refactor: reduce Prometheus metrics cardinality and clarify label names (#1059)

With many workspaces and flows, the prometheus metrics expand
greatly. Many significant metrics (e.g. latency) make more sense
measured globally. This change introduces a consolidation.

Histograms (request_latency, text_completion_duration, chunk_size, etc.)
are now global with only processor-level labels, eliminating per-workspace
time series multiplication. Counters and enums retain workspace and flow
labels for per-workspace debugging visibility.

Renamed metric labels for clarity: name -> consumer/producer/subscriber,
and added explicit workspace label alongside flow. Fixed bug where
request_response_spec labelled response subscriber with request_name
instead of response_name. Removed unused Histogram imports from model
and service files.

tests/unit/test_base/test_metrics.py

@@ -73,8 +73,8 @@ def test_consumer_metrics_reuses_singletons_and_records_events(monkeypatch):
     monkeypatch.setattr(metrics, "Histogram", histogram_factory)
     monkeypatch.setattr(metrics, "Counter", counter_factory)
 
-    first = metrics.ConsumerMetrics("proc", "flow", "name")
+    first = metrics.ConsumerMetrics("proc", "cons", workspace="ws", flow="fl")
-    second = metrics.ConsumerMetrics("proc-2", "flow-2", "name-2")
+    second = metrics.ConsumerMetrics("proc-2", "cons-2")
 
     assert enum_factory.call_count == 1
     assert histogram_factory.call_count == 1
@@ -96,7 +96,7 @@ def test_producer_metrics_increments_counter_once(monkeypatch):
     counter_factory.return_value.labels.return_value = labels
     monkeypatch.setattr(metrics, "Counter", counter_factory)
 
-    producer_metrics = metrics.ProducerMetrics("proc", "flow", "output")
+    producer_metrics = metrics.ProducerMetrics("proc", "output")
     producer_metrics.inc()
 
     counter_factory.assert_called_once()
@@ -133,7 +133,7 @@ def test_subscriber_metrics_tracks_received_state_and_dropped(monkeypatch):
     monkeypatch.setattr(metrics, "Enum", enum_factory)
     monkeypatch.setattr(metrics, "Counter", counter_factory)
 
-    subscriber_metrics = metrics.SubscriberMetrics("proc", "flow", "input")
+    subscriber_metrics = metrics.SubscriberMetrics("proc", "input")
     subscriber_metrics.received()
     subscriber_metrics.state("running")
     subscriber_metrics.dropped("ignored")

trustgraph-base/trustgraph/base/agent_service.py

@@ -8,7 +8,6 @@ from argparse import ArgumentParser
 
 import time
 import logging
-from prometheus_client import Histogram
 
 from .. schema import AgentRequest, AgentResponse, Error
 from .. exceptions import TooManyRequests

trustgraph-base/trustgraph/base/async_processor.py

@@ -84,7 +84,7 @@ class AsyncProcessor:
         config_subscriber_id = str(uuid.uuid4())
 
         config_consumer_metrics = ConsumerMetrics(
-            processor = self.id, flow = None, name = "config",
+            processor=self.id, consumer="config",
         )
 
         # Subscribe to config notify queue
@@ -112,10 +112,10 @@ class AsyncProcessor:
         config_rr_id = str(uuid.uuid4())
 
         config_req_metrics = ProducerMetrics(
-            processor = self.id, flow = None, name = "config-request",
+            processor=self.id, producer="config-request",
         )
         config_resp_metrics = SubscriberMetrics(
-            processor = self.id, flow = None, name = "config-response",
+            processor=self.id, subscriber="config-response",
         )
 
         return RequestResponse(

trustgraph-base/trustgraph/base/audit_publisher.py

@@ -21,8 +21,7 @@ class AuditPublisher:
             schema=AuditEvent,
             metrics=ProducerMetrics(
                 processor=processor_id or component_name,
-                flow=None,
+                producer="audit-events",
-                name="audit-events",
             ),
         )
 

trustgraph-base/trustgraph/base/consumer_spec.py

@@ -16,7 +16,8 @@ class ConsumerSpec(Spec):
     def add(self, flow: Any, processor: Any, definition: dict[str, Any]) -> None:
 
         consumer_metrics = ConsumerMetrics(
-            processor = flow.id, flow = flow.name, name = self.name,
+            processor=flow.id, consumer=self.name,
+            workspace=flow.workspace, flow=flow.name,
         )
 
         consumer = Consumer(

trustgraph-base/trustgraph/base/dynamic_tool_service.py

@@ -63,7 +63,7 @@ class DynamicToolService(AsyncProcessor):
 
