chore: publish new package versions

[github-actions]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.

Releases

@electric-ax/agents@0.4.3

Patch Changes

• e126eba: Harden pull-wake runner lifecycle with a state machine, heartbeat-driven stream resets, and exponential reconnect backoff (1s-30s). Add granular status field to PullWakeRunnerHealth (stopped | starting | connecting | streaming | reconnecting | stopping). The onError callback is now reporting-only ((Error) => void) - it can no longer control runner lifecycle. stop() rethrows drainWakes errors so callers observe wake handler failures. Event-driven heartbeat throttling avoids stale diagnostics between fixed-interval heartbeats. Durable Streams clients now append stream and __ds subscription control paths to the configured backend URL prefix without inferring a /v1/stream layout, so pull-wake subscriptions work behind arbitrary DS backend prefixes. Remove the stale StreamClient.getConsumerState() helper for the old Durable Streams /consumers endpoint.
• e126eba: Add pull-wake runner health check endpoint and rename owner_user_id to owner_principal across the runners system. The GET /_electric/runners/:id/health endpoint returns comprehensive diagnostics including runner state, client-reported stream/heartbeat/claim metrics, active claims, and dispatch stats with a derived health status (healthy/degraded/unhealthy). The PullWakeRunner now tracks internal diagnostics and reports them to the server via heartbeats, stored in a separate runner_runtime_diagnostics table so the main runners shape stays stable for normal UI sync. The owner_user_id → owner_principal rename stores canonical principal URLs instead of keys, with strict validation and canonicalization at route boundaries. The migration expires active runner claims and deletes existing runner rows as part of the principal rewrite. This is a breaking change with no backward compatibility — all callers must send principal URLs.
• Updated dependencies [e126eba]
• Updated dependencies [e126eba]
  • @electric-ax/agents-runtime@0.2.2

@electric-ax/agents-runtime@0.2.2

Patch Changes

• e126eba: Harden pull-wake runner lifecycle with a state machine, heartbeat-driven stream resets, and exponential reconnect backoff (1s-30s). Add granular status field to PullWakeRunnerHealth (stopped | starting | connecting | streaming | reconnecting | stopping). The onError callback is now reporting-only ((Error) => void) - it can no longer control runner lifecycle. stop() rethrows drainWakes errors so callers observe wake handler failures. Event-driven heartbeat throttling avoids stale diagnostics between fixed-interval heartbeats. Durable Streams clients now append stream and __ds subscription control paths to the configured backend URL prefix without inferring a /v1/stream layout, so pull-wake subscriptions work behind arbitrary DS backend prefixes. Remove the stale StreamClient.getConsumerState() helper for the old Durable Streams /consumers endpoint.
• e126eba: Add pull-wake runner health check endpoint and rename owner_user_id to owner_principal across the runners system. The GET /_electric/runners/:id/health endpoint returns comprehensive diagnostics including runner state, client-reported stream/heartbeat/claim metrics, active claims, and dispatch stats with a derived health status (healthy/degraded/unhealthy). The PullWakeRunner now tracks internal diagnostics and reports them to the server via heartbeats, stored in a separate runner_runtime_diagnostics table so the main runners shape stays stable for normal UI sync. The owner_user_id → owner_principal rename stores canonical principal URLs instead of keys, with strict validation and canonicalization at route boundaries. The migration expires active runner claims and deletes existing runner rows as part of the principal rewrite. This is a breaking change with no backward compatibility — all callers must send principal URLs.

@electric-ax/agents-server@0.4.4

Patch Changes

• e126eba: Harden pull-wake runner lifecycle with a state machine, heartbeat-driven stream resets, and exponential reconnect backoff (1s-30s). Add granular status field to PullWakeRunnerHealth (stopped | starting | connecting | streaming | reconnecting | stopping). The onError callback is now reporting-only ((Error) => void) - it can no longer control runner lifecycle. stop() rethrows drainWakes errors so callers observe wake handler failures. Event-driven heartbeat throttling avoids stale diagnostics between fixed-interval heartbeats. Durable Streams clients now append stream and __ds subscription control paths to the configured backend URL prefix without inferring a /v1/stream layout, so pull-wake subscriptions work behind arbitrary DS backend prefixes. Remove the stale StreamClient.getConsumerState() helper for the old Durable Streams /consumers endpoint.
• e126eba: Add pull-wake runner health check endpoint and rename owner_user_id to owner_principal across the runners system. The GET /_electric/runners/:id/health endpoint returns comprehensive diagnostics including runner state, client-reported stream/heartbeat/claim metrics, active claims, and dispatch stats with a derived health status (healthy/degraded/unhealthy). The PullWakeRunner now tracks internal diagnostics and reports them to the server via heartbeats, stored in a separate runner_runtime_diagnostics table so the main runners shape stays stable for normal UI sync. The owner_user_id → owner_principal rename stores canonical principal URLs instead of keys, with strict validation and canonicalization at route boundaries. The migration expires active runner claims and deletes existing runner rows as part of the principal rewrite. This is a breaking change with no backward compatibility — all callers must send principal URLs.
• Updated dependencies [e126eba]
• Updated dependencies [e126eba]
  • @electric-ax/agents-runtime@0.2.2

