Most software directories start with a tree: Software → Development → DevOps → CI/CD. It works beautifully for the first hundred products. Then you add GitHub, and the tree starts to creak.
GitHub is a code-hosting platform. It is also a collaboration tool, a CI/CD service, an automation ecosystem and a developer platform. Put it in one folder and users looking in the other four never find it. Put it in all five and you now have five copies, five sets of reviews and five records to keep in sync.
One canonical record, five ways in
BestToolsX solves this by keeping one canonical record per offering and describing it along five independent dimensions:
- Entity type — what is it? SaaS product, API, library, framework, platform, CLI tool, AI model and so on. A listing has one primary type and may carry others.
- Category — what problem does it solve? A governed hierarchy of domains (like Cloud & Infrastructure) and subcategories (like Kubernetes).
- Attributes — what features does it have? Shared core fields plus type-specific ones: an API has authentication methods and rate limits; a library has a package manager and dependencies.
- Use cases — who needs it and why? Jobs to be done, such as code review or release automation.
- Relationships — how does it connect? Part of, integrates with, depends on, alternative to, competes with, built with, SDK for.
Type, category and attributes are different things. Mixing them is the most common reason catalogs become impossible to navigate.
Why “type” and “category” must stay separate
“Python library” is not a category. It combines a type (library) with an attribute (language: Python). If you model it as a category you end up with “Python libraries”, “JavaScript libraries”, “Go libraries” and a combinatorial explosion that nobody can maintain. Keep the dimensions separate and the same result comes from a filter.
Stable IDs, changeable names
Every category has a stable identifier such as cloud.kubernetes. Names, synonyms and descriptions can evolve — “Container Orchestration” might become “Kubernetes” — without breaking links, filters or the listings attached to it. The taxonomy borrows from the W3C SKOS model of concepts, preferred and alternative labels, and broader/narrower/related links.
Relationships are where it gets useful
Once relationships are first-class data, the catalog can answer questions a flat directory can’t: What are the alternatives to this tool? What does my stack depend on? Which monitoring products integrate with Kubernetes? Amazon EKS can be “part of” AWS and “built with” Kubernetes at the same time — and the graph knows both.
Designed for growth
We would rather design the universal data model now than rebuild later. Going from a thousand listings to millions should mean expanding the taxonomy, attributes, data sources and search infrastructure — not starting again.
See the model in action on the GitHub sample listing, or read the full architecture overview.
Part of the BestToolsX editorial team, writing about software discovery, developer tools and technology catalogs.