         # Create consumer for requests
         consumer_metrics = ConsumerMetrics(
-            processor=self.id, flow=None, name="request"
+            processor=self.id, consumer="request",
         )
 
         self.consumer = Consumer(
@@ -79,7 +79,7 @@ class DynamicToolService(AsyncProcessor):
 
         # Create producer for responses
         producer_metrics = ProducerMetrics(
-            processor=self.id, flow=None, name="response"
+            processor=self.id, producer="response",
         )
 
         self.producer = Producer(
@@ -93,7 +93,7 @@ class DynamicToolService(AsyncProcessor):
             __class__.tool_service_metric = Counter(
                 'dynamic_tool_service_invocation_count',
                 'Dynamic tool service invocation count',
-                ["id"],
+                ["processor"],
             )
 
     async def start(self):
@@ -133,7 +133,7 @@ class DynamicToolService(AsyncProcessor):
             )
 
             __class__.tool_service_metric.labels(
-                id=self.id,
+                processor=self.id,
             ).inc()
 
         except TooManyRequests as e:

trustgraph-base/trustgraph/base/embeddings_service.py

@@ -8,7 +8,6 @@ from argparse import ArgumentParser
 
 import time
 import logging
-from prometheus_client import Histogram
 
 from .. schema import EmbeddingsRequest, EmbeddingsResponse, Error
 from .. exceptions import TooManyRequests

trustgraph-base/trustgraph/base/image_to_text_service.py

@@ -75,7 +75,7 @@ class ImageToTextService(FlowProcessor):
             __class__.image_to_text_metric = Histogram(
                 'image_to_text_duration',
                 'Image-to-text duration (seconds)',
-                ["id", "flow"],
+                ["processor"],
                 buckets=[
                     0.25, 0.5, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0,
                     8.0, 9.0, 10.0, 11.0, 12.0, 13.0, 14.0, 15.0, 16.0,
@@ -89,7 +89,7 @@ class ImageToTextService(FlowProcessor):
             __class__.image_to_text_model_metric = Info(
                 'image_to_text_model',
                 'Image-to-text model',
-                ["processor", "flow"]
+                ["processor"]
             )
 
     async def on_request(self, msg, consumer, flow):
@@ -105,8 +105,7 @@ class ImageToTextService(FlowProcessor):
             model = flow("model")
 
             with __class__.image_to_text_metric.labels(
-                    id=self.id,
+                    processor=self.id,
-                    flow=f"{flow.name}-{consumer.name}",
             ).time():
 
                 response = await self.describe_image(
@@ -126,8 +125,7 @@ class ImageToTextService(FlowProcessor):
             )
 
             __class__.image_to_text_model_metric.labels(
-                processor = self.id,
+                processor=self.id,
-                flow = flow.name
             ).info({
                 "model": str(model) if model is not None else "",
             })

trustgraph-base/trustgraph/base/librarian_client.py

@@ -43,10 +43,12 @@ class LibrarianClient:
             "librarian_subscriber", f"{id}-librarian",
         )
 
