Smart Library Assistant

Calvin University Senior Project

Project Vision

Within the domain of Smart Library Assistant, our project aims to enhance the digital library using Retrieval-Augmented Generation. Researchers, students, and theologians often struggle to find precise answers to their questions amid the large volume of Christian literature books. Traditional methods require reading through large texts and archives, such as the Christian Classics Ethereal Library (CCEL), which, unless you know where to look, is a time-consuming process. Existing search tools lack contextual understanding and may return irrelevant results that do not answer the user's question.

Our solution employs Agentic RAG (Retrieval-Augmented Generation), an advanced approach where Claude acts as an intelligent agent that dynamically adapts its retrieval strategy to each query's needs. Unlike traditional RAG that retrieves a fixed number of passages for every query, our agentic architecture allows the AI to: analyze the user's query to determine retrieval needs, formulate optimized search queries, decide whether retrieved results are sufficient or require additional searches, and synthesize information from multiple retrieval rounds. This approach significantly improves response quality and efficiency.

The key features of our system enhance the user experience and the accuracy of the results. First, it offers a domain-specific search, explicitly tailored to theological texts from CCEL and rigorously tested on 2,716 queries. Second, the system includes citation generation, with every response linking directly to the cited paragraph in the source material. Finally, we built a user-friendly chat interface featuring intuitive filters by author or work and support for follow-up questions within conversational context.

Live Demo

Try the Smart Library Assistant and explore theological texts with AI-powered search.

Smart Library Assistant

Ask questions about Christian classics and receive AI-generated answers with citations to original sources.

Try the Demo

Search

Semantic + Full-text

Filters

By Author/Work

Citations

Linked to Source

Our Team

Meet the talented individuals behind the Smart Library Assistant project.

David Kim

Testing framework, evaluation methodology, dataset generation

Jeton Cesaj

Backend development, RAG pipeline, Manticore integration

Zhonglin (Loya) Niu

Frontend development, Docker deployment, UI/UX design

Project Advisor

Dr. Harry Plantinga

Professor at Calvin University and Director of the Christian Classics Ethereal Library (CCEL).

Project Code

GitHub Repository

Our codebase is hosted on GitHub. You can explore the source code, contribute, or report issues through our organization repositories.

View on GitHub

Data Processing

Tools for processing CCEL documents and creating embeddings

Frontend

User interface for the Smart Library Assistant chatbot

Backend API

RAG implementation and LLM integration with Claude

Project Report

Project Completed

We have successfully delivered a fully functional AI-powered research tool for theological texts. The Smart Library Assistant is now deployed and accessible to users.

Key Features Implemented

  • Agentic RAG architecture that dynamically adapts retrieval strategy to query needs
  • Integration with Anthropic's Claude API and LangChain for intelligent orchestration
  • Semantic search using Manticore Search with vector and full-text capabilities
  • Citation generation with clickable links to source paragraphs in CCEL
  • Filter functionality to narrow searches by author or work
  • Usage statistics tracking (7-day and 30-day periods)

Evaluation Results

2,716

Queries Tested

48.1%

Overall Accuracy

52.8%

Perfect Match Rate

Top 3

Avg Matched Rank

Success Criteria

Criterion Status
Source Attribution Achieved
Retrieval Accuracy Baseline Achieved (48.1%)
Functional MVP Exceeded (Full Features)
Deployment Achieved (CCEL Servers)

Project Presentation

Final Presentation

CS/DATA 398 Senior Project Presentation - December 2025