R Rubric

Settings

The Settings screen is small on purpose. There are three things to configure, and they all live in your local Hive database — nothing in Settings is sent over the network unless and until you make an AI-powered call.

Provider

Pick which AI provider Rubric should call for interview generation and report drafting.

  • Anthropic (recommended) — Claude family. Strong at structured Markdown output and long-form reasoning.
  • OpenAI — GPT family. Also fine; pick this if you already have an OpenAI key handy.

You can switch providers at any time. The change applies to the next AI call; previous reports keep whatever provider drafted them.

If you don’t want any AI features, leave the provider unset (or the API key blank). Rubric runs fully without one — the Generate screen and the auto-draft step on End Session are simply unavailable.

API key

Paste the key for the selected provider.

Where the key is stored:

  • The key is persisted in the local Hive database at the platform’s standard application-support directory.
  • It is not synced across machines, not uploaded anywhere, not included in generated interviews or reports.
  • It is sent only as the Authorization header on requests to the provider’s own API.

If you commit the project to a public repo, the Hive database is outside the repo by default. Don’t paste keys into source files.

Model

Optional. Each provider has a default model that Rubric uses if you leave this blank:

  • Anthropic default: a recent Claude Sonnet — fast, strong on structured output.
  • OpenAI default: a recent GPT model.

Override the default if you want to:

  • Trade speed for quality — pick a more capable model for harder rounds.
  • Trade quality for cost — pick a smaller model for rapid iteration on prompts in the Generate screen.
  • Test a new release — when a provider ships a new model, point Rubric at it without waiting for an app update.

The model field accepts the provider’s exact model identifier (e.g., claude-sonnet-4-6, gpt-4o). Wrong identifiers fail at call time with a clear error.

Default storage

Pick where generated interviews, sessions, and reports get persisted. The three options:

  • Local filesystem — a directory you choose. Simplest. Good for solo use.
  • Git repository — a directory that’s a checked-out git repo. Generated content can be auto-committed. Good for team sharing with history.
  • AWS S3 — an S3 bucket you have credentials for. Good for sharing across machines without a git repo.

Each backend has its own configuration fields (path, repo URL, bucket + region + remote name, etc.). See Storage Backends for the trade-offs and detailed setup.

You can change storage later. Past sessions and reports stay where they were written; the new backend applies only to content created after the change.

What is not in Settings

For comparison, here is what Rubric deliberately doesn’t ask you to configure:

  • Theme — the letterpress brand is intentional, not a preference.
  • Telemetry — Rubric collects no analytics. The “off by default” is “the off switch doesn’t exist because the feature doesn’t exist.”
  • Account / login — Rubric is single-user, local-first. There is no account.

If you’d like one of these added, send the request through the support channel. See License & Disclaimer for license terms.

Restoring after a reinstall

Settings are persisted to the platform’s standard application-support directory. Reinstalling the app reads them back automatically. If you migrate to a new machine and want to bring settings with you, copy the Hive database files from the application-support directory on the old machine to the same path on the new one.