+        workspace = params.get("workspace")
         flow_name = params.get("flow_name")
 
         librarian_request_metrics = ProducerMetrics(
-            processor=id, flow=flow_name, name="librarian-request",
+            processor=id, producer="librarian-request",
+            workspace=workspace, flow=flow_name,
         )
 
         self._producer = Producer(
@@ -57,7 +59,8 @@ class LibrarianClient:
         )
 
         librarian_response_metrics = ConsumerMetrics(
-            processor=id, flow=flow_name, name="librarian-response",
+            processor=id, consumer="librarian-response",
+            workspace=workspace, flow=flow_name,
         )
 
         self._consumer = Consumer(

trustgraph-base/trustgraph/base/librarian_spec.py

@@ -25,6 +25,7 @@ class LibrarianSpec(Spec):
                 processor.id + "--" + flow.workspace + "--" +
                 flow.name + "--librarian--" + str(uuid.uuid4())
             ),
+            workspace=flow.workspace,
             flow_name=flow.name,
         )
 

trustgraph-base/trustgraph/base/llm_service.py

@@ -94,7 +94,7 @@ class LlmService(FlowProcessor):
             __class__.text_completion_metric = Histogram(
                 'text_completion_duration',
                 'Text completion duration (seconds)',
-                ["id", "flow"],
+                ["processor"],
                 buckets=[
                     0.25, 0.5, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0,
                     8.0, 9.0, 10.0, 11.0, 12.0, 13.0, 14.0, 15.0, 16.0,
@@ -108,7 +108,7 @@ class LlmService(FlowProcessor):
             __class__.text_completion_model_metric = Info(
                 'text_completion_model',
                 'Text completion model',
-                ["processor", "flow"]
+                ["processor"]
             )
 
     async def on_request(self, msg, consumer, flow):
@@ -133,8 +133,7 @@ class LlmService(FlowProcessor):
 
                 # Streaming mode
                 with __class__.text_completion_metric.labels(
-                        id=self.id,
+                        processor=self.id,
-                        flow=f"{flow.name}-{consumer.name}",
                 ).time():
 
                     async for chunk in self.generate_content_stream(
@@ -157,8 +156,7 @@ class LlmService(FlowProcessor):
 
                 # Non-streaming mode (original behavior)
                 with __class__.text_completion_metric.labels(
-                        id=self.id,
+                        processor=self.id,
-                        flow=f"{flow.name}-{consumer.name}",
                 ).time():
 
                     response = await self.generate_content(
@@ -179,8 +177,7 @@ class LlmService(FlowProcessor):
                 )
 
             __class__.text_completion_model_metric.labels(
-                processor = self.id,
+                processor=self.id,
-                flow = flow.name
             ).info({
                 "model": str(model) if model is not None else "",
                 "temperature": str(temperature) if temperature is not None else "",

trustgraph-base/trustgraph/base/metrics.py

@@ -6,82 +6,85 @@ from prometheus_client import start_http_server, Info, Enum, Histogram
 from prometheus_client import Counter
 
 class ConsumerMetrics:
-    """
-    Metrics tracking and reporting for flow consumers.
 
-    This class manages prometheus metrics specifically related to consumers 
+    def __init__(self, processor: str, consumer: str,
-    within the flow, including state, requests, processing time, and queues.
+                 workspace: str | None = None,
-    """
+                 flow: str | None = None) -> None:
-
-    def __init__(self, processor: str, flow: str, name: str) -> None:
 
         self.processor = processor
+        self.consumer = consumer
+        self.workspace = workspace
         self.flow = flow
-        self.name = name
 
         if not hasattr(__class__, "state_metric"):
             __class__.state_metric = Enum(
                 'consumer_state', 'Consumer state',
-                ["processor", "flow", "name"],
+                ["processor", "workspace", "flow", "consumer"],
                 states=['stopped', 'running']
             )
 
         if not hasattr(__class__, "request_metric"):
             __class__.request_metric = Histogram(
                 'request_latency', 'Request latency (seconds)',
-                ["processor", "flow", "name"],
+                ["processor", "consumer"],
             )
 
         if not hasattr(__class__, "processing_metric"):
             __class__.processing_metric = Counter(
                 'processing_count', 'Processing count',
-                ["processor", "flow", "name", "status"],
+                ["processor", "workspace", "flow", "consumer", "status"],
             )
 
