Random names are easy to generate badly. What is useful here is filling a sentence shape from several datasets at once — an adjective, a class, a quest and an artifact drawn together — so the output is a coherent prompt rather than four unrelated words. Look at the categories first; everything else takes a category and a dataset name.
A server over the Opera Omnia collection of JSON datasets, meant for games, storytelling and bots. Datasets sit in categories; you can read one whole, pull a random item, filter it, combine several, or fill a template from them.
- list_categories and list_datasets to see what exists before asking for anything
- get_dataset for a dataset's complete contents
- get_random_item for a single random pick from one dataset
- get_filtered_items for the entries matching a criterion
- combine_datasets to draw from several at once and get a set of random selections
- generate_content — a template with placeholders, each filled from a named dataset
- The same content also exposed as resources, so a client can browse categories and datasets by URI
A clone and a build — there is no published package. Node, and network access, since datasets are fetched from the upstream project rather than bundled.
Build from source — clone the repository and build it, then point your client at the binary