Boundary Boundary alpha
Docs Why Boundary Blog Home
Get Started

Declarative, rule-based audience segmentation. Segments are defined in code with the defaudience macro or persisted dynamically in the database, then evaluated against your user base through a hybrid SQL
predicate pipeline.

Key namespaces

Namespace Purpose

boundary.audience.core.audience

Pure: segment definition registry and the defaudience macro

boundary.audience.core.filter

Pure: built-in filter types and the filter→sql / filter→predicate multimethods

boundary.audience.core.compiler

Pure: compile segments into SQL + predicate execution plans

boundary.audience.core.composition

Pure: AND/OR/NOT composition of segment result sets

boundary.audience.core.ui

Pure: Hiccup UI components for segments

boundary.audience.ports

Protocols: IAudienceResolver, IAudienceRepository, IAudienceCache, IUserDataSource

boundary.audience.shell.service

Service orchestrating resolution, caching, and persistence

boundary.audience.shell.persistence

IAudienceRepository — DB persistence (next.jdbc + HoneySQL)

boundary.audience.shell.cache

IAudienceCache — DB-backed cache of evaluated results

boundary.audience.shell.http

REST + web endpoints (CRUD, preview, evaluate, members)

Built-in filter types

Filter Matches on

:demographics

Age, gender, and other profile attributes

:location

Country, region, city

:role

Assigned user roles

:account-tenure

Time since account creation

:last-active

Recency of last activity

:behavior

Recorded user events / actions

:feature-usage

Use of specific product features

Each filter is resolved through the filter→sql and filter→predicate multimethods. SQL-eligible filters are pushed to the database; the rest run as Clojure predicates over the candidate set. Define new methods on both multimethods to add custom filter types.

Composition

Segments compose with AND / OR / NOT. Circular references between composed segments are detected and rejected at definition time.

Dynamic segments

Segments can also be stored in the database as JSON and loaded at runtime. Incoming JSON is schema-validated, and fn-typed values are rejected so persisted definitions stay data-only.

Caching

Membership results are cached in an audience_memberships table with a configurable per-segment TTL, served through the IAudienceCache component.

Builder UI

A segment builder is served over HTMX, with Replicant widget mount points for the filter panel and the composition builder.

Endpoints

Endpoint group Purpose

CRUD

Create, read, update, delete segments

Preview

Count + sample of matching members

Evaluate

Run evaluation and cache the result

Members

List cached members of a segment