         if not hasattr(__class__, "rate_limit_metric"):
             __class__.rate_limit_metric = Counter(
                 'rate_limit_count', 'Rate limit event count',
-                ["processor", "flow", "name"],
+                ["processor", "workspace", "flow", "consumer"],
             )
 
     def process(self, status: str) -> None:
         __class__.processing_metric.labels(
-            processor = self.processor, flow = self.flow, name = self.name,
+            processor=self.processor, workspace=self.workspace,
-            status=status
+            flow=self.flow, consumer=self.consumer, status=status,
         ).inc()
 
     def rate_limit(self) -> None:
         __class__.rate_limit_metric.labels(
-            processor = self.processor, flow = self.flow, name = self.name,
+            processor=self.processor, workspace=self.workspace,
+            flow=self.flow, consumer=self.consumer,
         ).inc()
 
     def state(self, state: str) -> None:
         __class__.state_metric.labels(
-            processor = self.processor, flow = self.flow, name = self.name,
+            processor=self.processor, workspace=self.workspace,
+            flow=self.flow, consumer=self.consumer,
         ).state(state)
 
     def record_time(self) -> Any:
         return __class__.request_metric.labels(
-            processor = self.processor, flow = self.flow, name = self.name,
+            processor=self.processor, consumer=self.consumer,
         ).time()
 
 class ProducerMetrics:
 
-    def __init__(self, processor: str, flow: str, name: str) -> None:
+    def __init__(self, processor: str, producer: str,
+                 workspace: str | None = None,
+                 flow: str | None = None) -> None:
 
         self.processor = processor
+        self.producer = producer
+        self.workspace = workspace
         self.flow = flow
-        self.name = name
 
         if not hasattr(__class__, "producer_metric"):
             __class__.producer_metric = Counter(
                 'producer_count', 'Output items produced',
-                ["processor", "flow", "name"],
+                ["processor", "workspace", "flow", "producer"],
             )
 
     def inc(self) -> None:
         __class__.producer_metric.labels(
-            processor = self.processor, flow = self.flow, name = self.name
+            processor=self.processor, workspace=self.workspace,
+            flow=self.flow, producer=self.producer,
         ).inc()
 
 class ProcessorMetrics:
@@ -102,43 +105,48 @@ class ProcessorMetrics:
 
 class SubscriberMetrics:
 
-    def __init__(self, processor: str, flow: str, name: str) -> None:
+    def __init__(self, processor: str, subscriber: str,
+                 workspace: str | None = None,
+                 flow: str | None = None) -> None:
 
         self.processor = processor
+        self.subscriber = subscriber
+        self.workspace = workspace
         self.flow = flow
-        self.name = name
 
         if not hasattr(__class__, "state_metric"):
             __class__.state_metric = Enum(
                 'subscriber_state', 'Subscriber state',
-                ["processor", "flow", "name"],
+                ["processor", "workspace", "flow", "subscriber"],
                 states=['stopped', 'running']
             )
 
         if not hasattr(__class__, "received_metric"):
             __class__.received_metric = Counter(
                 'received_count', 'Received count',
-                ["processor", "flow", "name"],
+                ["processor", "workspace", "flow", "subscriber"],
             )
 
         if not hasattr(__class__, "dropped_metric"):
             __class__.dropped_metric = Counter(
                 'dropped_count', 'Dropped messages count',
-                ["processor", "flow", "name"],
+                ["processor", "workspace", "flow", "subscriber"],
             )
 
     def received(self) -> None:
         __class__.received_metric.labels(
-            processor = self.processor, flow = self.flow, name = self.name,
+            processor=self.processor, workspace=self.workspace,
+            flow=self.flow, subscriber=self.subscriber,
         ).inc()
 
     def state(self, state: str) -> None:
-
         __class__.state_metric.labels(
-            processor = self.processor, flow = self.flow, name = self.name,
+            processor=self.processor, workspace=self.workspace,
+            flow=self.flow, subscriber=self.subscriber,
         ).state(state)
 
