langgraph-api 0.3.0__tar.gz → 0.3.4__tar.gz

{langgraph_api-0.3.0 → langgraph_api-0.3.4}/Makefile

@@ -1,4 +1,4 @@
-.PHONY: build release lint format test test_watch start start-inmem start-inmem-license-oss start check-version
+.PHONY: build release lint format test test_watch start start-inmem start-inmem-license-oss start check-version check-base-imports
 
 # lint commands
 
@@ -11,6 +11,9 @@ format:
 	uv run ruff check --fix .
 	uv run ruff format .
 
+check-base-imports:
+	LANGGRAPH_RUNTIME_EDITION=inmem DATABASE_URI=:memory: REDIS_URI=_FAKE uv run python -c "from langgraph_api.config import *; from langgraph_runtime import *"
+
 # test commands
 
 TEST ?= tests/

{langgraph_api-0.3.0 → langgraph_api-0.3.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: langgraph-api
-Version: 0.3.0
+Version: 0.3.4
 Author-email: Nuno Campos <nuno@langchain.dev>, Will Fu-Hinthorn <will@langchain.dev>
 License: Elastic-2.0
 License-File: LICENSE

langgraph_api-0.3.4/langgraph_api/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.3.4"

{langgraph_api-0.3.0 → langgraph_api-0.3.4}/langgraph_api/api/meta.py

@@ -54,10 +54,16 @@ async def meta_metrics(request: ApiRequest):
         metadata.PROJECT_ID, metadata.HOST_REVISION_ID, metrics_format
     )
 
+    pg_redis_stats = pool_stats(
+        project_id=metadata.PROJECT_ID,
+        revision_id=metadata.HOST_REVISION_ID,
+        format=metrics_format,
+    )
+
     if metrics_format == "json":
         async with connect() as conn:
             resp = {
-                **pool_stats(),
+                **pg_redis_stats,
                 "queue": await Runs.stats(conn),
                 **http_metrics,
             }
@@ -93,6 +99,7 @@ async def meta_metrics(request: ApiRequest):
                 )
 
             metrics.extend(http_metrics)
+            metrics.extend(pg_redis_stats)
 
         metrics_response = "\n".join(metrics)
         return PlainTextResponse(metrics_response)

{langgraph_api-0.3.0 → langgraph_api-0.3.4}/langgraph_api/config.py

@@ -180,6 +180,7 @@ REDIS_CLUSTER = env("REDIS_CLUSTER", cast=bool, default=False)
 REDIS_MAX_CONNECTIONS = env("REDIS_MAX_CONNECTIONS", cast=int, default=2000)
 REDIS_CONNECT_TIMEOUT = env("REDIS_CONNECT_TIMEOUT", cast=float, default=10.0)
 REDIS_MAX_IDLE_TIME = env("REDIS_MAX_IDLE_TIME", cast=float, default=120.0)
+REDIS_STREAM_TIMEOUT = env("REDIS_STREAM_TIMEOUT", cast=float, default=30.0)
 REDIS_KEY_PREFIX = env("REDIS_KEY_PREFIX", cast=str, default="")
 RUN_STATS_CACHE_SECONDS = env("RUN_STATS_CACHE_SECONDS", cast=int, default=60)
 
@@ -374,6 +375,7 @@ API_VARIANT = env("LANGSMITH_LANGGRAPH_API_VARIANT", cast=str, default="")
 # UI
 UI_USE_BUNDLER = env("LANGGRAPH_UI_BUNDLER", cast=bool, default=False)
 IS_QUEUE_ENTRYPOINT = False
+IS_EXECUTOR_ENTRYPOINT = False
 ref_sha = None
 if not os.getenv("LANGCHAIN_REVISION_ID") and (
     ref_sha := os.getenv("LANGSMITH_LANGGRAPH_GIT_REF_SHA")

{langgraph_api-0.3.0 → langgraph_api-0.3.4}/langgraph_api/executor_entrypoint.py

@@ -20,4 +20,7 @@ if __name__ == "__main__":
         uvloop.install()
     except ImportError:
         pass
+    from langgraph_api import config
+
+    config.IS_EXECUTOR_ENTRYPOINT = True
     asyncio.run(main(grpc_port=args.grpc_port, entrypoint_name="python-executor"))

{langgraph_api-0.3.0 → langgraph_api-0.3.4}/langgraph_api/queue_entrypoint.py

@@ -11,10 +11,8 @@ if not (
 
 import asyncio
 import contextlib
-import http.server
 import json
 import logging.config
-import os
 import pathlib
 import signal
 from contextlib import asynccontextmanager
@@ -22,6 +20,7 @@ from typing import cast
 
 import structlog
 
+from langgraph_runtime.database import pool_stats
 from langgraph_runtime.lifespan import lifespan
 from langgraph_runtime.metrics import get_metrics
 
