langgraph-api 0.10.0rc2__tar.gz → 0.11.0.dev3__tar.gz

{langgraph_api-0.10.0rc2 → langgraph_api-0.11.0.dev3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: langgraph-api
-Version: 0.10.0rc2
+Version: 0.11.0.dev3
 Author-email: Will Fu-Hinthorn <will@langchain.dev>, Josh Rogers <josh@langchain.dev>, Parker Rule <parker@langchain.dev>
 License: Elastic-2.0
 License-File: LICENSE
@@ -22,8 +22,10 @@ Requires-Dist: langgraph<2,>=0.4.10
 Requires-Dist: langsmith[otel]>=0.6.3
 Requires-Dist: opentelemetry-api>=0.0.1
 Requires-Dist: opentelemetry-exporter-otlp-proto-http>=0.0.1
+Requires-Dist: opentelemetry-exporter-prometheus>=0.62b1
 Requires-Dist: opentelemetry-sdk>=0.0.1
 Requires-Dist: orjson>=3.9.7
+Requires-Dist: prometheus-client>=0.0.1
 Requires-Dist: protobuf<7.0.0,>=6.32.1
 Requires-Dist: pyjwt>=2.9.0
 Requires-Dist: sse-starlette<3.4.0,>=2.1.3

langgraph_api-0.11.0.dev3/langgraph_api/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.11.0.dev3"

{langgraph_api-0.10.0rc2 → langgraph_api-0.11.0.dev3}/langgraph_api/config/__init__.py

@@ -578,6 +578,8 @@ METRIC_MAX_EMITTING_TIER = env(
     "METRIC_MAX_EMITTING_TIER", cast=int, default=_METRIC_MAX_EMITTING_TIER_DEFAULT
 )
 DATADOG_METRICS_ENABLED = bool(LSD_DD_API_KEY)
+LSD_PROM_METRICS_ENABLED = env("LSD_PROM_METRICS_ENABLED", cast=bool, default=False)
+LSD_PROM_METRICS_PORT = env("LSD_PROM_METRICS_PORT", cast=int, default=9464)
 LANGGRAPH_LOGS_ENDPOINT = env("LANGGRAPH_LOGS_ENDPOINT", cast=str, default=None)
 LANGGRAPH_LOGS_ENABLED = env("LANGGRAPH_LOGS_ENABLED", cast=bool, default=False)
 
@@ -672,6 +674,8 @@ __all__ = [
     "LSD_GRPC_SERVER_ADDRESS",
     "LSD_GRPC_SERVER_MAX_RECV_MSG_BYTES",
     "LSD_GRPC_SERVER_MAX_SEND_MSG_BYTES",
+    "LSD_PROM_METRICS_ENABLED",
+    "LSD_PROM_METRICS_PORT",
     "MAX_STREAM_CHUNK_SIZE_BYTES",
     "METRIC_MAX_EMITTING_TIER",
     "METRIC_PREFIX",

{langgraph_api-0.10.0rc2 → langgraph_api-0.11.0.dev3}/langgraph_api/event_streaming/session.py

@@ -39,7 +39,7 @@ from langgraph_api.event_streaming.types import (
     NamespaceInfo,
     Subscription,
 )
-from langgraph_api.metrics_datadog import (
+from langgraph_api.metrics_otlp import (
     COUNTER_PROTOCOL_V2_BUFFER_EVICTED,
     COUNTER_PROTOCOL_V2_EVENT_EMITTED,
     COUNTER_PROTOCOL_V2_RESUME_GAP,
@@ -47,7 +47,7 @@ from langgraph_api.metrics_datadog import (
     COUNTER_STREAMING_DATA_LOSS,
     GAUGE_PROTOCOL_V2_BUFFER_SIZE,
     HISTOGRAM_PROTOCOL_V2_REPLAYED_EVENTS,
-    get_datadog_metrics_reporter,
+    get_otlp_metrics_reporter,
 )
 
