lfai/lfai-landscape: Data Files, Not Software, for the LF AI & Data Project Map

The repository is a dataset, and the README says so

People arrive at lfai/lfai-landscape expecting a web application and find a directory of YAML and SVG files. The README states the position directly: this repository contains the data files and images required to generate the landscape, and the software that generates it lives in cncf/landscape2. So there is no build system here to audit, no dependency tree, no server code. What you get is landscape.yml plus a hosted_logos directory of SVG files, and a pointer to someone else's tool. That split is the first thing to understand, because it determines what a contribution to this repository can and cannot change. You can change which projects appear, what category they sit in, and which logo renders beside them. You cannot change the page layout, the filtering behaviour, the card design, or how the search index is built. Those decisions belong to landscape2, which is shared with the CNCF landscape and other foundation landscapes. The homepage at l.lfai.foundation is the rendered output; the repository is the input.

The inclusion rule is a star threshold, and the README admits it

The stated requirements for a new entry are narrow and worth reading before drafting a pull request. Projects must be open source and hosted on or mirrored to GitHub. AI, ML and DL projects with at least 300 GitHub stars that clearly fit an existing category are generally included. The word generally does the work in that sentence, and the README is honest about the second constraint: put the project in the single category where it best fits, and do not expect a new category, because the maintainers would rather find the best home among the current options. A logo is also required, and the logo has to include the name. The Crunchbase organization field should name the company or organization that controls the software, normally the trademark owner whether or not a trademark has been formally filed. That last rule matters more than it looks. It ties the funding column to a legal entity rather than to a maintainer team, so a project governed by a foundation but commercially backed by one vendor will resolve to that vendor. If your project has no clear controlling entity, the Crunchbase field is where the entry gets murky.

How to build and serve the landscape locally

The README gives two commands, run after installing landscape2. The build step takes the local data file, pulls settings from a URL, reads logos from the local directory, and writes a static site:

landscape2 build --data-file landscape.yml --settings-url https://raw.githubusercontent.com/cncf/landscape2-sites/refs/heads/main/lfai/settings.yml --logos-path hosted_logos --output-dir build

Then serve it:

landscape2 serve --landscape-dir build

Two details in that invocation are easy to miss. The settings file is fetched over the network from the cncf/landscape2-sites repository, not from this repository, so the categories, filters and card fields you see locally come from a file you do not control here. And the output is written to build, a directory that the README does not list as tracked content, which fits the pattern of a repo whose committed artefacts are inputs rather than outputs. If you are reviewing a proposed entry, this pair of commands is the only way the README offers to see the change rendered before it merges.

Edits to member rows are overwritten every night

The most consequential operational detail in the README concerns the category named LF AI & Data Member Company. Those entries are built nightly using the LFX Landscape Tools, and the README warns that any changes made directly in the data files for that category will be overwritten. The correct path is the LFX Organization Dashboard, or a helpdesk ticket if you cannot reach it. This is a real failure mode for contributors: a pull request that edits a member company row can be correct, reviewed, and merged, and still disappear on the next nightly run. The same caution applies to Crunchbase-sourced fields. The README tells you to open a Crunchbase account and edit the data there rather than patching the YAML, and to fix a project description in GitHub rather than in the landscape file. If a licence is displaying incorrectly, the suggested fix is to paste the unmodified licence text into a LICENSE file at the root of the project's own repository so GitHub can serve it correctly. Three different fields, three different systems of record, none of them this repository.

The licence carve-out is the part to read twice

The repository is Apache-2.0, but the README carves out an exception that changes what you can do with the data. The generated landscape contains data received from Crunchbase, and that data is not licensed under Apache-2.0. It is subject to Crunchbase's Data Access Terms and is only permitted to be used with LF AI & Data Landscape projects. Everything else is Apache-2.0, except project and product logos, which are generally copyrighted by the company that created them and are cached here for reliability. The generated landscape and landscape.yml are alternatively available under Creative Commons Attribution 4.0. Read together, those sentences mean the funding and market-cap style fields are the constrained ones, the logos are third-party marks held for rendering, and the rest of the file is permissively licensed. If you were planning to lift the funding columns into an internal dashboard or a comparison site, that is the clause to check with your own counsel. This is a description of what the README says, not legal advice.

landscape2 is the alternative, and it is a different kind of thing

The obvious alternative to using this repository is to use cncf/landscape2 directly with your own data file, and the difference is not a matter of taste. This repository is a curated instance: someone decides which projects qualify, which category each one lands in, and which logo is canonical. Running landscape2 yourself gives you the generator and none of that curation, so you own the inclusion decisions, the category taxonomy, the logo sourcing, and the settings file that this repo fetches from cncf/landscape2-sites. The trade is control against maintenance. A foundation-hosted landscape carries a governance process and a nightly pipeline for member data; a self-hosted one carries neither, and the data goes stale the moment you stop updating it. There is also a third option worth naming: read the hosted site at l.lfai.foundation and never touch the repository at all. For someone trying to work out which open source AI projects sit in which part of the ecosystem, that is often enough, and it costs nothing to maintain.

Maintenance cost falls on the people who care about one row

There are no releases in the material provided, and no versioned artefacts to upgrade, which is consistent with a repository whose content is a data file and a logo directory. The maintenance cost is therefore not a dependency upgrade treadmill. It is the recurring work of keeping entries accurate across systems you do not own: Crunchbase for organisation data, GitHub for descriptions and licence detection, the LFX Organization Dashboard for member rows, and pull requests here for everything else. The nightly rebuild of member data means the repository's own state is partly a derived artefact, so treating a local clone as authoritative will mislead you. For a contributor, the practical loop is short: edit landscape.yml, add an SVG to hosted_logos, open a pull request, and expect the maintainers to weigh your project against the 300-star threshold and the no-new-categories preference. For anyone else, the honest summary is that this repository is a contribution surface, not a tool you install.

Editorial conclusion

Adopt this repo if your job is to add or correct an entry in the LF AI & Data map: edit landscape.yml, drop an SVG in hosted_logos, open a pull request. Do not adopt it as a library, because it ships no code; the generator is cncf/landscape2. Before you spend time on a member-company row, confirm whether it comes from the LFX Organization Dashboard, since nightly runs overwrite direct edits to the LF AI & Data Member Company category. Verify the Crunchbase terms before reusing any funding field outside an LF AI & Data landscape.

Official sources

1. Issues
2. lfai/lfai-landscape on GitHub
3. License: Apache-2.0
4. Project website
5. README