Relay

Relay is an event-driven runtime designed for building declarative applications that utilize serverless functions, schedules, and persistent services. It operates on infrastructure under user control, ensuring greater flexibility and reliability in execution.

This platform excels at executing short-lived functions within isolated runtimes, facilitating the publishing of scheduled events alongside external occurrences to a unified Redis Stream and consumer group. Relay continuously reconciles long-running services based on a shared declarative model without necessitating their involvement in the event pipeline.

Key Features

• Event-Driven Execution: Leverages Redis Streams for declarative matching, ensuring retries, dead-lettering, and at-least-once delivery for each function invocation.
• Managed Runtimes: Supports Python 3.14 and Node 24 with integrated TypeScript support, optimizing performance through reusable warm containers and bounded concurrency.
• Cron Schedules: Implement minute-precision cron schedules with IANA timezones, offering features like deterministic occurrence identity, cluster-wide deduplication, and bounded retries.
• Persistent Services: Enables the deployment of APIs, background workers, and other long-lived processes while managing runtime environments, resource limits, and optional routing via Traefik.
• Resource Controls: Define per-container limits on memory, CPU, and process identifiers to manage system resources effectively.
• Configuration and Secrets Management: Share environment variables and secret references seamlessly across applications, schedules, and services.
• Observability Features: Gain insights through persisted statistics, Prometheus metrics, structured logs, and OpenTelemetry tracing.
• Operational Tools: Execute health checks, manual invocations, and inspect/replay dead-letter queues (DLQs), supplemented by Git synchronization and ordered teardown processes.

Workload Dynamics

Relay's architecture encompasses two primary workloads:

• Functions: Designed for ephemeral, one-shot executions triggered by external events or schedules, executing through a common runtime.
• Services: Persistent workloads that are continuously reconciled to their declared state, suitable for long-running processes.

Workflow Example

Events flow seamlessly through Relay as follows:

Producer → Redis Stream → Relay → App → Function (handler)

Event classifications and retries are managed within a consistent infrastructure to ensure reliability and performance. Relay smartly distinguishes between events and scheduled occurrences, streamlining the execution process for efficient handling of workloads.

Quick Start

To get started with Relay, you need:

• Redis installed
• Access to a Docker Engine

Simply use the bundled compose.yaml to launch Relay alongside Redis and deploy your apps. Functions and services can then be declared and executed directly, with a shared model facilitating seamless integration.

Relay emphasizes a distinct separation between function invocations and service convergence, allowing developers to start with simple functions and scale to more complex services as needed. This adaptability makes it particularly useful for modern application development, where event-driven architectures and persistent workloads must coexist harmoniously.