server-library

A media library backend that extracts book, music, and movie references from natural language text, looks them up against public APIs, and persists them to a per-user SQLite database.

What it does

When given a Discord username, a media type, and a raw message, the pipeline:

1. Sends the text to the free chatjimmy.ai LLM and asks it to extract entity names as a JSON array
2. Looks each name up against the appropriate public API
3. Upserts the result into SQLite and records the user → media association with a timestamp

Supported media types:

Type	Extraction	API
books	Book titles	Open Library — no key required
music	Artist names	TheAudioDB — no key required
movies	Movie titles	OMDB — free key required

Usage

# Install dependencies
uv sync

# Add media from a message
python main.py add <username> books  "I've been rereading 1984 and Animal Farm"
python main.py add <username> music  "been listening to a lot of Radiohead and Portishead"
python main.py add <username> movies "just watched Parasite and Everything Everywhere All at Once"

# Show a user's library
python main.py show <username>          # all categories
python main.py show <username> books
python main.py show <username> music
python main.py show <username> movies

OMDB requires an API key (free at https://www.omdbapi.com/apikey.aspx):

export OMDB_API_KEY=your_key_here

Static site

The repo now includes an Astro-based static site that reads from library.db at build time.

# Install Astro dependencies
npm install

# Export SQLite data to JSON and build the site
npm run build

Generated files are written to dist/.

The build pipeline is:

1. uv run python export_site_data.py
2. astro build

Connecting to nmemos

nmemos needs to call into this library in two places:

1. /library <type> command handler

When a user runs /library books, /library music, or /library movies, nmemos should grab the content of the previous message and call:

from db import connect, init_db
from library import process_message

conn = connect()
init_db(conn)

result = process_message(conn, discord_username, previous_message_text, media_type)
# result = {"user": "...", "added": [...], "skipped": [...]}

Reply to the user with result["added"] (saved) and result["skipped"] (no API results found).

2. Displaying a user's library

When a user wants to see their saved media:

from db import connect, get_user_books, get_user_artists, get_user_movies

conn = connect()
books   = get_user_books(conn, discord_username)    # list[UserBook]
artists = get_user_artists(conn, discord_username)  # list[UserArtist]
movies  = get_user_movies(conn, discord_username)   # list[UserMovie]

Each entry has a .added_at timestamp and the full API metadata object (.book, .artist, .movie).

The connect() call will use library.db in the project root by default. Pass a Path to connect() if nmemos needs the database at a different location.

Roadmap

• Book extraction + Open Library API
• Music extraction + TheAudioDB API
• Movie extraction + OMDB API
• SQLite database with per-user history
• SSG to generate a static website from the database
• GitHub Action to auto-deploy the website
• nmemos /library command integration