Reminder

Quran, hadith and names of Allah in one app and API.

Overview

The Reminder is an app and API for the Quran, Hadith and Names of Allah. It provides search and query answers using an LLM. RAG (Retrieval Augmented Generation) and references ensures answers are grounded in authentic Islamic texts.

The goal is to consolidate these texts into a single app and API, providing helpful daily reminders using web push notifications and leveraging AI as a tool for search summaries.

Contents

• Features
• Install
• Usage
• API
• Web
• Notes
• Sources

Features

• Quran in English & Arabic
• Quran Audio Recitation (Arabic & English)
• Names of Allah & Meaning
• Hadith (Bukhari) in English
• Index & Search with RAG and LLMs
• Optional Claude, Ollama or OpenAI integration
• API to query Quran, hadith, names
• Daily reminder web notifications

Install

Find the latest release

Or Go install

go install github.com/asim/reminder@latest

Usage

Quick Start (requires OpenAI API key for fast embeddings):

export OPENAI_API_KEY=xxx
reminder --serve

Alternative: Fully Local (slower, requires Ollama):

# Install and start Ollama (https://ollama.ai/)
ollama pull nomic-embed-text  # For embeddings
ollama pull gemma3:1b          # For LLM responses

# Don't set OPENAI_API_KEY - will use Ollama for both
reminder --serve

LLM Configuration (optional) - used for answer summaries:

# OpenAI (default: gpt-6-luna)
export OPENAI_API_KEY=xxx
export OPENAI_MODEL=gpt-6-luna  # optional override

# Use a specific Ollama model (default: gemma3:1b)
export OLLAMA_LLM_MODEL=gemma3:1b

# Use Claude when OpenAI and an explicit Ollama model are not configured
export ANTHROPIC_API_KEY=xxx

An explicit Ollama model takes priority, then OpenAI, then Anthropic. OpenAI API credentials are required; a ChatGPT/Codex subscription is not an API key.

The reader links back to the full chapter at the selected verse.

Summaries are optional. With no model configured the summarise controls are hidden and search still returns references.

Embedding Configuration:

For best performance, use OpenAI embeddings (fast & cheap - $0.02/1M tokens):

export OPENAI_API_KEY=xxx  # Uses text-embedding-3-small (fast, 1536 dims)

For local/offline (slower, but no API costs):

# Install Ollama first: https://ollama.ai/
ollama pull nomic-embed-text

# Optional: use different model or instance
export OLLAMA_EMBED_MODEL=nomic-embed-text  # default
export OLLAMA_BASE_URL=http://localhost:11434/api  # default

Migration Note: If switching embedding providers, delete the old index cache:

rm -rf ~/.reminder/data/reminder.idx.gob.gz

Run the server

reminder --serve

Go to localhost:8080

API

All queries are returned as JSON

• /api/latest - for the latest reminder with LLM-generated contextual message
  • Returns: verse, hadith, name of Allah, links, and an AI-generated spiritual message
  • The message field contains an LLM-generated reflection (2-3 sentences) based on the verse, hadith, and name
  • Falls back to default message ("In the Name of Allah—the Most Beneficent, Most Merciful") if LLM is unavailable
• /api/quran - to get the entire quran
• /api/names - to get the list of names
• /api/hadith - to get the entire hadith
• /api/search - to get summarised answer
  • q param for the query
  • POST using content-type as application/json
  • curl -d '{"q": "what is islam"}' http://localhost:8080/api/search

See /api for more details

App

The reminder bakes in a "lite" app by default. This can be replaced by a featureful react app.

To build the react app

# requires pnpm
make setup

Build the app

make build

Pass the --web flag to use the react app

reminder --web --serve

Notes

The Quran says in 6:90

Say,
“I ask no reward of you for this (Quran) —
it is a reminder to the whole world.”

Sources

See the Arabic source audit for verification scope and limitations.

The sources used for data

• Arabic Quran
• English Quran
• Names of Allah
• Sahih Bukhari
• Word by word

Development

File an issue to discuss

Offline reading

The web app automatically saves the Quran (including Arabic, translation, word-by-word text, transliteration and stored commentary), Bukhari and names on this device in the background while online. The initial download needs a connection and resumes after an interruption. Texts and the app shell survive reloads and browser restarts, subject to the browser's storage limits and eviction policy.

Saved reading modes, bookmarks and unopened chapters work offline. Audio, server search and AI features need a connection. Unsaved content shows a reconnect message instead of waiting indefinitely. Previously viewed daily reminders are cached, but new reminders require a connection.

npm run build generates the versioned service worker after building all route bundles. Deploy the complete web/build/client output (as make web/make build already do). Development mode does not install an offline cache. Production must use HTTPS (or localhost) for service workers.