     def dropped(self, state: str) -> None:
         __class__.dropped_metric.labels(
-            processor = self.processor, flow = self.flow, name = self.name,
+            processor=self.processor, workspace=self.workspace,
+            flow=self.flow, subscriber=self.subscriber,
         ).inc()

trustgraph-base/trustgraph/base/producer_spec.py

@@ -14,7 +14,8 @@ class ProducerSpec(Spec):
     def add(self, flow: Any, processor: Any, definition: dict[str, Any]) -> None:
 
         producer_metrics = ProducerMetrics(
-            processor = flow.id, flow = flow.name, name = self.name
+            processor=flow.id, producer=self.name,
+            workspace=flow.workspace, flow=flow.name,
         )
 
         producer = Producer(

trustgraph-base/trustgraph/base/request_response_spec.py

@@ -132,11 +132,13 @@ class RequestResponseSpec(Spec):
             return
 
         request_metrics = ProducerMetrics(
-            processor = flow.id, flow = flow.name, name = self.request_name
+            processor=flow.id, producer=self.request_name,
+            workspace=flow.workspace, flow=flow.name,
         )
 
         response_metrics = SubscriberMetrics(
-            processor = flow.id, flow = flow.name, name = self.request_name
+            processor=flow.id, subscriber=self.response_name,
+            workspace=flow.workspace, flow=flow.name,
         )
 
         rr = self.impl(

trustgraph-base/trustgraph/base/subscriber_spec.py

@@ -15,7 +15,8 @@ class SubscriberSpec(Spec):
     def add(self, flow: Any, processor: Any, definition: dict[str, Any]) -> None:
 
         subscriber_metrics = SubscriberMetrics(
-            processor = flow.id, flow = flow.name, name = self.name
+            processor=flow.id, subscriber=self.name,
+            workspace=flow.workspace, flow=flow.name,
         )
 
         subscriber = Subscriber(

trustgraph-base/trustgraph/base/tool_service.py

@@ -50,7 +50,7 @@ class ToolService(FlowProcessor):
         if not hasattr(__class__, "tool_invocation_metric"):
             __class__.tool_invocation_metric = Counter(
                 'tool_invocation_count', 'Tool invocation count',
-                ["id", "flow", "name"],
+                ["processor", "workspace", "flow", "tool"],
             )
 
     async def on_request(self, msg, consumer, flow):
@@ -89,7 +89,8 @@ class ToolService(FlowProcessor):
                 )
 
             __class__.tool_invocation_metric.labels(
-                id = self.id, flow = flow.name, name = request.name,
+                processor=self.id, workspace=flow.workspace,
+                flow=flow.name, tool=request.name,
             ).inc()
 
         except TooManyRequests as e:

trustgraph-flow/trustgraph/agent/react/tools.py

@@ -277,13 +277,11 @@ class ToolServiceImpl:
 
         request_metrics = ProducerMetrics(
             processor=self.processor.id,
-            flow="tool-service",
+            producer=self.request_queue,
-            name=self.request_queue
         )
         response_metrics = SubscriberMetrics(
             processor=self.processor.id,
-            flow="tool-service",
+            subscriber=self.response_queue,
-            name=self.response_queue
         )
 
         # Create unique subscription for responses

trustgraph-flow/trustgraph/bootstrap/bootstrapper/service.py

@@ -173,12 +173,12 @@ class Processor(AsyncProcessor):
             request_topic=config_request_queue,
             request_schema=ConfigRequest,
             request_metrics=ProducerMetrics(
-                processor=self.id, flow=None, name="config-request",
+                processor=self.id, producer="config-request",
             ),
             response_topic=config_response_queue,
             response_schema=ConfigResponse,
             response_metrics=SubscriberMetrics(
-                processor=self.id, flow=None, name="config-response",
+                processor=self.id, subscriber="config-response",
             ),
         )
 