@@ -29,69 +28,68 @@ logger = structlog.stdlib.get_logger(__name__)
 
 
 async def health_and_metrics_server():
+    import uvicorn
+    from starlette.applications import Starlette
+    from starlette.responses import JSONResponse, PlainTextResponse
+    from starlette.routing import Route
+
     port = int(os.getenv("PORT", "8080"))
-    ok = json.dumps({"status": "ok"}).encode()
-    ok_len = str(len(ok))
-
-    class HealthAndMetricsHandler(http.server.SimpleHTTPRequestHandler):
-        def log_message(self, format, *args):
-            # Skip logging for /ok and /metrics endpoints
-            if getattr(self, "path", None) in ["/ok", "/metrics"]:
-                return
-            # Log other requests normally
-            super().log_message(format, *args)
-
-        def do_GET(self):
-            path = getattr(self, "path", None)
-            if path == "/ok":
-                self.send_response(200)
-                self.send_header("Content-Type", "application/json")
-                self.send_header("Content-Length", ok_len)
-                self.end_headers()
-                self.wfile.write(ok)
-            elif path == "/metrics":
-                metrics = get_metrics()
-                worker_metrics = cast(dict[str, int], metrics["workers"])
-                workers_max = worker_metrics["max"]
-                workers_active = worker_metrics["active"]
-                workers_available = worker_metrics["available"]
-
-                project_id = os.getenv("LANGSMITH_HOST_PROJECT_ID")
-                revision_id = os.getenv("LANGSMITH_HOST_REVISION_ID")
-
-                metrics = [
-                    "# HELP lg_api_workers_max The maximum number of workers available.",
-                    "# TYPE lg_api_workers_max gauge",
-                    f'lg_api_workers_max{{project_id="{project_id}", revision_id="{revision_id}"}} {workers_max}',
-                    "# HELP lg_api_workers_active The number of currently active workers.",
-                    "# TYPE lg_api_workers_active gauge",
-                    f'lg_api_workers_active{{project_id="{project_id}", revision_id="{revision_id}"}} {workers_active}',
-                    "# HELP lg_api_workers_available The number of available (idle) workers.",
-                    "# TYPE lg_api_workers_available gauge",
-                    f'lg_api_workers_available{{project_id="{project_id}", revision_id="{revision_id}"}} {workers_available}',
-                ]
-
-                metrics_response = "\n".join(metrics).encode()
-                metrics_len = str(len(metrics_response))
-
-                self.send_response(200)
-                self.send_header(
-                    "Content-Type", "text/plain; version=0.0.4; charset=utf-8"
-                )
-                self.send_header("Content-Length", metrics_len)
-                self.end_headers()
-                self.wfile.write(metrics_response)
-            else:
-                self.send_error(http.HTTPStatus.NOT_FOUND)
-
-    with http.server.ThreadingHTTPServer(
-        ("0.0.0.0", port), HealthAndMetricsHandler
-    ) as httpd:
-        logger.info(f"Health and metrics server started at http://0.0.0.0:{port}")
-        try:
-            await asyncio.to_thread(httpd.serve_forever)
-        finally:
-            httpd.shutdown()
+
+    async def health_endpoint(request):
+        return JSONResponse({"status": "ok"})
+
+    async def metrics_endpoint(request):
+        metrics = get_metrics()
+        worker_metrics = cast(dict[str, int], metrics["workers"])
+        workers_max = worker_metrics["max"]
+        workers_active = worker_metrics["active"]
+        workers_available = worker_metrics["available"]
+
+        project_id = os.getenv("LANGSMITH_HOST_PROJECT_ID")
+        revision_id = os.getenv("LANGSMITH_HOST_REVISION_ID")
+
+        metrics_lines = [
+            "# HELP lg_api_workers_max The maximum number of workers available.",
+            "# TYPE lg_api_workers_max gauge",
+            f'lg_api_workers_max{{project_id="{project_id}", revision_id="{revision_id}"}} {workers_max}',
+            "# HELP lg_api_workers_active The number of currently active workers.",
+            "# TYPE lg_api_workers_active gauge",
+            f'lg_api_workers_active{{project_id="{project_id}", revision_id="{revision_id}"}} {workers_active}',
+            "# HELP lg_api_workers_available The number of available (idle) workers.",
+            "# TYPE lg_api_workers_available gauge",
+            f'lg_api_workers_available{{project_id="{project_id}", revision_id="{revision_id}"}} {workers_available}',
+        ]
+
+        metrics_lines.extend(
+            pool_stats(
+                project_id=project_id,
+                revision_id=revision_id,
+            )
+        )
+
+        return PlainTextResponse(
+            "\n".join(metrics_lines),
+            media_type="text/plain; version=0.0.4; charset=utf-8",
+        )
+
+    app = Starlette(
+        routes=[
+            Route("/ok", health_endpoint),
+            Route("/metrics", metrics_endpoint),
+        ]
+    )
+
+    config = uvicorn.Config(
+        app,
+        host="0.0.0.0",
+        port=port,
+        log_level="error",
+        access_log=False,
+    )
+    server = uvicorn.Server(config)
+
+    logger.info(f"Health and metrics server started at http://0.0.0.0:{port}")
+    await server.serve()
 
 
 async def entrypoint(

{langgraph_api-0.3.0 → langgraph_api-0.3.4}/langgraph_api/stream.py

@@ -2,7 +2,7 @@ import uuid
 from collections.abc import AsyncIterator, Callable
 from contextlib import AsyncExitStack, aclosing, asynccontextmanager
 from functools import lru_cache
-from typing import Any, cast
+from typing import Any, Literal, cast
 
 import langgraph.version
 import langsmith
@@ -423,7 +423,7 @@ async def consume(
     stream: AnyStream,
     run_id: str | uuid.UUID,
     resumable: bool = False,
-    stream_modes: set[StreamMode] | None = None,
+    stream_modes: set[StreamMode | Literal["metadata"]] | None = None,
 ) -> None:
     stream_modes = stream_modes or set()
     if "messages-tuple" in stream_modes:

langgraph_api-0.3.0/langgraph_api/__init__.py

@@ -1 +0,0 @@
-__version__ = "0.3.0"