Progressive Web App

LCSH Cataloging Assistant

A Progressive Web App that combines AI-powered subject analysis with real-time Library of Congress validation to produce accurate LCSH headings and MARC records.

Key Features

🤖

AI-Powered Suggestions

Configurable AI models analyze bibliographic information and suggest appropriate subject headings with detailed reasoning for each recommendation.

Dual Authority Validation

Every suggestion is validated in real time against both LCSH (subjects) and LCNAF (names) databases using the Library of Congress suggest2 API.

📋

MARC Record Generation

Automatically generates properly formatted MARC records with correct tags (650, 600, 610), indicators, and subfields — ready for your ILS.

🖼️

Image Analysis

Upload book covers or title pages. Vision-capable AI models extract additional subject information from visual materials.

🔄

Multi-Provider AI

Bring your own API key from OpenAI, Google Gemini, DeepSeek, Qwen, Z.ai, Kimi, Minimax, OpenRouter (400+ models with free tier), or connect to local OpenAI-compatible endpoints.

📱

Offline PWA

Install on any device — desktop, tablet, or phone. Cached assets load instantly and the app works offline after initial setup.

📊

Similarity Scoring

Levenshtein distance-based scoring shows match quality at a glance with color-coded percentages from 0% (no match) to 100% (exact match).

💾

History & Export

Save sessions to browser storage, review past recommendations in detail, export individual or bulk CSV files, and copy all MARC records at once.

🔒

Privacy-First

Zero-knowledge architecture. All data stays in your browser. API keys are stored locally and masked in the UI.

Supported AI Providers

Bring your own API key from any of these providers, or connect to a local model.

Provider Example Models
OpenAIGPT-5.2, GPT-4.5, o3-pro, o3-mini
Google GeminiGemini 3 Flash, Gemini 3 Pro, Gemini 2.5 Pro, Gemini 2.5 Flash
DeepSeekDeepSeek V3.2, DeepSeek R1
QwenQwen3.5-Plus, Qwen3, Qwen2.5
Z.aiGLM-5
KimiKimi K2.5
MinimaxM2.5, M2.1
OpenRouter400+ models — free tier includes Llama 3.3 70B, Gemini 2.5 Flash, DeepSeek R1
Custom EndpointOllama, LM Studio, vLLM, or any OpenAI-compatible API

How It Works

1

Enter Bibliographic Info

Provide the title, author, abstract, table of contents, and optionally upload book covers or title pages.

2

AI Suggests & Validates

AI analyzes your input and generates LCSH candidates. Each one is validated against the Library of Congress in real time.

3

Export MARC Records

Review validated headings with similarity scores, then copy MARC records to your clipboard or export as CSV.

Complete Tutorial

Step-by-step guide with screenshots covering settings, the 3-step wizard, MARC records, validation scores, history management, and troubleshooting.