@workflowbuilder/temporal@0.3.0

A run's status turns running the moment it starts, and TemporalWorkflowEngine recovers from a failed connection. Event Histories recorded by 0.1.0 and 0.2.0 still replay.

Added

• lazyClient(connectionOptions, clientOptions?) in @workflowbuilder/temporal/client: a client factory for TemporalWorkflowEngine that closes the connection of a failed attempt, so retries during an outage leave no channel open.
• reasonRequired on a resume action, as on a reject: the type now carries it on every routed action.
• TESTING.md in the package: what the test suites cover, what they do not, and how CI gates a release.

Changed

• A run's status moves to running as soon as it starts, so updateExecutionStatus receives running before the first node runs and a store can stamp the start. Runs started on an earlier version keep their old sequence and still replay.
• The README documents what Event History records, where secrets belong, and the payload and history size limits a run can reach.
• The README and TESTING.md state that the package is also tested weekly against the newest Temporal TypeScript SDK release.

Fixed

• TemporalWorkflowEngine connects again on the next call after a failed connection attempt, instead of failing every call until the process restarts.