@@ -191,12 +191,12 @@ class Processor(AsyncProcessor):
             request_topic=f"{flow_request_queue}:{workspace}",
             request_schema=FlowRequest,
             request_metrics=ProducerMetrics(
-                processor=self.id, flow=None, name="flow-request",
+                processor=self.id, producer="flow-request",
             ),
             response_topic=f"{flow_response_queue}:{workspace}",
             response_schema=FlowResponse,
             response_metrics=SubscriberMetrics(
-                processor=self.id, flow=None, name="flow-response",
+                processor=self.id, subscriber="flow-response",
             ),
         )
 
@@ -209,12 +209,12 @@ class Processor(AsyncProcessor):
             request_topic=iam_request_queue,
             request_schema=IamRequest,
             request_metrics=ProducerMetrics(
-                processor=self.id, flow=None, name="iam-request",
+                processor=self.id, producer="iam-request",
             ),
             response_topic=iam_response_queue,
             response_schema=IamResponse,
             response_metrics=SubscriberMetrics(
-                processor=self.id, flow=None, name="iam-response",
+                processor=self.id, subscriber="iam-response",
             ),
         )
 

trustgraph-flow/trustgraph/chunking/recursive/chunker.py

@@ -53,7 +53,7 @@ class Processor(ChunkingService):
         if not hasattr(__class__, "chunk_metric"):
             __class__.chunk_metric = Histogram(
                 'chunk_size', 'Chunk size',
-                ["id", "flow"],
+                ["processor"],
                 buckets=[100, 160, 250, 400, 650, 1000, 1600,
                          2500, 4000, 6400, 10000, 16000]
             )
@@ -184,7 +184,7 @@ class Processor(ChunkingService):
             )
 
             __class__.chunk_metric.labels(
-                id=consumer.id, flow=consumer.flow
+                processor=self.id,
             ).observe(chunk_length)
 
             await flow("output").send(r)

trustgraph-flow/trustgraph/chunking/token/chunker.py

@@ -51,7 +51,7 @@ class Processor(ChunkingService):
         if not hasattr(__class__, "chunk_metric"):
             __class__.chunk_metric = Histogram(
                 'chunk_size', 'Chunk size',
-                ["id", "flow"],
+                ["processor"],
                 buckets=[100, 160, 250, 400, 650, 1000, 1600,
                          2500, 4000, 6400, 10000, 16000]
             )
@@ -180,7 +180,7 @@ class Processor(ChunkingService):
             )
 
             __class__.chunk_metric.labels(
-                id=consumer.id, flow=consumer.flow
+                processor=self.id,
             ).observe(chunk_length)
 
             await flow("output").send(r)

trustgraph-flow/trustgraph/config/service/service.py

@@ -106,13 +106,13 @@ class Processor(AsyncProcessor):
         )
 
         config_request_metrics = ConsumerMetrics(
-            processor = self.id, flow = None, name = "config-request"
+            processor=self.id, consumer="config-request",
         )
         config_response_metrics = ProducerMetrics(
-            processor = self.id, flow = None, name = "config-response"
+            processor=self.id, producer="config-response",
         )
         config_push_metrics = ProducerMetrics(
-            processor = self.id, flow = None, name = "config-push"
+            processor=self.id, producer="config-push",
         )
 
         self.config_request_queue_base = config_request_queue
@@ -225,8 +225,8 @@ class Processor(AsyncProcessor):
             topic=resp_queue,
             schema=ConfigResponse,
             metrics=ProducerMetrics(
-                processor=self.id, flow=None,
+                processor=self.id, producer="config-response",
-                name=f"config-response-{workspace_id}",
+                workspace=workspace_id,
             ),
         )
 
@@ -242,8 +242,8 @@ class Processor(AsyncProcessor):
                 workspace=workspace_id,
             ),
             metrics=ConsumerMetrics(
-                processor=self.id, flow=None,
+                processor=self.id, consumer="config-request",
-                name=f"config-request-{workspace_id}",
+                workspace=workspace_id,
             ),
         )
 

trustgraph-flow/trustgraph/cores/service.py

@@ -120,8 +120,8 @@ class Processor(WorkspaceProcessor):
             topic=resp_queue,
             schema=KnowledgeResponse,
             metrics=ProducerMetrics(
-                processor=self.id, flow=None,
+                processor=self.id, producer="knowledge-response",
-                name=f"knowledge-response-{workspace}",
+                workspace=workspace,
             ),
         )
 
