Built with Micronaut, z4j focuses on low memory use, fast startup, and native GraalVM support.
Compared to traditional reflection-based clients like CloudBees:
- Zero reflection: Serialization is handled at compile time without runtime reflection caches.
- Low memory overhead: Lean execution across cloud containers, serverless workloads, and CLI tools.
Here are the memory allocation results when parsing Zendesk tickets into domain models:
| Ticket Cohort | Payload Size | CloudBees (Jackson AST + Reflection) | z4j (Compile-Time Serde Records) | z4j Memory Savings |
|---|---|---|---|---|
| 25 Tickets | 43 KB JSON | 50,344 KB (cold start) | 24,401 KB (cold start) | 52% less memory |
| 50 Tickets | 86 KB JSON | 1,661 KB | 520 KB | 69% less memory |
| 100 Tickets | 173 KB JSON | 3,799 KB | 1,035 KB | 73% less memory |
Run the benchmark locally:
./gradlew benchmarkGradle
| Gradle Kotlin | Gradle |
|---|---|
implementation("lol.pbu:z4j:0.4.3") |
implementation 'lol.pbu:z4j:0.4.3' |
Maven
<dependency>
<groupId>lol.pbu</groupId>
<artifactId>z4j</artifactId>
<version>0.4.3</version>
</dependency>- zcmi: Cross-platform CLI to manage Zendesk tickets and articles with automatic rate-limit backoff.
- zenith: MCP server bringing rate-safe Zendesk tools to AI agents on any platform.
Please read our Code of Conduct and Contributing Guide for details on our development process, style guides, and testing strategy.