OmniNode's founder, a veteran iOS developer who turned to AI agents after developing neuropathy, shares how he built a contract-driven AI agent orchestration platform. The post details the migration from Redis Streams to Redpanda as the event bus grew beyond 100 topic types across 12 repositories. The core engineering challenge was topic naming drift — silent failures caused by hyphen/underscore mismatches, pluralization differences, and version suffix inconsistencies. The solution: a single ContractTopicExtractor that reads topic names exclusively from contract.yaml files, enforced at three independent call sites (pre-deploy CI, runtime boot, and post-boot validation). Redpanda was chosen for its Kafka API compatibility and low operational footprint, enabling the same broker to run in local dev, CI, and production without fake broker paths.

11m read timeFrom redpanda.com
Post cover image
Table of contents
How I used AI to help solve coding and workflow challengesThe origins of OmniNode: Building a system to coordinate across agentsWhat this buys you: cheapest-capable routing with a receiptRedis Streams worked until scale changed the problemMigrating to Redpanda and addressing our topic naming complicationThe solution: contract owns the topic namesOne extractor for contract topics, three call sitesWhat runtime provisioning actually doesThe benefits of Redpanda’s broker affordability for topic identityUp next: Configuration driftRecommendations
2.2K Impressions