@@ -136,8 +136,8 @@ class Processor(WorkspaceProcessor):
                 self.on_knowledge_request, workspace=workspace,
             ),
             metrics=ConsumerMetrics(
-                processor=self.id, flow=None,
+                processor=self.id, consumer="knowledge-request",
-                name=f"knowledge-request-{workspace}",
+                workspace=workspace,
             ),
         )
 

trustgraph-flow/trustgraph/flow/service/service.py

@@ -55,10 +55,10 @@ class Processor(WorkspaceProcessor):
         )
 
         config_req_metrics = ProducerMetrics(
-            processor=self.id, flow=None, name="config-request",
+            processor=self.id, producer="config-request",
         )
         config_resp_metrics = SubscriberMetrics(
-            processor=self.id, flow=None, name="config-response",
+            processor=self.id, subscriber="config-response",
         )
 
         config_rr_id = str(uuid.uuid4())
@@ -100,8 +100,8 @@ class Processor(WorkspaceProcessor):
             topic=resp_queue,
             schema=FlowResponse,
             metrics=ProducerMetrics(
-                processor=self.id, flow=None,
+                processor=self.id, producer="flow-response",
-                name=f"flow-response-{workspace}",
+                workspace=workspace,
             ),
         )
 
@@ -116,8 +116,8 @@ class Processor(WorkspaceProcessor):
                 self.on_flow_request, workspace=workspace,
             ),
             metrics=ConsumerMetrics(
-                processor=self.id, flow=None,
+                processor=self.id, consumer="flow-request",
-                name=f"flow-request-{workspace}",
+                workspace=workspace,
             ),
         )
 

trustgraph-flow/trustgraph/gateway/auth.py

@@ -164,12 +164,12 @@ class IamAuth:
             request_topic=iam_request_queue,
             request_schema=IamRequest,
             request_metrics=ProducerMetrics(
-                processor=self.id, flow=None, name="iam-request",
+                processor=self.id, producer="iam-request",
             ),
             response_topic=iam_response_queue,
             response_schema=IamResponse,
             response_metrics=SubscriberMetrics(
-                processor=self.id, flow=None, name="iam-response",
+                processor=self.id, subscriber="iam-response",
             ),
         )
 

trustgraph-flow/trustgraph/gateway/config/receiver.py

@@ -96,12 +96,10 @@ class ConfigReceiver:
         id = str(uuid.uuid4())
 
         config_req_metrics = ProducerMetrics(
-            processor="api-gateway", flow=None,
+            processor="api-gateway", producer="config-request",
-            name="config-request",
         )
         config_resp_metrics = SubscriberMetrics(
-            processor="api-gateway", flow=None,
+            processor="api-gateway", subscriber="config-response",
-            name="config-response",
         )
 
         return RequestResponse(

trustgraph-flow/trustgraph/iam/noauth/service.py

@@ -44,10 +44,10 @@ class Processor(AsyncProcessor):
         super().__init__(**params)
 
         iam_request_metrics = ConsumerMetrics(
-            processor=self.id, flow=None, name="iam-request",
+            processor=self.id, consumer="iam-request",
         )
         iam_response_metrics = ProducerMetrics(
-            processor=self.id, flow=None, name="iam-response",
+            processor=self.id, producer="iam-response",
         )
 
         self.iam_request_topic = iam_req_q
@@ -88,10 +88,10 @@ class Processor(AsyncProcessor):
     def _create_config_client(self):
         config_rr_id = str(uuid.uuid4())
         config_req_metrics = ProducerMetrics(
-            processor=self.id, flow=None, name="config-request",
+            processor=self.id, producer="config-request",
         )
         config_resp_metrics = SubscriberMetrics(
-            processor=self.id, flow=None, name="config-response",
+            processor=self.id, subscriber="config-response",
         )
         return RequestResponse(
             backend=self.pubsub,

