TIPS: Using LLMs on OSLC projects

In this post, I will give a few examples of techniques that work with LLMs when working on the OSLC/Lyo-related codebases as of 2025-08. If you have tips of your own - post them below, this is intended to be a living document.

Use the docs, Luke

We keep most of our docs in a separate repo. If you concatenate all Markdown sources of the docs site, you will get an up-to-date context for an LLM (35k tokens as of 2025-08):

Gitingest

https://uithub.com/OSLC/developing-oslc-applications/tree/gh-pages/docs-new/eclipse_lyo?accept=text%2Fhtml&maxTokens=50000&ext=md (seems to be down at the moment)

Build an up-to-date context of the domains

While the frontier models produce a relatively good general code (request this, update that), they are prone to getting lost and hallucinating on specific methods in the domain model. You can build the context using the following commands:

# inside lyo-domains
mvn clean compile
find ./target/classes/ -name '*.class' -exec javap -s {} \; | tee llm.txt

The output produced from a recent 7.0.0-SNAPSHOT tree is uploaded here: