langgraph-api 0.8.0__tar.gz → 0.8.2__tar.gz

{langgraph_api-0.8.0 → langgraph_api-0.8.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: langgraph-api
-Version: 0.8.0
+Version: 0.8.2
 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
@@ -15,7 +15,7 @@ Requires-Dist: httpx>=0.25.0
 Requires-Dist: jsonschema-rs<0.45,>=0.20.0
 Requires-Dist: langchain-core>=0.3.64
 Requires-Dist: langgraph-checkpoint<5,>=3.0.1
-Requires-Dist: langgraph-runtime-inmem<0.28.0,>=0.27.0
+Requires-Dist: langgraph-runtime-inmem<0.29.0,>=0.28.0
 Requires-Dist: langgraph-sdk>=0.3.5
 Requires-Dist: langgraph<2,>=0.4.10
 Requires-Dist: langsmith[otel]>=0.6.3

langgraph_api-0.8.2/langgraph_api/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.8.2"

{langgraph_api-0.8.0 → langgraph_api-0.8.2}/langgraph_api/api/runs.py

@@ -1,7 +1,7 @@
 import asyncio
 from collections.abc import AsyncIterator, Awaitable, Callable
 from typing import Any, Literal, cast
-from uuid import UUID, uuid4
+from uuid import UUID
 
 import orjson
 import structlog
@@ -374,7 +374,7 @@ async def stream_run_stateless(
     payload["if_not_exists"] = "create"
     on_disconnect = payload.get("on_disconnect", "continue")
     run_id = uuid7()
-    thread_id = uuid4()
+    thread_id = uuid7()
 
     sub = await Runs.Stream.subscribe(run_id, thread_id)
     try:
@@ -468,7 +468,7 @@ async def wait_run_stateless(request: ApiRequest):
     payload["if_not_exists"] = "create"
     on_disconnect = payload.get("on_disconnect", "continue")
     run_id = uuid7()
-    thread_id = uuid4()
+    thread_id = uuid7()
 
     sub = await Runs.Stream.subscribe(run_id, thread_id)
     try:
@@ -533,7 +533,7 @@ async def list_runs(
 
     async with connect() as conn:
         thread, runs = await asyncio.gather(
-            Threads.get(conn, thread_id),
+            Threads.get(conn, thread_id, read_mask_paths=[]),
             Runs.search(
                 conn,
                 thread_id,
@@ -562,7 +562,7 @@ async def get_run(request: ApiRequest):
 
     async with connect() as conn:
         thread, run = await asyncio.gather(
-            Threads.get(conn, thread_id),
+            Threads.get(conn, thread_id, read_mask_paths=[]),
             Runs.get(
                 conn,
                 run_id,

{langgraph_api-0.8.0 → langgraph_api-0.8.2}/langgraph_api/api/threads.py

@@ -1,5 +1,4 @@
 from typing import get_args
-from uuid import uuid4
 
 from starlette.exceptions import HTTPException
 from starlette.responses import Response
@@ -26,6 +25,7 @@ from langgraph_api.state import state_snapshot_to_thread_state
 from langgraph_api.utils import (
     fetchone,
     get_pagination_headers,
+    uuid7,
     validate_select_columns,
     validate_stream_id,
     validate_uuid,
@@ -69,7 +69,7 @@ async def create_thread(
             ["metadata"],
         )
 
-    thread_id = thread_id or str(uuid4())
+    thread_id = thread_id or str(uuid7())
     supersteps = payload.get("supersteps")
 
     if IS_POSTGRES_OR_GRPC_BACKEND and not supersteps:

{langgraph_api-0.8.0 → langgraph_api-0.8.2}/langgraph_api/config/__init__.py

@@ -45,6 +45,17 @@ DATABASE_URI = env("DATABASE_URI", cast=str, default=getenv("POSTGRES_URI", unde
 # Not in public docs: infrastructure, set by platform
 MIGRATIONS_PATH = env("MIGRATIONS_PATH", cast=str, default="/storage/migrations")
 POSTGRES_POOL_MAX_SIZE = env("LANGGRAPH_POSTGRES_POOL_MAX_SIZE", cast=int, default=150)
+
+# Checkpoint ingestion batch controls
+# Go defaults (core/config/config.go): CHECKPOINT_MAX_BATCH_SIZE=1000, CHECKPOINT_BATCH_DELAY=0.005 (5ms)
+# CHECKPOINT_BATCH_DELAY uses float seconds to match Go's time.Duration env parsing convention.
+# storage_postgres/Makefile sets these to the Go defaults for CI/local dev.
+# TODO(braa): flip these to match Go defaults once validated in production (LSD-1404)
+CHECKPOINT_MAX_BATCH_SIZE: int | None = env(
+    "CHECKPOINT_MAX_BATCH_SIZE", cast=int, default=None
+)
+CHECKPOINT_BATCH_DELAY: float = env("CHECKPOINT_BATCH_DELAY", cast=float, default=0.0)
+
 RESUMABLE_STREAM_TTL_SECONDS = env(
     "RESUMABLE_STREAM_TTL_SECONDS",
     cast=int,
@@ -579,6 +590,8 @@ __all__ = [
     "BG_JOB_SHUTDOWN_GRACE_PERIOD_SECS",
     "BG_JOB_TIMEOUT_SECS",
     "CHECKPOINTER_CONFIG",
+    "CHECKPOINT_BATCH_DELAY",
+    "CHECKPOINT_MAX_BATCH_SIZE",
     "CORS_ALLOW_ORIGINS",
     "CORS_CONFIG",
     "CRON_SCHEDULER_SLEEP_TIME",

{langgraph_api-0.8.0 → langgraph_api-0.8.2}/langgraph_api/grpc/ops/threads.py

@@ -64,6 +64,12 @@ if TYPE_CHECKING:
 logger = structlog.stdlib.get_logger(__name__)
 
 
+# Read-mask preset for Threads.get callers that need metadata/config but not
+# the large "values" column. Used by state get/update/bulk/history which
+# only need graph_id (from metadata) and the thread config.
+THREAD_METADATA_READ_MASK_PATHS: list[str] = ["metadata", "config"]
+
+
 THREAD_STATUS_TO_PB = {
     "idle": enum_thread_status.idle,
     "busy": enum_thread_status.busy,
@@ -471,6 +477,7 @@ class Threads(Authenticated):
         ctx: Auth.types.BaseAuthContext | None = None,
         filters: Auth.types.FilterType | list[pb.AuthFilter] | None = None,
         include_ttl: bool = False,
+        read_mask_paths: list[str] | None = None,
     ) -> AsyncIterator[Thread]:
         """Get a thread by ID.
 
@@ -482,6 +489,11 @@ class Threads(Authenticated):
                      Accepts either raw dict filters (FilterType) or pre-processed
                      proto filters (list[pb.AuthFilter]).
             include_ttl: When True, include TTL information in the response.
+            read_mask_paths: Optional list of field names to fetch from the DB.
+                None (default) fetches all columns. When set, only those columns
+                (plus always-included identity/status columns) are queried.
+                Use ["metadata", "config"] to skip the large "values" column, or
+                [] for an auth/existence check that discards the thread body.
         """
         auth_filters = await Threads.handle_event(
             ctx, "read", {"thread_id": str(thread_id)}
@@ -496,10 +508,16 @@ class Threads(Authenticated):
                 # Raw dict format, convert to proto
                 auth_filters = (auth_filters or []) + _filters_to_proto(filters)
 
+        read_mask = (
+            field_mask_pb2.FieldMask(paths=read_mask_paths)
+            if read_mask_paths is not None
+            else None
+        )
         request = pb.GetThreadRequest(
             thread_id=pb.UUID(value=_normalize_uuid(thread_id)),
             filters=auth_filters,
             include_ttl=include_ttl if include_ttl else None,
+            read_mask=read_mask,
         )
         client = await get_shared_client()
         response = await client.threads.Get(request)
@@ -974,7 +992,12 @@ class Threads(Authenticated):
 
             async with conn.pipeline():
                 thread, checkpoint_iter, graph_id = await asyncio.gather(
-                    Threads.get(conn, thread_id, ctx=ctx),
+                    Threads.get(
+                        conn,
+                        thread_id,
+                        ctx=ctx,
+                        read_mask_paths=THREAD_METADATA_READ_MASK_PATHS,
+                    ),
                     checkpointer.aget_iter(config),
                     Threads.get_graph_id(thread_id),
                 )
@@ -1037,7 +1060,13 @@ class Threads(Authenticated):
 