 logger = structlog.stdlib.get_logger(__name__)
@@ -937,7 +937,7 @@ class EventStreamingSession:
 
     async def _push_event(self, event: dict[str, Any]) -> None:
         self._buffer.append(event)
-        reporter = get_datadog_metrics_reporter()
+        reporter = get_otlp_metrics_reporter()
         if len(self._buffer) > self._max_buffer_size:
             evicted = len(self._buffer) - self._max_buffer_size
             self._buffer = self._buffer[-self._max_buffer_size :]
@@ -1023,7 +1023,7 @@ class EventStreamingSession:
             min_available_seq = self._buffer[0].get("seq", 0)
             if isinstance(min_available_seq, int) and since + 1 < min_available_seq:
                 try:
-                    get_datadog_metrics_reporter().inc_counter(
+                    get_otlp_metrics_reporter().inc_counter(
                         COUNTER_PROTOCOL_V2_RESUME_GAP
                     )
                 except Exception:
@@ -1058,7 +1058,7 @@ class EventStreamingSession:
 
         subscription.active = True
         try:
-            get_datadog_metrics_reporter().record_histogram(
+            get_otlp_metrics_reporter().record_histogram(
                 HISTOGRAM_PROTOCOL_V2_REPLAYED_EVENTS,
                 float(replayed),
             )
@@ -1129,7 +1129,7 @@ class EventStreamingSession:
             # session down shortly after the first failure.
             self._transport_broken = True
             try:
-                get_datadog_metrics_reporter().inc_counter(
+                get_otlp_metrics_reporter().inc_counter(
                     COUNTER_PROTOCOL_V2_TRANSPORT_SEND_FAILURE,
                     attributes={"method": _metric_event_method(message)},
                 )

langgraph_api-0.10.0rc2/langgraph_api/metrics_datadog.py → langgraph_api-0.11.0.dev3/langgraph_api/metrics_otlp.py

@@ -48,6 +48,16 @@ else:
         MetricExporter = object
         OTEL_AVAILABLE = False
 
+    try:
+        from opentelemetry.exporter.prometheus import PrometheusMetricReader
+        from prometheus_client import start_http_server
+
+        PROMETHEUS_EXPORTER_AVAILABLE = True
+    except ModuleNotFoundError:
+        PrometheusMetricReader = None
+        start_http_server = None
+        PROMETHEUS_EXPORTER_AVAILABLE = False
+
 logger = structlog.stdlib.get_logger(__name__)
 
 SERVICE_NAME = "lsd_langgraph_api"
@@ -213,7 +223,7 @@ class _FilteringExporter(MetricExporter):
         return self._exporter.force_flush(timeout_millis)
 
 
-class DatadogMetricsReporter:
+class OTelMetricsReporter:
     def __init__(self) -> None:
         self._lock = threading.Lock()
         self._initialized = False
@@ -222,6 +232,8 @@ class DatadogMetricsReporter:
         self._meter = None
         self._max_tier = _normalize_emitting_tier(config.METRIC_MAX_EMITTING_TIER)
         self._instruments: dict[str, Any] = {}
+        self._prom_enabled = False
+        self._prom_server: Any | None = None
 
     @property
     def enabled(self) -> bool:
@@ -233,12 +245,17 @@ class DatadogMetricsReporter:
                 return
             self._initialized = True
 
-            if not config.DATADOG_METRICS_ENABLED:
-                logger.info("Datadog metrics disabled (no DD API key configured)")
+            if (
+                not config.DATADOG_METRICS_ENABLED
+                and not config.LSD_PROM_METRICS_ENABLED
+            ):
+                logger.info(
+                    "OTel metrics disabled (no DD API key and Prometheus not enabled)"
+                )
                 return
             if not OTEL_AVAILABLE:
                 logger.warning(
-                    "Datadog metrics disabled because OpenTelemetry dependencies are not installed"
+                    "OTel metrics disabled because OpenTelemetry dependencies are not installed"
                 )
                 return
 
