rototo
How-to: Authoring

How to Select a Value for a Runtime Condition

Use this when a named runtime condition should select a different value in an environment, such as enterprise accounts, one country, one tenant class, or a stable bucketed rollout.

The operational goal is to keep the condition definition in the workspace instead of branching in application code.

Expected outcome

After this change:

Define the condition as a qualifier

Create a qualifier file with a product-readable name:

qualifiers/enterprise-accounts.toml

schema_version = 1

[qualifier]
description = "Accounts on the enterprise plan with at least 100 seats"

[[qualifier.predicate]]
attribute = "account.plan"
op = "eq"
value = "enterprise"

[[qualifier.predicate]]
attribute = "account.seats"
op = "gte"
value = 100

The qualifier id is enterprise-accounts. It turns raw context fields into a named condition that can be reviewed and reused.

Add the rollout value

Add the target value to the variable. For a primitive value:

[variable.values]
standard = 1000
enterprise = 2000

For larger structured values, put the value in an external value file such as:

variables/llm-agent-config-values/enterprise.toml

The value key is the file stem, so this file defines enterprise.

Add a rule to the environment

Add a rule under the environment where the condition should select the new value:

[variable.env.prod]
value = "standard"

[[variable.env.prod.rule]]
description = "Enterprise accounts get the larger value"
qualifier = "enterprise-accounts"
value = "enterprise"

Rules are evaluated before the environment default. If the qualifier matches, rototo selects enterprise. If it does not match, rototo selects standard.

Verify matching and non-matching context

Lint the workspace:

rototo workspace lint config/

Resolve with matching context:

rototo variable resolve max-output-tokens \
  --workspace config/ \
  --env prod \
  --context '{"account":{"plan":"enterprise","seats":250}}'

Resolve with non-matching context:

rototo variable resolve max-output-tokens \
  --workspace config/ \
  --env prod \
  --context '{"account":{"plan":"team","seats":25}}'

The two commands should return different value_key values.

Common mistakes

Do not duplicate the same predicate in many variables. Define a qualifier once and reuse it from rules.

Do not skip the non-matching test. Rollout mistakes often come from defaults, not only from the matching branch.

Do not rely on missing context fields. Add or update a context schema so the application/config boundary is explicit.

Related docs