Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
61 changes: 61 additions & 0 deletions CONTRIBUTING.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,61 @@
This document is the beginning of an LLM contribution policy.

Areas we want to address:
1. Pull Requests
2. Code Review
3. Feature Development
4. Writing/Communication
5. Testing
6. Community

```
You are responsible for your code. It represents your work. Use whichever tools you want, but you need to understand your code, be able to explain and answer questions, and understand the choices being made.
```

## Pull Requests

If you use LLMs for a significant portion of the work, it is helpful to just say so in the PR description, and how. It can help us guide you especially if you're a new contributor. Or even a seasoned contributor.

You are accountable for your PR, and have clear understanding of the work.
A successful PR will be understood by other people.

We like to keep PRs small when possible. Stacked PRs are a great way to break up your work so that your work can be given a proper review.

It's good to start small if you're new so we can establish trust. It also helps if you start by establishing a plan and getting our feedback on it early.

## Code Review

Generally, LLMs are taking the shortest path to meet an objective. Which means they can cut corners.
##### Reviewing others' PRs
LLMs are great for summarizing commits, or walking you through the changes. Use your own judgement when you make comments, however. Don't paste LLM review output. We can do LLM review without your help.

Asking questions to learn about what is being done is an amazing way to review code. It helps all of us build a shared understanding, and it helps the author build more confidence in their choices.

##### Responding to review
Reply in your own words. Please don't feed the reviewer comments into LLM and paste it back out. It's frustrating and a big waste of time.

If someone is asking you questions, help them understand where you're coming from. Don't assume bad faith.

## Picking a Task
It's good to look through the open issues to see where we need help, or open a new issue.

If you want to be assigned you can either ask on the issue, or reach out to us on the discord in the #libprocessing channel. As mentioned above, it's really great if you do some planning and let us offer early feedback.

## Writing/Communication

Writing should be done by you if possible. Imperfect english is perfectly wonderful, but we understand the need to translate from another language and will not stop you. Try to make sure the meaning is your own and try not to let the LLM add or subtract from your voice.

Outside of that specific case, if you are just asking an LLM to do some work and then pasting output, then we can just do that ourselves. Think about what human value you're adding, because that is what we want.

## Testing

Tests help us to create shared understanding for both humans and LLMs so we can ensure high code quality and mitigate regressions. We have visual tests, unit tests, and have plans for performance benchmarking.

LLMs can be helpful for generating tests. But do make sure you've reviewed them and make sure they are actually testing something useful. Again, LLMs are taking the shortest path so don't let them cut corners!

## Community
We prioritize community as we always have. We are here because of community and we stay for it. However, LLMs do change how work is done, it's changing the nature of the medium and we want to accommodate the new ways in which people contribute and enter into the world of software development and computing.

If you are someone who wants to truly understand how these systems fit together then speak to us in the #libprocessing or #devs-chat discord channels and someone will happily guide you.

We are so happy to have you.
Loading