Skip to main contentThis guide covers the fastest path to a first evaluation run in Evaris.
Before you start
- Access to the Evaris platform
- At least one dataset or a small set of test prompts
- A solver and scorer configuration for your use case
1. Create a project
Create a project in the Evaris app to group datasets, suites, and runs.
2. Add a small dataset
Start with a small dataset, usually 5 to 20 samples, so you can iterate quickly and inspect failures easily.
Choose:
- A solver that represents your generation or agent flow
- One or more scorers aligned to the quality signals you care about
4. Launch the run
Submit the run from the platform.
5. Inspect results
Review:
- Run status and aggregate scores
- Sample-level outputs
- Event and trace artifacts for debugging
Local development
If you are self-hosting Evaris from this monorepo, see the platform and runtime setup in the repository-level engineering docs. 