@@ -247,7 +264,6 @@ class DatadogMetricsReporter:
                     {
                         "service.name": SERVICE_NAME,
                         "host.id": os.getenv("HOSTNAME", ""),
-                        # Not in public docs: these LANGSMITH_* vars are set by SaaS control plane
                         "api_version": os.getenv("LANGSMITH_LANGGRAPH_API_VERSION")
                         or __version__,
                         "project_id": os.getenv("LANGSMITH_HOST_PROJECT_ID", ""),
@@ -260,58 +276,117 @@ class DatadogMetricsReporter:
                         "deployment_type": os.getenv("LSD_DEPLOYMENT_TYPE", ""),
                     }
                 )
-                base_exporter = OTLPMetricExporter(
-                    endpoint=f"https://{config.LSD_DD_ENDPOINT}/v1/metrics",
-                    headers={
-                        "dd-api-key": config.LSD_DD_API_KEY,
-                        "dd-otel-metric-config": DD_OTEL_METRIC_CONFIG,
-                    },
-                    preferred_temporality={
-                        Counter: AggregationTemporality.DELTA,
-                        Histogram: AggregationTemporality.DELTA,
-                    },
-                )
-                # Burn after reading: remove DD keys from env so user code
-                # and child processes cannot access them.
-                for key in (
-                    "LSD_DD_API_KEY",
-                    "LSD_DD_ENDPOINT",
-                ):
-                    os.environ.pop(key, None)
-
-                reader = PeriodicExportingMetricReader(
-                    _FilteringExporter(base_exporter),
-                    export_interval_millis=10_000,
-                )
+
+                readers: list[Any] = []
+
+                if config.DATADOG_METRICS_ENABLED:
+                    base_exporter = OTLPMetricExporter(
+                        endpoint=f"https://{config.LSD_DD_ENDPOINT}/v1/metrics",
+                        headers={
+                            "dd-api-key": config.LSD_DD_API_KEY,
+                            "dd-otel-metric-config": DD_OTEL_METRIC_CONFIG,
+                        },
+                        preferred_temporality={
+                            Counter: AggregationTemporality.DELTA,
+                            Histogram: AggregationTemporality.DELTA,
+                        },
+                    )
+                    # Burn after reading: remove DD keys from env so user code
+                    # and child processes cannot access them.
+                    for key in (
+                        "LSD_DD_API_KEY",
+                        "LSD_DD_ENDPOINT",
+                    ):
+                        os.environ.pop(key, None)
+
+                    readers.append(
+                        PeriodicExportingMetricReader(
+                            _FilteringExporter(base_exporter),
+                            export_interval_millis=10_000,
+                        )
+                    )
+
+                if config.LSD_PROM_METRICS_ENABLED:
+                    if not PROMETHEUS_EXPORTER_AVAILABLE:
+                        logger.warning(
+                            "Prometheus metrics disabled: opentelemetry-exporter-prometheus not installed"
+                        )
+                    else:
+                        # PrometheusMetricReader registers its collector with the
+                        # global prometheus_client REGISTRY, which start_http_server
+                        # serves by default. MeterProvider.shutdown() unregisters it.
+                        readers.append(PrometheusMetricReader())
+                        self._prom_enabled = True
+
+                if not readers:
+                    logger.info(
+                        "OTel metrics reporter disabled (no backend configured)"
+                    )
+                    return
+
                 self._meter_provider = MeterProvider(
-                    resource=resource, metric_readers=[reader]
+                    resource=resource, metric_readers=readers
                 )
                 self._meter = self._meter_provider.get_meter(SERVICE_NAME)
                 self._enabled = True
+
+                if config.DATADOG_METRICS_ENABLED:
+                    logger.info(
+                        "Datadog OTLP metrics reader initialized",
+                        endpoint=f"https://{config.LSD_DD_ENDPOINT}/v1/metrics",
+                    )
+                if self._prom_enabled:
+                    self._start_prometheus_server()
+
                 logger.info(
-                    "Initialized Datadog metrics reporter",
-                    endpoint=f"https://{config.LSD_DD_ENDPOINT}/v1/metrics",
+                    "OTel metrics reporter initialized",
                     metric_prefix=METRIC_NAME_PREFIX,
                     max_emitting_tier=self._max_tier,
+                    backends=[
+                        b
+                        for b in (
+                            "datadog" if config.DATADOG_METRICS_ENABLED else None,
+                            "prometheus" if self._prom_enabled else None,
+                        )
+                        if b
+                    ],
                 )
-                logger.info("Datadog metrics reporter initialized")
             except Exception:
                 self._enabled = False
                 self._meter_provider = None
                 self._meter = None
-                logger.exception("Failed to initialize Datadog metrics reporter")
+                logger.exception("Failed to initialize OTel metrics reporter")
                 raise
 
