kafka4beam/wolff

Wolff is an open-source Erlang library that functions as a Kafka producer, providing a structured API for publishing messages to Kafka clusters. It supports both static and dynamic topic configurations, offering features such as asynchronous and synchronous sending, configurable batching, retry policies, and memory-pressure handling. The system utilizes disk-based replay queues for buffering and integrates with Telemetry for observability, while also including comprehensive test suites and scripts for managing local Kafka test environments.

The scripts directory now includes a new test environment setup, adding a \.env\ file for network configuration, \docker-compose.yml\ for Zookeeper-based Kafka clusters, and \docker-compose-kraft.yml\ for KRaft mode. A \setup-test-env.sh\ script automates the provisioning of these environments, supporting Kafka versions from 0.9 through 4.0 by selecting the appropriate compose file based on the version (Zookeeper for \<3.0, KRaft for \>=3.0) and creating initial test topics.

scripts · high confidence

This change introduces the Wolff Kafka publisher as an open-source project, adding the Apache 2.0 license, a NOTICE file acknowledging its origin as a fork of the EMQ Technologies upstream, and a comprehensive README detailing its features such as disk-based replay queues, per-partition connection management, and auto-partition refresh. It also establishes the build infrastructure with a Makefile, rebar3 configuration, and test environment scripts.

(repo-wide) · high confidence

This change introduces the Wolff library, a new Erlang-based Kafka publisher. It provides a structured API for managing supervised clients and producers, supporting both static and dynamic topic configurations. Key capabilities include asynchronous and synchronous message sending, configurable batching (via \max\_linger\_ms\/\bytes\), retry policies (\max\_retry\), and memory-pressure handling (\drop\_if\_highmem\). The library integrates with \telemetry\ for observability, exposing metrics for inflight requests, queue sizes, and drop reasons, and relies on \kafka\_protocol\ and \replayq\ for underlying transport and disk buffering.

src · high confidence

The Wolff library now exposes a central header file defining key constants and macros for the system. This includes error reason atoms (such as \message\_expired\ and \max\_retry\_exceeded\), the default Kafka message byte limit (1,000,000 bytes), and the names of the global ETS tables used for client and producer registration. It also defines macros for namespace-topic pairing and partition handling, establishing the foundational types and identifiers used by the client.

include · high confidence

Added new test files including \wolff\_bench.erl\ for performance benchmarking, \wolff\_client\_SUITE.erl\ for client lifecycle and connection handling, \wolff\_dynamic\_topics\_SUITE.erl\ for dynamic topic support, and \wolff\_supervised\_tests.erl\ for supervised producer/client management. These tests cover scenarios such as auto-topic creation, leader connection retries, reconnection on recent disconnect, dynamic topic expiration, and telemetry event logging.

test · high confidence

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

• CAI 72 → 71 (-0.7)
• Rubric changed (rubric-2026.09.12 → rubric-2026.09.18) — scores are not directly comparable.

• Code Health 96 → 96 (+0.0)
• Architecture 100 → 100 (+0.0)
• Maturity 59 → 59 (+0.0)
• Readiness 68 → 65 (-2.8)
• Security 100 → 100 (+0.0)

• Coverage not measured — no coverage collector is wired up

• Dependency hygiene PARTLY measured — rebar3 pinning read, dependency currency not (no rebar.lock-pinned Hex declaration to grade)
• Documentation: no contributor guidance (README.md)
• Documentation: no usage examples (README.md)
• No dependency lockfile committed (rebar.config)

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

kafka4beam/wolff was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.