This project is an AI-powered theological reasoning engine designed to answer questions from structured belief systems using Retrieval-Augmented Generation (RAG). The goal is to build a machine that reasons from source material — not general knowledge — with full theological integrity.
- Represent the belief system faithfully
- Isolate and control knowledge by canon, group, and author
- Let users query within a belief context and receive context-faithful answers
- Enable structured commentary, user annotations, and viewpoint overlays
- Modular chunking pipeline
- Vector DB powered retrieval system
- Group-specific LLM prompts
- Expandable to new traditions without compromising others
/data/ - Raw texts (Bible, etc.)
/chunks/ - Exported, processed JSONL files
/scripts/ - Python tools (chunking, embedding, enrichment)
/prompts/ - Prompt templates by group/perspective
/docs/ - Schemas, source maps, design notes
README.md - This file
PROJECT_TRACKER.md - Feature roadmap, todos
chunk_schema.json - JSON schema for chunk consistency