Getting Started

OpenResearch is a workbench for research. Stack experiments off a baseline, run them on your compute, and compare results.

1. Create a project

Start from one of our templates (pre-wired with a baseline and run command) or bring your own GitHub repo.

2. Press Run

Open the baseline experiment, point it at an instance, and press Run. OpenResearch checks out the branch, executes your run command, and produces an EVAL.md plus a code diff.

3. Stack experiments

Once you have a baseline result, stack a child experiment off it to test a variation, then stack another off the best result to go deeper. That's the loop. See Experiment Flow for the full lifecycle.