SifterSearch Documentation

The search and research intelligence behind OceanLibrary.com

SifterSearch is the backend search and AI research engine that powers OceanLibrary.com — a free interfaith library for scholars, seekers, and researchers exploring sacred texts from the world's religious traditions. Built on a multi-agent AI architecture, it goes beyond simple keyword matching to understand the deeper meaning behind your questions.

46,503 Documents
4.9M Passages
8 Languages
11 Traditions

What Makes SifterSearch Different

Traditional search engines find words. SifterSearch finds meaning.

  • Semantic Understanding — Understands concepts, not just keywords. Search for "detachment from material things" and find passages about spiritual freedom, even if they don't use those exact words.
  • Cross-Language Search — Semantic search operates across language boundaries. An English query finds Arabic passages; a Persian query finds English ones. Learn how →
  • RAG-Enhanced Retrieval — Advanced query disambiguation, hypothetical passage expansion (HyPE), and cross-encoder reranking surface the most relevant passages — not just the most popular ones. See the strategy →
  • Cross-Tradition Discovery — Explore how different religions approach the same themes. Find connections across Baha'i writings, Islamic texts, and Buddhist sutras.
  • AI-Powered Analysis — Each search result is scored, summarized, and highlighted by AI to show you why it's relevant to your query.
  • Scholarly Focus — Built for serious study with proper citations, source attribution, and access to original texts.

Core Features

Hybrid Search

Combines traditional keyword search with semantic vector search for best-of-both-worlds results. Find exact quotes when you need them, or explore conceptually related passages.

Advanced RAG Pipeline

Beyond basic retrieval, SifterSearch uses a multi-step enhancement strategy: query disambiguation resolves ambiguous terms, HyPE generates hypothetical ideal passages to improve semantic matching, and a cross-encoder reranks results for precision. Full strategy details →

Multi-Agent System

Under the hood, specialized AI agents collaborate to understand your query, design search strategies, analyze results, and present findings. Learn more about our agents →

Comprehensive Library

46,503 documents and 4.9 million passages spanning 12 religious traditions in 8 languages. Explore the library →

Developer API

Build your own applications on top of SifterSearch with our public REST API. Full AI analysis, filtering, and rate limiting included. View API docs →

Quick Start

  1. Ask a question — Type naturally: "What does Baha'u'llah say about justice?"
  2. Review results — Each passage shows its relevance score and key phrases highlighted
  3. Explore deeper — Click any result to see the full context and document
  4. Refine your search — Filter by religion, collection, or language

Architecture Overview

Query Intent Classification Hybrid Search keyword + semantic Cross-Encoder Reranking AI Summary + Results

Getting Help

  • Questions? — Use the search bar to ask Sifter directly
  • Bug reportsGitHub Issues
  • Feature requests — We'd love to hear your ideas