+    def _start_prometheus_server(self) -> None:
+        port = config.LSD_PROM_METRICS_PORT
+        # ``start_http_server`` spins up a WSGI server in a daemon thread serving
+        # the global prometheus_client REGISTRY, and returns the (server, thread)
+        # handle. Keep the server so ``shutdown`` can stop it cleanly (e.g. in
+        # tests); in production the daemon thread exits with the process anyway.
+        server, _thread = start_http_server(port=port)
+        self._prom_server = server
+        logger.info("Prometheus metrics scrape server started", port=port)
+
     def shutdown(self) -> None:
         with self._lock:
             if self._meter_provider:
                 try:
                     self._meter_provider.shutdown()
                 except Exception:
-                    logger.exception("Failed to shutdown Datadog metrics reporter")
+                    logger.exception("Failed to shutdown OTel metrics reporter")
                 finally:
                     self._meter_provider = None
                     self._meter = None
+            if self._prom_server is not None:
+                try:
+                    self._prom_server.shutdown()
+                    self._prom_server.server_close()
+                except Exception:
+                    logger.exception("Failed to stop Prometheus scrape server")
+                finally:
+                    self._prom_server = None
+            self._prom_enabled = False
             self._enabled = False
             self._initialized = False
             self._instruments.clear()
@@ -352,7 +427,7 @@ class DatadogMetricsReporter:
         try:
             instrument.add(value, attributes or {})
         except Exception:
-            logger.warning("Failed to add Datadog counter", metric_name=metric.name)
+            logger.warning("Failed to add counter", metric_name=metric.name)
 
     def record_histogram(
         self,
@@ -368,9 +443,7 @@ class DatadogMetricsReporter:
         try:
             instrument.record(value, attributes or {})
         except Exception:
-            logger.warning(
-                "Failed to record Datadog histogram", metric_name=metric.name
-            )
+            logger.warning("Failed to record histogram", metric_name=metric.name)
 
     def record_latency(
         self,
@@ -391,7 +464,7 @@ class DatadogMetricsReporter:
         try:
             instrument.record(value, attributes or {})
         except Exception:
-            logger.warning("Failed to record Datadog latency", metric_name=metric.name)
+            logger.warning("Failed to record latency", metric_name=metric.name)
 
     def record_gauge(
         self,
@@ -407,7 +480,7 @@ class DatadogMetricsReporter:
         try:
             instrument.set(value, attributes or {})
         except Exception:
-            logger.warning("Failed to record Datadog gauge", metric_name=metric.name)
+            logger.warning("Failed to record gauge", metric_name=metric.name)
 
     @contextmanager
     def track_latency_ms(
@@ -429,11 +502,11 @@ class DatadogMetricsReporter:
             )
 
 
-_metrics_reporter: DatadogMetricsReporter | None = None
+_metrics_reporter: OTelMetricsReporter | None = None
 
 
-def get_datadog_metrics_reporter() -> DatadogMetricsReporter:
+def get_otlp_metrics_reporter() -> OTelMetricsReporter:
     global _metrics_reporter
     if _metrics_reporter is None:
-        _metrics_reporter = DatadogMetricsReporter()
+        _metrics_reporter = OTelMetricsReporter()
     return _metrics_reporter

