A few days ago, I wrote about why I think Context Repositories matter.
This time I’d like to talk about the project I’m building around that idea.
It’s called Renma.
Renma is an open-source project, but I don’t think of it as “just another AI tool.”
I see it as an experiment.
An experiment on how we can build, organize, and maintain reusable context for LLMs.
Why I started Renma
While using LLMs every day, I noticed something.
The more I used them, the less time I spent writing complicated prompts.
Instead, I spent more time collecting useful knowledge.
Coding conventions.
Architecture notes.
Review checklists.
Testing strategies.
Design decisions.
These pieces of knowledge were much more stable than prompts themselves.
Unfortunately, they were scattered everywhere.
Some lived in Git.
Some were in Markdown files.
Some were in Confluence.
Some only existed in my memory.
I wanted a better way to organize them.
That became the starting point of Renma.
Context Assets
In Renma, the basic building block is a Context Asset.
A Context Asset is a small piece of reusable knowledge.
Not a huge document.
Not a complete specification.
Just one focused piece of context that can be understood and reused independently.
For example,
- how to review API changes
- naming conventions
- security requirements
- testing guidelines
- project-specific terminology
Each asset should have one responsibility.
Just like we try to keep software components small and focused.
Skills
One Context Asset is useful.
Multiple Context Assets become much more powerful.
That’s where Skills come in.
A Skill is simply a collection of Context Assets designed for a specific task.
For example,
A code review skill might include:
- coding conventions
- architecture principles
- review checklist
- project glossary
Instead of creating a huge prompt every time, the right context can be assembled from reusable assets.
Opinionated by design
One thing I intentionally did with Renma is making it opinionated.
I don’t think every way of organizing context is equally maintainable.
That’s why Renma encourages a specific structure.
It validates relationships between assets.
It checks references.
It reports common mistakes.
In other words, it treats context more like software than documentation.
Some people may prefer complete flexibility.
That’s perfectly fine.
Renma intentionally chooses consistency over flexibility.
More than a prompt library
Sometimes people ask if Renma is a prompt management tool.
I don’t think it is.
Prompts are usually the final output.
Renma focuses on everything before that.
The knowledge.
The structure.
The relationships.
The review process.
If prompts disappear one day because models become good enough, I believe these assets will still be useful.
Still an experiment
Renma is still young.
There are many things I want to explore.
Better validation.
Better tooling.
Integrations with different AI systems.
New ways to organize context.
I don’t know yet what the final shape will look like.
But that’s exactly why I’m building it in the open.
If Context Repositories become an important part of software development in the future, I hope Renma can contribute some ideas to that journey.
~ my tech diary ~ 