electric-ax@0.2.3

Patch Changes

• e126eba: Add pull-wake runner health check endpoint and rename owner_user_id to owner_principal across the runners system. The GET /_electric/runners/:id/health endpoint returns comprehensive diagnostics including runner state, client-reported stream/heartbeat/claim metrics, active claims, and dispatch stats with a derived health status (healthy/degraded/unhealthy). The PullWakeRunner now tracks internal diagnostics and reports them to the server via heartbeats, stored in a separate runner_runtime_diagnostics table so the main runners shape stays stable for normal UI sync. The owner_user_id → owner_principal rename stores canonical principal URLs instead of keys, with strict validation and canonicalization at route boundaries. The migration expires active runner claims and deletes existing runner rows as part of the principal rewrite. This is a breaking change with no backward compatibility — all callers must send principal URLs.
• Updated dependencies [e126eba]
• Updated dependencies [e126eba]
  • @electric-ax/agents-runtime@0.2.2
  • @electric-ax/agents@0.4.3

@electric-ax/example-deep-survey@0.1.10

Patch Changes

• Updated dependencies [e126eba]
• Updated dependencies [e126eba]
  • @electric-ax/agents-runtime@0.2.2

@electric-ax/agents-desktop@0.1.4

Patch Changes

• e4acb1d: Use the Electric Cloud service query parameter for tenant-specific agents URLs so desktop cloud requests target the root agents endpoint while preserving tenant auth routing.
• e126eba: Route local desktop mutating agents-server requests through the Electron main process so CORS preflights cannot stall behind renderer connection limits.
• e126eba: Default unauthenticated local desktop sessions to the system:dev-local principal and resolve optimistic send principals at mutation time so pending messages do not render as unknown.
• e126eba: Add pull-wake runner health check endpoint and rename owner_user_id to owner_principal across the runners system. The GET /_electric/runners/:id/health endpoint returns comprehensive diagnostics including runner state, client-reported stream/heartbeat/claim metrics, active claims, and dispatch stats with a derived health status (healthy/degraded/unhealthy). The PullWakeRunner now tracks internal diagnostics and reports them to the server via heartbeats, stored in a separate runner_runtime_diagnostics table so the main runners shape stays stable for normal UI sync. The owner_user_id → owner_principal rename stores canonical principal URLs instead of keys, with strict validation and canonicalization at route boundaries. The migration expires active runner claims and deletes existing runner rows as part of the principal rewrite. This is a breaking change with no backward compatibility — all callers must send principal URLs.

@electric-ax/agents-server-ui@0.4.4

Patch Changes

• e126eba: Route local desktop mutating agents-server requests through the Electron main process so CORS preflights cannot stall behind renderer connection limits.
• e126eba: Default unauthenticated local desktop sessions to the system:dev-local principal and resolve optimistic send principals at mutation time so pending messages do not render as unknown.
• e126eba: Send new-session initial messages through the spawn request so pull-wake sessions can start without waiting for the UI to preload the entity stream.
• Updated dependencies [e126eba]
• Updated dependencies [e126eba]
  • @electric-ax/agents-runtime@0.2.2

@core/electric-telemetry@0.2.7

Patch Changes

• eac6dea: Wrap telemetry-poller MFAs in ElectricTelemetry.Poller.safe_invoke/3 so that transient collector failures (:noproc, :timeout, :shutdown/:normal exits, ArgumentError from not-yet-created ETS tables) no longer cause :telemetry_poller to permanently remove the measurement from its polling list. Unexpected errors are now logged as warnings with the offending MFA and the collector keeps being polled on subsequent ticks. Strips now-redundant defensive try/catch / with-fallthrough code from individual collectors.

  Note: user-supplied periodic measurement functions no longer have exceptions propagated up to :telemetry_poller's own error logger — they are caught and logged via ElectricTelemetry.Poller instead.

@core/sync-service@1.6.6

Patch Changes

• 8a94d7f: Fix @> (and other array operators) returning 400 when the right-hand side uses a non-foldable ARRAY[...]::T[] outer cast, e.g. "organization_ids" @> ARRAY[$1]::uuid[] with a column or parameter inside the constructor. The where-clause parser was assigning the element type (:uuid) instead of the array type ({:array, :uuid}) to array-cast and array-implicit-cast functions, which made the @> operator overload lookup fail with Could not select an operator overload.

• 4590862: Remove the single-function PartialMode module by moving its query_subset() function into SnapshotQuery, so that its implementation could sit close to SnapshotQuery.execute_for_shape().

• eac6dea: Wrap telemetry-poller MFAs in ElectricTelemetry.Poller.safe_invoke/3 so that transient collector failures (:noproc, :timeout, :shutdown/:normal exits, ArgumentError from not-yet-created ETS tables) no longer cause :telemetry_poller to permanently remove the measurement from its polling list. Unexpected errors are now logged as warnings with the offending MFA and the collector keeps being polled on subsequent ticks. Strips now-redundant defensive try/catch / with-fallthrough code from individual collectors.

  Note: user-supplied periodic measurement functions no longer have exceptions propagated up to :telemetry_poller's own error logger — they are caught and logged via ElectricTelemetry.Poller instead.