# rag-from-scratch

**Repository Path**: summry/rag-from-scratch

## Basic Information

- **Project Name**: rag-from-scratch
- **Description**: lang-chain
https://github.com/langchain-ai/rag-from-scratch.git
- **Primary Language**: Python
- **License**: Not specified
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-01-09
- **Last Updated**: 2025-01-09

## Categories & Tags

**Categories**: Uncategorized

**Tags**: langchain

## README

# RAG From Scratch

LLMs are trained on a large but fixed corpus of data, limiting their ability to reason about private or recent information. Fine-tuning is one way to mitigate this, but is often [not well-suited for facutal recall](https://www.anyscale.com/blog/fine-tuning-is-for-form-not-facts) and [can be costly](https://www.glean.com/blog/how-to-build-an-ai-assistant-for-the-enterprise).
Retrieval augmented generation (RAG) has emerged as a popular and powerful mechanism to expand an LLM's knowledge base, using documents retrieved from an external data source to ground the LLM generation via in-context learning. 
These notebooks accompany a [video playlist](https://youtube.com/playlist?list=PLfaIDFEXuae2LXbO1_PKyVJiQ23ZztA0x&feature=shared) that builds up an understanding of RAG from scratch, starting with the basics of indexing, retrieval, and generation. 
![rag_detail_v2](https://github.com/langchain-ai/rag-from-scratch/assets/122662504/54a2d76c-b07e-49e7-b4ce-fc45667360a1)