Does it include tools for performance testing?

Yes, it includes specialized Python scripts for benchmarking latency, throughput, and cost, as well as evaluating retrieval quality using metrics like MRR and NDCG.

What retrieval frameworks does this skill support?

This skill provides implementation patterns and templates for both LlamaIndex and LangChain, the leading frameworks for building RAG systems.

When should I use hybrid search over semantic search?

Hybrid search is recommended as the default for production systems because it combines the conceptual understanding of vector search with the keyword precision of BM25.

Does this skill help with conversational RAG?

Yes, it includes specific examples for conversational retrieval that handle chat history compression and query rewriting for multi-turn interactions.

How does reranking improve my RAG system?

Reranking acts as a second-stage filter that uses more powerful models to re-order the top results from your initial search, significantly improving the relevance of the context provided to the LLM.

RAG Retrieval Patterns

Name: RAG Retrieval Patterns
Author: vanman2024

byvanman2024

Ciencia de Datos y ML

Implements and optimizes advanced search strategies including semantic, hybrid, and reranking for high-performance RAG systems.

Acerca de

This skill provides a comprehensive framework for building and tuning RAG (Retrieval-Augmented Generation) pipelines using LlamaIndex and LangChain. It offers ready-to-use templates for semantic vector search, hybrid (keyword + vector) search, and cross-encoder reranking, alongside sophisticated benchmarking scripts to measure latency and retrieval quality metrics like Precision, Recall, and NDCG. Whether you are building a prototype or a production-grade system, this skill provides the decision trees and implementation patterns necessary to move beyond basic retrieval to high-precision, context-aware results.

Características Principales

Advanced reranking pipelines using cross-encoders and LLM-based scoring
0 GitHub stars
Comprehensive benchmarking scripts for measuring latency, cost, and throughput
Quality evaluation framework for Precision@k, Recall, MRR, and NDCG
Standardized templates for semantic, hybrid, and multi-query search strategies
Conversational retrieval patterns for context-aware chat history handling

Casos de Uso

Optimizing retrieval relevance for production-grade RAG applications
Implementing complex multi-query expansion and fusion for ambiguous user requests
Benchmarking different embedding models and retrieval architectures to find the best cost-to-quality ratio

Acerca de

Características Principales

Advanced reranking pipelines using cross-encoders and LLM-based scoring
0 GitHub stars
Comprehensive benchmarking scripts for measuring latency, cost, and throughput
Quality evaluation framework for Precision@k, Recall, MRR, and NDCG
Standardized templates for semantic, hybrid, and multi-query search strategies
Conversational retrieval patterns for context-aware chat history handling

Casos de Uso

Optimizing retrieval relevance for production-grade RAG applications
Implementing complex multi-query expansion and fusion for ambiguous user requests
Benchmarking different embedding models and retrieval architectures to find the best cost-to-quality ratio