business-rules-reference

Rossum Business Rules & Validation Reference

Rossum validates extracted data and blocks automation in two ways. This pack covers the feature end-to-end for both:

1. Native Rossum Rules (POST /v1/rules) — the modern, platform-native path: a boolean trigger_condition plus an actions[] array. Each action names the event it runs on: validation for messages, blockers and show/hide, or a lifecycle event (import, confirm, export, and the four import/export success/failure outcome events) for send_email, change_queue, change_status, labels and exclude_from_training. The condition is TxScript (field.X access) — for that expression syntax see the txscript-reference skill. This pack owns the entity, the JSON shape, actions[] types, FIRE-vs-PASS polarity, and action pairings.
2. Legacy Business Rules Validation extension — a Store hook configured with a checks[] array, running at the end of the extension chain. It has its own curly-brace expression engine (has_value({field}), {value, default=0}, similarity()), documented in full here because nothing else uses it.

See reference.md for the full reference. Consult it when authoring or debugging either kind of rule, choosing between them, or reading a checks[] config.

It also covers porting expressions between engines — legacy checks[] to native Rules, legacy calculation hooks to formulas — as a defect taxonomy plus a verification method built on real field values and predicted message counts rather than on translating expression text.

Cross-references:

• txscript-reference — the TxScript expression language for native Rule trigger_conditions (and hooks/formulas). Native Rules use field.X; the BRV extension uses a different {field}-brace engine documented here.
• rossum-reference — the /v1/rules and /v1/triggers endpoint basics and the platform overview.
• upgrade — the skill that drives deprecated-extension migrations; the porting method here is the value-level verification step for the rules and formulas it produces.
• test-behavioral-equivalence — corpus-wide before/after replay across two environments, the heavier companion to the in-place count check.