This repository contains a series of complete examples showcasing the powerful mechanics that Nyx offers for data retrieval in Retrieval-Augmented Generation (RAG) applications. These examples demonstrate how Nyx enables sophisticated retrieval capabilities to underpin and enhance RAG workflows.
This example demonstrates a basic chatbot application with a naive implementation of retrieval. It focuses on:
- Retrieval Method: Using a simple metadata-based search, where the chatbot identifies relevant datalinks based on Genres and Categories inferred from the user's query.
- Chatbot Workflow: Captures user queries, retrieves relevant datalinks from Nyx, and provides responses using OpenAI models.
This example is ideal for understanding the foundational mechanics of Nyx in a RAG context.
This example builds upon the Simple RAG chatbot and introduces enhancements to improve retrieval and user interaction:
- Advanced Search and Discovery:
- Leverages Nyx's powerful metadata capabilities for fine-grained discovery, including SPARQL-based queries for advanced discovery.
- Human-in-the-Loop Validation:
- Includes a phase where users validate retrieved documents to ensure relevance and accuracy before they are analysed.
- Enhanced Chatbot Workflow:
- Supports iterative question-and-answer sessions, allowing users to refine or expand their queries dynamically. This example demonstrates how Nyx can be used to build robust and interactive RAG applications for production-level scenarios.