             async with conn.pipeline():
                 thread, checkpoint_iter, graph_id, run_count = await asyncio.gather(
-                    Threads.get(conn, thread_id, ctx=ctx, filters=filters),
+                    Threads.get(
+                        conn,
+                        thread_id,
+                        ctx=ctx,
+                        filters=filters,
+                        read_mask_paths=THREAD_METADATA_READ_MASK_PATHS,
+                    ),
                     checkpointer.aget_iter(config),
                     Threads.get_graph_id(thread_id),
                     Runs.count(thread_id=thread_id, statuses=["pending", "running"]),
@@ -1130,7 +1159,13 @@ class Threads(Authenticated):
 
             async with conn.pipeline():
                 thread, graph_id = await asyncio.gather(
-                    Threads.get(conn, thread_id, ctx=ctx, filters=filters),
+                    Threads.get(
+                        conn,
+                        thread_id,
+                        ctx=ctx,
+                        filters=filters,
+                        read_mask_paths=THREAD_METADATA_READ_MASK_PATHS,
+                    ),
                     Threads.get_graph_id(config["configurable"]["thread_id"]),
                 )
             thread = await fetchone(thread)
@@ -1218,7 +1253,12 @@ class Threads(Authenticated):
             """Get the history of a thread."""
             async with conn.pipeline():
                 thread, graph_id = await asyncio.gather(
-                    Threads.get(conn, config["configurable"]["thread_id"], ctx=ctx),
+                    Threads.get(
+                        conn,
+                        config["configurable"]["thread_id"],
+                        ctx=ctx,
+                        read_mask_paths=THREAD_METADATA_READ_MASK_PATHS,
+                    ),
                     Threads.get_graph_id(config["configurable"]["thread_id"]),
                 )
             thread = await fetchone(thread)

{langgraph_api-0.8.0 → langgraph_api-0.8.2}/langgraph_api/js/package.json

@@ -25,7 +25,7 @@
     "tsx": "^4.21.0",
     "typescript": "^5.9.3",
     "undici": "^7.24.7",
-    "uuid": "^13.0.0",
+    "uuid": "^14.0.0",
     "vite": "^7.3.2",
     "winston": "^3.19.0",
     "zod": "^4.3.6"

{langgraph_api-0.8.0 → langgraph_api-0.8.2}/langgraph_api/js/yarn.lock

@@ -1655,19 +1655,10 @@ postcss-value-parser@^4.1.0, postcss-value-parser@^4.2.0:
   resolved "https://registry.yarnpkg.com/postcss-value-parser/-/postcss-value-parser-4.2.0.tgz#723c09920836ba6d3e5af019f92bc0971c02e514"
   integrity sha512-1NNCs6uurfkVbeXG4S8JFT9t19m45ICnif8zWLd5oPSZ50QnwMfK+H3jv408d4jw/7Bttv5axS5IiHoLaVNHeQ==
 
-postcss@^8.4.41, postcss@^8.5.1, postcss@^8.5.3:
-  version "8.5.6"
-  resolved "https://registry.yarnpkg.com/postcss/-/postcss-8.5.6.tgz#2825006615a619b4f62a9e7426cc120b349a8f3c"
-  integrity sha512-3Ybi1tAuwAP9s0r1UQ2J4n5Y0G05bJkpUIO0/bI9MhwmD70S5aTWbXGBwxHrelT+XM1k6dM0pk+SwNkpTRN7Pg==
-  dependencies:
-    nanoid "^3.3.11"
-    picocolors "^1.1.1"
-    source-map-js "^1.2.1"
-
-postcss@^8.5.6:
-  version "8.5.9"
-  resolved "https://registry.yarnpkg.com/postcss/-/postcss-8.5.9.tgz#f6ee9e0b94f0f19c97d2f172bfbd7fc71fe1cca4"
-  integrity sha512-7a70Nsot+EMX9fFU3064K/kdHWZqGVY+BADLyXc8Dfv+mTLLVl6JzJpPaCZ2kQL9gIJvKXSLMHhqdRRjwQeFtw==
+postcss@^8.4.41, postcss@^8.5.1, postcss@^8.5.3, postcss@^8.5.6:
+  version "8.5.10"
+  resolved "https://registry.yarnpkg.com/postcss/-/postcss-8.5.10.tgz#8992d8c30acf3f12169e7c09514a12fed7e48356"
+  integrity sha512-pMMHxBOZKFU6HgAZ4eyGnwXF/EvPGGqUr0MnZ5+99485wwW41kW91A4LOGxSHhgugZmSChL5AlElNdwlNgcnLQ==
   dependencies:
     nanoid "^3.3.11"
     picocolors "^1.1.1"
@@ -1950,6 +1941,11 @@ uuid@^13.0.0:
   resolved "https://registry.yarnpkg.com/uuid/-/uuid-13.0.0.tgz#263dc341b19b4d755eb8fe36b78d95a6b65707e8"
   integrity sha512-XQegIaBTVUjSHliKqcnFqYypAd4S+WCYt5NIeRs6w/UAry7z8Y9j5ZwRRL4kzq9U3sD6v+85er9FvkEaBpji2w==
 