{langgraph_api-0.10.0rc2 → langgraph_api-0.11.0.dev3}/langgraph_api/stream.py

@@ -44,12 +44,12 @@ from langgraph_api.graph import get_graph
 from langgraph_api.grpc.ops.runs import StreamPublishException
 from langgraph_api.js.base import BaseRemotePregel
 from langgraph_api.metadata import HOST, PLAN, USER_API_URL, incr_nodes
-from langgraph_api.metrics_datadog import (
+from langgraph_api.metrics_otlp import (
     COUNTER_STREAMING_DATA_LOSS,
     GAUGE_PUBLISH_QUEUE_AVAILABILITY,
     HISTOGRAM_STREAM_DATA_SIZE,
     LATENCY_STREAM_PUBLISH,
-    get_datadog_metrics_reporter,
+    get_otlp_metrics_reporter,
 )
 from langgraph_api.schema import Run, StreamMode
 from langgraph_api.serde import json_dumpb
@@ -569,7 +569,7 @@ async def consume(
     *,
     thread_id: str | uuid.UUID,
 ) -> None:
-    reporter = get_datadog_metrics_reporter()
+    reporter = get_otlp_metrics_reporter()
     stream_modes = stream_modes or set()
     if "messages-tuple" in stream_modes:
         stream_modes.add("messages")

{langgraph_api-0.10.0rc2 → langgraph_api-0.11.0.dev3}/langgraph_api/worker.py

@@ -29,7 +29,7 @@ from langgraph_api.feature_flags import IS_POSTGRES_OR_GRPC_BACKEND
 from langgraph_api.graph import restore_dd_trace_context
 from langgraph_api.js.errors import RemoteException
 from langgraph_api.metadata import incr_runs
-from langgraph_api.metrics_datadog import (
+from langgraph_api.metrics_otlp import (
     COUNTER_GRAPH_RECURSION_LIMIT_ERROR,
     COUNTER_RUN_ATTEMPT_STARTED,
     COUNTER_RUN_CANCELED_BY_REQUEST,
@@ -39,7 +39,7 @@ from langgraph_api.metrics_datadog import (
     COUNTER_RUN_SET_STATUS_ERROR,
     COUNTER_RUN_SUCCESS,
     LATENCY_RUN_EXECUTION,
-    get_datadog_metrics_reporter,
+    get_otlp_metrics_reporter,
 )
 from langgraph_api.otel_context import restore_otel_trace_context
 from langgraph_api.schema import RUN_KWARGS_ENCRYPTION_SUBFIELDS, Run, StreamMode
@@ -114,7 +114,7 @@ async def worker(
     *,
     encryption_context: dict[str, Any] | None = None,
 ) -> WorkerResult:
-    reporter = get_datadog_metrics_reporter()
+    reporter = get_otlp_metrics_reporter()
     run_id = run["run_id"]
     if attempt == 1:
         incr_runs()

{langgraph_api-0.10.0rc2 → langgraph_api-0.11.0.dev3}/pyproject.toml

@@ -48,6 +48,8 @@ dependencies = [
     "opentelemetry-api>=0.0.1",
     "opentelemetry-sdk>=0.0.1",
     "opentelemetry-exporter-otlp-proto-http>=0.0.1",
+    "opentelemetry-exporter-prometheus>=0.62b1",
+    "prometheus-client>=0.0.1",
     "uuid-utils>=0.12.0",
     "zstandard>=0.23.0",
     "langchain-protocol>=0.0.16,<0.1",

