At LGND, our mission is to build the infrastructure that decodes our changing planet — infrastructure that supports the best models and the best sensors for your use case, so you never have to rebuild your stack. Today we're adding a new option to our menu of models: FarSLIP, a vision-language model.

More than one way to search by text

The LGND API already lets you search Clay collections with natural language, by resolving text queries against Clay embeddings for you behind the scenes. FarSLIP adds a second path: a model that's language-native from the ground up, where text and imagery are represented together in a single shared embedding space.

Both models work from the same search-by-text endpoint, and both are there to be chosen per task — so you can reach for whichever fits the data and the question in front of you, without changing how you work.

On the LGND API, FarSLIP is queryable from the same endpoint as every other model. You select it in your ModelSpec, point a search-by-text query at a NAIP embeddings collection, and the rest of your workflow stays exactly where it is.

Often the strongest workflow uses multiple models: start broad with a FarSLIP text query to find candidate locations from a plain-language description, then switch to Clay to refine and scale the search using the best of those hits as examples. Simply swap the ModelSpec argument from farslip to clay between runs. 

Get started

We built LGND to be a sandbox for the planet, where the best models are always at your fingertips, ready to scale. FarSLIP is available on the LGND API today. Create an account at developer.lgnd.ai to begin, and see the API reference for the full ModelSpec and endpoint details.