Tag integration spans with instrumentation scope

.changesets/add-collector-attribute-options.md

@@ -0,0 +1,10 @@
+---
+bump: minor
+type: add
+---
+
+Add configuration options that map to OpenTelemetry resource attributes under collector mode: `service_name`, `filter_attributes`, `filter_function_parameters`, `filter_request_query_parameters`, `filter_request_payload`, `response_headers`, `send_function_parameters`, `send_request_query_parameters`, and `send_request_payload`. These tell the AppSignal Collector how to filter and forward telemetry data.
+
+When collector mode is active, existing configuration options (`name`, environment, `hostname`, `revision`, `ignore_actions`, `ignore_errors`, `ignore_namespaces`, `request_headers`, `filter_session_data`, `send_session_data`) are now passed to the collector as OpenTelemetry resource attributes.
+
+Setting any of these options without `collector_endpoint`, or `filter_parameters`/`filter_metadata`/`send_params` with `collector_endpoint`, now logs a warning at startup.

.changesets/add-collector-mode.md

@@ -0,0 +1,19 @@
+---
+bump: minor
+type: add
+---
+
+Add a new `collector_endpoint` configuration option (`APPSIGNAL_COLLECTOR_ENDPOINT` environment variable) that puts the integration in _collector mode_. When set, AppSignal additionally configures an OpenTelemetry SDK that exports OTLP/HTTP protobuf traces, metrics, and logs to the configured endpoint. The existing AppSignal agent continues to run unchanged; no AppSignal-collected data flows through the OpenTelemetry SDK yet.
+
+Collector mode requires Ruby 3.1 or newer and the OpenTelemetry gems, which are optional and not installed by default. To use it, add them to your application's `Gemfile`:
+
+```ruby
+gem "opentelemetry-sdk", ">= 1.8.0"
+gem "opentelemetry-metrics-sdk", ">= 0.7.1"
+gem "opentelemetry-logs-sdk", ">= 0.2.0"
+gem "opentelemetry-exporter-otlp", ">= 0.30.0"
+gem "opentelemetry-exporter-otlp-metrics", ">= 0.4.0"
+gem "opentelemetry-exporter-otlp-logs", ">= 0.2.0"
+```
+
+If these gems are missing or older than the minimum versions, AppSignal logs a warning and falls back to the bundled agent.

.changesets/add-que-distributed-tracing.md

@@ -0,0 +1,6 @@
+---
+bump: minor
+type: add
+---
+
+Improve Que support. In collector mode, AppSignal now propagates trace context when enqueuing Que jobs, so each job links back to the trace that enqueued it. This covers single and bulk enqueues, on both Que 1 and Que 2.

.changesets/complete-transaction-when-a-block-raises.md

@@ -0,0 +1,10 @@
+---
+bump: patch
+type: fix
+---
+
+Complete the transaction even when a block handed to it raises.
+
+The error block given to `Appsignal.send_error` and `Appsignal.report_error`, and the internal `after_create` and `before_complete` hooks, are user code that can raise. An error block runs at completion in agent mode, and the hooks run during creation and completion, so these blocks run far from where they were defined.
+
+Until now, an exception from any of these blocks propagated out of transaction creation or completion. That surfaced the error in unrelated code, and it could leave the transaction unfinished. These blocks are now run defensively. The failure is logged, naming the error and where the block was defined, and creation and completion carry on.

.changesets/drop-actionless-collector-transactions.md

@@ -0,0 +1,6 @@
+---
+bump: patch
+type: fix
+---
+
+In collector mode, drop a transaction that never set an action name instead of reporting it under the placeholder `appsignal.transaction <namespace>` action. A request that is not routed to an action -- for example a static asset served without a controller -- has no action to group by. Agent mode does not report such a transaction, but collector mode was surfacing every one of them under a single shared placeholder action. The transaction's root span is now flagged with `appsignal.ignore_subtrace` on completion, so the AppSignal Collector (0.10.0 and newer) drops it, matching agent mode.

.changesets/sequel-query-spans-client-kind.md

@@ -0,0 +1,11 @@
+---
+bump: patch
+type: fix
+---
+
+Mark Sequel query spans as CLIENT kind in collector mode. The sequel-rails gem
+emits its queries as `sql.sequel` ActiveSupport::Notifications events, which are
+recorded through the generic notifications integration rather than the dedicated
+Sequel hook. That path only tagged `sql.active_record` as an outgoing datastore
+call, so Sequel queries were exported with the default span kind. They now carry
+CLIENT kind, matching ActiveRecord and the dedicated Sequel hook.

.changesets/tag-integration-instrumentation-scope.md

@@ -0,0 +1,6 @@
+---
+bump: patch
+type: change
+---
+
+In collector mode, each AppSignal integration now records its spans under its own OpenTelemetry instrumentation scope. The scope is named after the integration, for example `appsignal-ruby/sidekiq` or `appsignal-ruby/net_http`. The Rails components that are bridged through `ActiveSupport::Notifications` are named after the component, for example `appsignal-ruby/active_record` and `appsignal-ruby/action_view`. This makes it possible to group spans by the integration that produced them. Custom instrumentation keeps the default `appsignal-ruby` scope, and so does anything else without a more specific scope.