RAGO Documentation¶

Welcome to the RAGO (Retrieval Augmented Generation Optimizer) documentation!

📚 Documentation Structure¶

    📁 docs/
    ├── 📁 code_architecture                 # Code Architecture
    │   ├── 📄 overview.md
    ├── 📁 installation                      # Installation guide
    │   ├── 📄 elasticsearch.md
    │   └── 📄 ollama.md
    └── 📁 usage_guide                       # Usage guide
        ├── 📁 dataset                       # Generate and load datasets
        │   ├── 📄 data_loader.md
        │   └── 📄 generator.md
        ├── 📁 evaluation                    # Evaluation and metrics
        │   └── 📄 metrics.md
        ├── 📁 optimization                  # Optimization methods and strategies
        │   ├── 📄 run_experiment.md
        │   └── 📄 tpe.md
        └── 📁 rag                           # RAG concepts, configurations and components
            ├── 📄 rag_concepts.md
            ├── 📄 rag_configuration.md
            ├── 📄 reader.md
            └── 📄 retriever.md

🚀 Getting Started¶

Installation - Setup & ollama configuration
Quick Start - Your first optimization

📖 Core Documentation¶

RAG Concepts - Understanding RAG
RAG Configuration - Parameters & search space
Retriever - Retrieval methods
Reader - Generation strategies

⚙️ Optimization¶

Run Optimization - Optimization parameters and strategies
TPE Algorithm - Bayesian optimization theory

🔧 Evaluation & Datasets¶

Evaluators - Evaluators overview (BertScore, SimilarityScore, LLM-as-Judge)
Metrics - Evaluation metrics
Dataset Loader - Dataset loading and format
Dataset Generator - Dataset generators

🔬 Core Concepts¶

RAG (Retrieval Augmented Generation) combines: 1. Retrieve relevant documents from knowledge base 2. Augment LLM prompt with context 3. Generate informed answers

RAG Optimization automatically finds the best configuration (retriever, embeddings, LLM params) for your use case using the Optuna optimization framework (Bayesian method by default).

→ Learn more: RAG Concepts | Config Space

📖 External Resources¶

Research Papers¶

Tree-structured Parzen Estimator - TPE optimization algorithm
BERTScore - Semantic evaluation metrics
LLM-as-a-Judge - Using LLMs for evaluation

Dependencies¶

Optuna - Hyperparameter optimization framework
LangChain - LLM application framework
LlamaIndex - Data framework for LLMs
Ollama - Run LLMs locally

💡 Need Help?¶

💬 Ask in GitHub Discussions
🐛 Report bugs in Issues