+uuid@^14.0.0:
+  version "14.0.0"
+  resolved "https://registry.yarnpkg.com/uuid/-/uuid-14.0.0.tgz#0af883220163d264ffe0c084f6b8a89b9666966d"
+  integrity sha512-Qo+uWgilfSmAhXCMav1uYFynlQO7fMFiMVZsQqZRMIXp0O7rR7qjkj+cPvBHLgBqi960QCoo/PH2/6ZtVqKvrg==
+
 "vite@^6.0.0 || ^7.0.0 || ^8.0.0", vite@^6.4.2:
   version "6.4.2"
   resolved "https://registry.yarnpkg.com/vite/-/vite-6.4.2.tgz#a4e548ca3a90ca9f3724582cab35e1ba15efc6f2"

{langgraph_api-0.8.0 → langgraph_api-0.8.2}/langgraph_api/queue_entrypoint.py

@@ -34,11 +34,15 @@ shutdown_reason: str | None = None
 
 def _ensure_port_available(host: str, port: int) -> None:
     host = normalize_host(host)
+    # Pin AF_INET6 for IPv6 literals: glibc's getaddrinfo can return EAI_ADDRFAMILY
+    # on IPv6-only hosts when the family is unspecified.
+    family = socket.AF_INET6 if ":" in host else socket.AF_UNSPEC
     last_error: OSError | None = None
     try:
         addrinfos = socket.getaddrinfo(
             host,
             port,
+            family=family,
             type=socket.SOCK_STREAM,
             flags=socket.AI_PASSIVE,
         )

{langgraph_api-0.8.0 → langgraph_api-0.8.2}/langgraph_api/worker.py

@@ -15,7 +15,6 @@ from typing_extensions import TypedDict
 import langgraph_api.logging as lg_logging
 from langgraph_api.auth.custom import SimpleUser, normalize_user
 from langgraph_api.config import (
-    BG_JOB_ISOLATED_LOOPS,
     BG_JOB_MAX_RETRIES,
     BG_JOB_TIMEOUT_SECS,
     USE_CUSTOM_CHECKPOINTER,
@@ -238,17 +237,11 @@ async def worker(
 
                 error_message = (
                     f"Run {run['run_id']} exceeded max attempts ({BG_JOB_MAX_RETRIES}).\n\n"
-                    "This may happen if your code blocks the event loop with synchronous I/O bound calls (network requests, database queries, etc.).\n\n"
-                    "If that is the case, your issues may be resolved by converting synchronous operations to async (e.g., use aiohttp instead of requests).\n\n"
+                    "This usually means the pod processing the run became unhealthy and the run was re-queued repeatedly. "
+                    "Check for OOM kills, pod restarts, liveness/readiness probe failures, or other signs of pod instability "
+                    "in your deployment's logs and metrics.\n\n"
                 )
 
-                if not BG_JOB_ISOLATED_LOOPS:
-                    error_message += (
-                        "Also consider setting BG_JOB_ISOLATED_LOOPS=true in your environment. This will isolate I/O-bound operations to avoid"
-                        " blocking the main API server.\n\n"
-                        "See: https://langchain-ai.github.io/langgraph/cloud/reference/env_var/#bg_job_isolated_loops\n\n"
-                    )
-
                 raise RuntimeError(error_message)
             configurable = run["kwargs"].get("config", {}).get("configurable", {})
             async with set_auth_ctx_for_run(run["kwargs"]):