Skip to main content

How to generate Prethink context with the CLI

Prethink recipes generate structured context that gives AI coding agents a clear, accurate understanding of your entire codebase. In this guide, we will walk you through how to generate this Prethink context locally using the Moderne CLI.

For more background on what Prethink is and the available recipes, please check out our Prethink documentation.

Prerequisites​

This guide assumes that you have already installed and configured the CLI and you are aware of what Moderne Prethink is.

Step 1: Clone the repositories​

The first thing you'll need to do is come up with the list of repositories that you want to generate Prethink context for. Once you have that list, please ensure they are cloned to a shared directory locally. For instance, this might look like:

prethink-demo
├── service-a
├── service-b
└── shared-library
tip

Start with a small subset of repositories to test and iterate quickly. Once everything works, add more repositories as needed.

We recommend using the mod git sync command to create this shared directory. It can clone repositories from a CSV file or from an existing organization.

Step 2: Build the LSTs​

With the repositories cloned, build or download LSTs to run Prethink recipes on:

mod build prethink-demo

Step 3: Install Prethink recipes​

The Moderne module depends on the OpenRewrite module, so install both of these JARs:

mod config recipes jar install org.openrewrite.recipe:rewrite-prethink:LATEST
mod config recipes jar install io.moderne.recipe:rewrite-prethink:LATEST

If you only need the core building blocks for custom discovery recipes, you can install just the OpenRewrite module.

Step 4: Run a Prethink recipe​

You can either run the starter recipe without AI:

mod run prethink-demo --recipe io.moderne.prethink.UpdatePrethinkContextNoAiStarter

Or, if you want AI-generated code comprehension and test summaries, you can run the AI version with your LLM provider credentials:

mod run prethink-demo --recipe io.moderne.prethink.UpdatePrethinkContextStarter \
  -P provider=openai \
  -P apiKey=sk-...

In either case, the recipe will:

  • Discover service endpoints, database connections, external service calls, and messaging patterns
  • Map test methods to implementation methods
  • Generate a CALM architecture diagram
  • Update agent configuration files

After you've run the recipe, make sure to apply the changes so the generated files appear in your working tree:

mod git apply prethink-demo --last-recipe-run

You'll find the generated context in each repository's .moderne/context/ directory. For details on what files are generated, see our documentation on what Prethink generates.

Step 5: Commit and push your changes​

After reviewing the generated context, commit the changes:

mod git commit prethink-demo -m "Add Prethink context for AI agents" --last-recipe-run

Then push to your remote repository:

mod git push prethink-demo --last-recipe-run

Once pushed, create pull requests through your git provider (GitHub, GitLab, etc.) to merge the Prethink context into your main branch.

Creating custom Prethink recipes​

If you want to customize which context is generated, you can create your own Prethink recipe. See creating custom Prethink recipes for examples and guidance.

Once you have a custom recipe, install and run it with:

mod config recipes yaml install MyPrethink.yml
mod run prethink-demo --recipe com.acme.MyCustomPrethink

Keeping context up to date​

Prethink context should be regenerated as your codebase evolves. You should consider:

  • Running Prethink as part of your CI pipeline
  • Scheduling periodic regeneration (e.g., weekly)
  • Regenerating the context after every major refactor or dependency update