RAGenesis is an open-source AI-powered tool designed as a reading and exploration platform for religious and classic texts, leveraging Retrieval-Augmented Generation (RAG) to provide interactive, ecumenical engagement with scriptures. 1 2 12 It functions as a “generative book” that compiles wisdom from various traditions, promoting unity and mutual understanding across cultures by allowing users to query and compare texts in innovative ways. 1

Supported Texts

The system supports multiple foundational religious and philosophical texts, enabling cross-tradition exploration:

• Torah
• Bible
• Quran
• Bhagavad Gita
• Analects (of Confucius) 2 12

This selection allows for comparative analysis, such as finding semantic similarities between verses from different sources.

Core RAG Functionality

RAGenesis integrates external knowledge from these texts into AI responses to enhance accuracy, relevance, and dynamic updates. 1 Key aspects include:

• Retrieval Pipelines: It offers three distinct RAG pipelines, each guided by different instruction prompts to the generative model. These pipelines retrieve relevant verses or chunks from the texts based on user queries, then augment the AI’s generation process. 1
• Embeddings and Vector Search: Utilizes Hugging Face’s sentence-transformers/all-MiniLM-L6-v2 model for semantic embeddings, along with Jina AI’s jina-clip-v1 for multimodal capabilities. Embeddings are stored in a Milvus VectorDB for efficient similarity searches. 12
• GraphRAG Elements: Incorporates graph-based methods, defining Semantic Similarity Networks (a subtype of Semantic Networks) to model relationships between text chunks. It applies a “Main Chunk” methodology derived from basic graph theory to identify central or connected ideas, enabling deeper structural analysis beyond simple keyword matching. 2 12

Interactive Capabilities

Users interact via a conversational interface, typically built with Streamlit (in the open-source version), where they can pose questions like “Compare teachings on compassion in the Bible and Bhagavad Gita” or explore specific verses. 12 Features include:

• Query and Response Generation: AI retrieves semantically similar passages, then generates contextual explanations or syntheses.
• Multi-Text Comparison: Supports ecumenical queries that bridge traditions, highlighting shared themes or differences.
• Deployment and Accessibility: The backend is deployed on AWS using CloudFormation and SAM for scalability. The open-source repo allows users to run it locally or customize it. 2 12

Overall, RAGenesis transforms static reading into an AI-assisted, graph-enhanced discovery process, making it a valuable tool for researchers, theologians, or anyone interested in interfaith dialogue. 1