Versalist Cloud
Evaluate agent behavior with a Versalist-operated application on Google Cloud Run in us-central1.
Agent evaluation and evidence
Run a skill and model against challenge cases. Review rubric scores and public-case results. When capture and access are enabled, inspect available trace metadata.
Explore results and tracesWhat runs where
- Web application
- Google Cloud Run, us-central1
- Builds and container images
- Google Cloud Build and Artifact Registry, us-central1
- Model-provider credentials you add
- Google Secret Manager, one secret per user and provider
- Application database and sign-in
- Supabase: managed Postgres with row-level security, and Supabase Auth
- Background jobs
- Inngest
- Platform default model
- Google Gemini API. Workspaces can add their own providers, including Vertex AI in their own project.
- Sandboxed code execution
- Not generally available
You provide
- An account for each person who signs in.
- Provider credentials, if runs should use your own model accounts.
Versalist provides
- Operation, upgrades, and logs for the application.
- Credential storage in Secret Manager. Deleting a credential destroys its secret.
- Workspace isolation through Postgres row-level security.
Current limits
- Hosting regions, subprocessors, retention, and service levels are not yet verified for procurement.
- No uptime commitment or certification is claimed.
- Application data is stored by Supabase. Its hosting region is not yet verified.
- Sandboxed code execution is not generally available.
How to start
- 1Create an accountSign up with email. The developer platform is free.
- 2Add a model providerUse the platform default model, or add credentials for your own provider.
- 3Run a challengeRun on a challenge page, or connect the CLI to work from your repository.
Related documentation
Other deployment options
- Your Google CloudPlanned
- On-premisesIn development
- Local developmentAvailable