{langgraph_api-0.10.0rc2 → langgraph_api-0.11.0.dev3}/uv.lock

@@ -732,8 +732,10 @@ dependencies = [
     { name = "langsmith", extra = ["otel"] },
     { name = "opentelemetry-api" },
     { name = "opentelemetry-exporter-otlp-proto-http" },
+    { name = "opentelemetry-exporter-prometheus" },
     { name = "opentelemetry-sdk" },
     { name = "orjson" },
+    { name = "prometheus-client" },
     { name = "protobuf" },
     { name = "pyjwt" },
     { name = "sse-starlette" },
@@ -793,8 +795,10 @@ requires-dist = [
     { name = "langsmith", extras = ["otel"], specifier = ">=0.6.3" },
     { name = "opentelemetry-api", specifier = ">=0.0.1" },
     { name = "opentelemetry-exporter-otlp-proto-http", specifier = ">=0.0.1" },
+    { name = "opentelemetry-exporter-prometheus", specifier = ">=0.62b1" },
     { name = "opentelemetry-sdk", specifier = ">=0.0.1" },
     { name = "orjson", specifier = ">=3.9.7" },
+    { name = "prometheus-client", specifier = ">=0.0.1" },
     { name = "protobuf", specifier = ">=6.32.1,<7.0.0" },
     { name = "pyjwt", specifier = ">=2.9.0" },
     { name = "sse-starlette", specifier = ">=2.1.3,<3.4.0" },
@@ -1123,6 +1127,20 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/ba/4d/ef07ff2fc630849f2080ae0ae73a61f67257905b7ac79066640bfa0c5739/opentelemetry_exporter_otlp_proto_http-1.41.1-py3-none-any.whl", hash = "sha256:1a21e8f49c7a946d935551e90947d6c3eb39236723c6624401da0f33d68edcb4", size = 22673, upload-time = "2026-04-24T13:15:21.313Z" },
 ]
 
+[[package]]
+name = "opentelemetry-exporter-prometheus"
+version = "0.62b1"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "opentelemetry-api" },
+    { name = "opentelemetry-sdk" },
+    { name = "prometheus-client" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/1b/03/e1fbf14386ef171b949b71fba7c18643691a9390ab38c221df582e916569/opentelemetry_exporter_prometheus-0.62b1.tar.gz", hash = "sha256:7ecbac9aa76e7abb44082ab0ff2983e0a573e4091c4653f7db483b02bae03506", size = 15446, upload-time = "2026-04-24T13:15:43.783Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/c8/d2/ee4002b88e20c59fae52bed008a63c6f7eff7d498f302032f6b0434a6de7/opentelemetry_exporter_prometheus-0.62b1-py3-none-any.whl", hash = "sha256:7a0b8a6402e107e1f93e38f074a668797e1103936b189561959531a67ffeba55", size = 13278, upload-time = "2026-04-24T13:15:22.485Z" },
+]
+
 [[package]]
 name = "opentelemetry-proto"
 version = "1.41.1"
@@ -1314,6 +1332,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/54/20/4d324d65cc6d9205fabedc306948156824eb9f0ee1633355a8f7ec5c66bf/pluggy-1.6.0-py3-none-any.whl", hash = "sha256:e920276dd6813095e9377c0bc5566d94c932c33b27a3e3945d8389c374dd4746", size = 20538, upload-time = "2025-05-15T12:30:06.134Z" },
 ]
 
+[[package]]
+name = "prometheus-client"
+version = "0.25.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/1b/fb/d9aa83ffe43ce1f19e557c0971d04b90561b0cfd50762aafb01968285553/prometheus_client-0.25.0.tar.gz", hash = "sha256:5e373b75c31afb3c86f1a52fa1ad470c9aace18082d39ec0d2f918d11cc9ba28", size = 86035, upload-time = "2026-04-09T19:53:42.359Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/8d/9b/d4b1e644385499c8346fa9b622a3f030dce14cd6ef8a1871c221a17a67e7/prometheus_client-0.25.0-py3-none-any.whl", hash = "sha256:d5aec89e349a6ec230805d0df882f3807f74fd6c1a2fa86864e3c2279059fed1", size = 64154, upload-time = "2026-04-09T19:53:41.324Z" },
+]
+
 [[package]]
 name = "protobuf"
 version = "6.33.6"

langgraph_api-0.10.0rc2/langgraph_api/__init__.py

@@ -1 +0,0 @@
-__version__ = "0.10.0rc2"