Find a Mathlib lemma with AI
Search Lean declarations by mathematical shape, inspect close matches and get the import and name needed to try the result in your project.
Find a Mathlib lemma with AI when the obvious search term comes back empty. An ordering fact you learnt for real numbers might live under a more general type. Start with what has to follow from which hypotheses. The lean-search skill follows those clues through declarations, module names and the code around promising hits.
A useful answer gives you something you can try. Ask for the full declaration name, its import and the assumptions under which it holds. If a close match needs an extra premise, you can decide whether you have that premise before building a proof around it.
What it does
- Search by the shape of a statement as well as by its name.
- Read the source around a likely result for related declarations.
- Locate the module used by a working example.
- Identify the precise change that would make a near match usable.
- Record the search paths when the library appears to lack a result.
Install and use
- Give the agent access to the permitted Lean 4 project and the Mathlib version selected by that project.
- Ask, “Does the library already prove this implication? Compare the assumptions of the two nearest results and include their imports.”
- Try the suggested declaration in context. If nothing matches, retain the closest result and the proposed general statement for a follow-up search.
FAQ
Source
Made by Texra (texra-ai). Original skill: lean-search, published on GitHub under the MIT licence.
View on GitHub