techniques like document loading, splitting, and embedding. It emphasizes the crucial role of converting text into numerical vectors for efficient semantic search and retrieval, showcasing practical implementation with code examples and troubleshooting tips.

4mFeb 16, 2025

Lesson 2

RAG From Scratch Part 3 (Retrieval)

This lesson demonstrates building a Retrieval Augmented Generation (RAG) system using LangChain, covering indexing, embedding, and similarity search for efficient document retrieval. The process is visualized using LangSmith, showcasing the workflow from question to answer, highlighting the importance of task decomposition and LangChain's various integrations.

Lesson 3

RAG From Scratch Part 4 (Generation)

This lesson shows how to build a Retrieval Augmented Generation (RAG) pipeline using Langchain, focusing on prompt engineering and LLM integration for efficient question answering. The process is demonstrated step-by-step, from document preparation and embedding to prompt creation and answer generation using a Langchain chain.

Lesson 4

RAG from scratch Part 5 (Query Translation - Multi Query)

This lesson demonstrates how to significantly improve Retrieval Augmented Generation (RAG) systems by employing multi-query techniques. By transforming single questions into multiple perspectives, the lesson showcases how to enhance retrieval accuracy and overcome limitations of traditional semantic search, leveraging LangChain and various vector databases for efficient parallel processing.

Lesson 5

RAG from scratch Part 6 (Query Translation -- RAG Fusion)

This lesson demonstrates building a robust Retrieval Augmented Generation (RAG) system using Langchain, focusing on improving retrieval accuracy through query translation and fusion techniques. The system intelligently routes queries to various databases, optimizes chunk sizes, and employs advanced ranking and refinement methods to deliver precise and efficient answers.

Lesson 6

RAG from scratch Part 7 (Query Translation -- Decomposition)

This lesson demonstrates building robust Retrieval Augmented Generation (RAG) pipelines using LangChain, focusing on advanced query decomposition techniques like Least-to-Most prompting and RAG fusion to improve accuracy and efficiency in complex question answering. The lesson covers various database interactions, LLM integration, and practical implementation using Python and LangChain, showcasing a complete workflow from question decomposition to final answer synthesis.

Lesson 7

RAG from scratch Part 8 (Query Translation -- Step Back)

This lesson demonstrates how "step-back prompting" improves Retrieval Augmented Generation (RAG) systems by reformulating complex questions into more abstract forms for better context retrieval. The process, visualized with flowcharts and code examples using Langchain, enhances Large Language Model (LLM) accuracy through a multi-stage query translation and dual retrieval approach.