refactor: keep correlated tracing out of OVOS core

Author: goldyfruit

docs/performance-metrics.md

@@ -85,17 +85,3 @@ histogram_quantile(
 These histograms are cumulative and reset when the process restarts. They use
 no session, client, utterance, or skill labels, avoiding unbounded cardinality
 and user-content leakage.
-
-## Request-correlated benchmark traces
-
-Set `OVOS_PERFORMANCE_TRACE=true` only for a controlled benchmark to emit an
-opaque request ID and a wall-clock nanosecond timestamp when the intent service
-receives an utterance. The structured log record is prefixed with
-`performance_trace` and uses the stage `runtime_receive`.
-
-Request IDs are intentionally absent from the Prometheus endpoint. The trace
-contains no utterance, skill, client, credential, or session payload and does
-not change the message or public bus contract. Join this opt-in event with the
-matching Workshop `skill_reply_emit`, listener stages, and client-receipt
-timestamp outside the runtime. Cluster nodes must have synchronized clocks
-before interpreting cross-process intervals.

ovos_core/_performance_trace.py

@@ -54,7 +54,6 @@
     UTTERANCE_TRANSFORM,
     pipeline_matching_histogram,
 )
-from ovos_core._performance_trace import trace_performance_stage
 from ovos_core.intent_services.dispatcher import (
     DEFAULT_HANDLER_TIMEOUT,
     IntentDispatcher,
@@ -639,7 +638,6 @@ def handle_utterance(self, message: Message):
         Args:
             message (Message): The messagebus data
         """
-        trace_performance_stage("runtime_receive", message=message)
         with UTTERANCE_PREPROCESS.measure():
             # Get utterance utterance_plugins additional context
             with UTTERANCE_TRANSFORM.measure():

ovos_core/intent_services/service.py

@@ -10,11 +10,6 @@
     performance_histograms,
     pipeline_matching_histogram,
 )
-from ovos_core._performance_trace import (
-    message_request_id,
-    trace_performance_stage,
-)
-from ovos_bus_client.message import Message
 from ovos_core._prometheus import (
     collect_histograms,
     load_metric_collectors,
@@ -55,43 +50,6 @@ def test_histogram_rejects_non_finite_observations():
         histogram.observe_ms(float("nan"))
 
 
-def test_trace_extracts_nested_metadata_request_id():
-    message = Message(
-        "recognizer_loop:utterance",
-        {},
-        {"metadata": {"qa_query_id": "request-runtime"}},
-    )
-
-    assert message_request_id(message) == "request-runtime"
-
-
-def test_correlated_trace_is_opt_in(monkeypatch):
-    message = Message(
-        "recognizer_loop:utterance",
-        {},
-        {"query_id": "request-runtime"},
-    )
-    logged = []
-    monkeypatch.setattr(
-        "ovos_core._performance_trace._LOG.info",
-        lambda template, payload: logged.append(template % payload),
-    )
-    monkeypatch.delenv("OVOS_PERFORMANCE_TRACE", raising=False)
-
-    trace_performance_stage("runtime_receive", message=message)
-
-    assert logged == []
-    monkeypatch.setenv("OVOS_PERFORMANCE_TRACE", "true")
-    monkeypatch.setattr(
-        "ovos_core._performance_trace.time.time_ns",
-        lambda: 456_000_000,
-    )
-    trace_performance_stage("runtime_receive", message=message)
-    assert '"stage":"runtime_receive"' in logged[0]
-    assert '"request_id":"request-runtime"' in logged[0]
-    assert '"at_unix_ns":456000000' in logged[0]
-
-
 def test_prometheus_renderer_converts_milliseconds_to_seconds():
     payload = render_prometheus({
         "test_stage_ms": {