My RAG (Retriever-Augmented Generation) Example

This codebase demonstrates a production-quality, object-oriented approach to mitigating the "lost in the middle" effect when providing long context inputs to GPT-4.

Overview

• Object-Oriented Design:
  The codebase uses abstract base classes and concrete implementations for embeddings, vector stores, transformers, LLMs, and chains.
  The "lost in the middle" mitigation is implemented as a document transformer that can be easily replaced or extended.

• Modular Codebase:
  Classes and logic are separated into modules, grouped by functionality.

• Documentation & Logging:
  Comprehensive docstrings are provided, and logging is used throughout.

• Exception Handling:
  Potential errors (e.g., retrieval issues, API failures) are handled gracefully.

• "Lost in the Middle" Mitigation:
  Implemented via a LongContextReorder transformer. Other strategies can be added by extending BaseDocumentTransformer.

Usage

1. Install dependencies:

   • pip install openai
   • Set your OpenAI API key as an environment variable: export OPENAI_API_KEY=your_key_here.

2. Run the example:

   python main.py