🤖 AI Summary
Restate 1.5 is the new open-source release that accompanies Restate Cloud going public, bringing major observability, execution-management, performance, and docs improvements aimed at durable function and workflow orchestration. The UI now shows live execution timelines (retries, nested RPCs, events, cancellation), links invocations across services like a distributed call stack, and retains per-invocation journals by default in the single-binary’s RocksDB tables (configurable per deployment/service/handler). Crucially, failed TerminalErrors can be re-triggered via a “Restart as new” UI action, offering a built-in alternative to dead-letter queues — and paused invocations can be listed and resumed (or moved to another deployment) to support versioned rollouts and controlled recovery.
For developers building AI agents, long-running ML workflows, or microservice sagas, 1.5 adds fine-grained retry policies (service/handler level), a new pause state that preserves journals and virtual-object ownership (avoiding noisy retries and unnecessary billing), and the ability to resume invocations on different deployments. Under the hood the embedded SQL/query engine is 5–20× faster, making the UI snappier, and AWS Lambda payloads are now automatically compressed (plus Rust support) to avoid the 6 MB limit. The release also overhauls docs with tutorials for agents, workflows and microservice orchestration, lowering the barrier for building robust, observable, long-running AI/ML systems.
Loading comments...
login to comment
loading comments...
no comments yet