trustgraph-flow/trustgraph/iam/service/service.py

@@ -120,10 +120,10 @@ class Processor(AsyncProcessor):
         )
 
         iam_request_metrics = ConsumerMetrics(
-            processor=self.id, flow=None, name="iam-request",
+            processor=self.id, consumer="iam-request",
         )
         iam_response_metrics = ProducerMetrics(
-            processor=self.id, flow=None, name="iam-response",
+            processor=self.id, producer="iam-response",
         )
 
         self.iam_request_topic = iam_req_q
@@ -179,10 +179,10 @@ class Processor(AsyncProcessor):
         import uuid
         config_rr_id = str(uuid.uuid4())
         config_req_metrics = ProducerMetrics(
-            processor=self.id, flow=None, name="config-request",
+            processor=self.id, producer="config-request",
         )
         config_resp_metrics = SubscriberMetrics(
-            processor=self.id, flow=None, name="config-response",
+            processor=self.id, subscriber="config-response",
         )
         return RequestResponse(
             backend=self.pubsub,

trustgraph-flow/trustgraph/librarian/service.py

@@ -171,7 +171,7 @@ class Processor(WorkspaceProcessor):
 
         # Config service client for collection management
         config_request_metrics = ProducerMetrics(
-            processor = id, flow = None, name = "config-request"
+            processor=id, producer="config-request",
         )
 
         self.config_request_producer = Producer(
@@ -182,7 +182,7 @@ class Processor(WorkspaceProcessor):
         )
 
         config_response_metrics = ConsumerMetrics(
-            processor = id, flow = None, name = "config-response"
+            processor=id, consumer="config-response",
         )
 
         self.config_response_consumer = Consumer(
@@ -258,8 +258,8 @@ class Processor(WorkspaceProcessor):
             topic=lib_resp_queue,
             schema=LibrarianResponse,
             metrics=ProducerMetrics(
-                processor=self.id, flow=None,
+                processor=self.id, producer="librarian-response",
-                name=f"librarian-response-{workspace}",
+                workspace=workspace,
             ),
         )
 
@@ -268,8 +268,8 @@ class Processor(WorkspaceProcessor):
             topic=col_resp_queue,
             schema=CollectionManagementResponse,
             metrics=ProducerMetrics(
-                processor=self.id, flow=None,
+                processor=self.id, producer="collection-response",
-                name=f"collection-response-{workspace}",
+                workspace=workspace,
             ),
         )
 
@@ -284,8 +284,8 @@ class Processor(WorkspaceProcessor):
                 self.on_librarian_request, workspace=workspace,
             ),
             metrics=ConsumerMetrics(
-                processor=self.id, flow=None,
+                processor=self.id, consumer="librarian-request",
-                name=f"librarian-request-{workspace}",
+                workspace=workspace,
             ),
         )
 
@@ -300,8 +300,8 @@ class Processor(WorkspaceProcessor):
                 self.on_collection_request, workspace=workspace,
             ),
             metrics=ConsumerMetrics(
-                processor=self.id, flow=None,
+                processor=self.id, consumer="collection-request",
-                name=f"collection-request-{workspace}",
+                workspace=workspace,
             ),
         )
 

trustgraph-flow/trustgraph/model/text_completion/azure/llm.py

@@ -6,7 +6,6 @@ serverless endpoint service.  Input is prompt, output is response.
 
 import requests
 import json
-from prometheus_client import Histogram
 import os
 import logging
 

trustgraph-flow/trustgraph/model/text_completion/azure_openai/llm.py

@@ -5,7 +5,6 @@ OpenAI endpoit service.  Input is prompt, output is response.
 """
 
 import json
-from prometheus_client import Histogram
 from openai import AzureOpenAI, RateLimitError
 import os
 import logging

trustgraph-flow/trustgraph/model/text_completion/cohere/llm.py

@@ -6,7 +6,6 @@ Input is prompt, output is response.
 
 import cohere
 from cohere.errors import TooManyRequestsError, ServiceUnavailableError
-from prometheus_client import Histogram
 import os
 import logging