fix: nested subgraph inherits parent checkpoint_ns (regression in 1.2.3) (#8053)

Closes #8038

## Description

The `ensure_config` merge introduced in #7926 caused a child graph
invoked inside a parent node to inherit the parent task's
`checkpoint_ns` from the ambient run context
(`var_child_runnable_config`), so the child's checkpoints were written
under an unreadable namespace and re-ran from scratch each turn (#8038).
The first explicitly passed `configurable` that carries a checkpoint
coordinate (a `thread_id`, or any
`checkpoint_ns`/`checkpoint_id`/`checkpoint_map`) now replaces the
ambient one, while subsequent explicit configs still shallow-merge —
preserving #7926's `with_config(...)` semantics.

## Contract

`ensure_config` merges an explicit `configurable` over the ambient run
context (`var_child_runnable_config`) with one rule: **an explicit
`configurable` that supplies its own checkpoint coordinate addresses its
own checkpoint lineage, so the ambient `configurable` is dropped rather
than merged over.** Coordinate keys are `thread_id`, `checkpoint_ns`,
`checkpoint_id`, and `checkpoint_map` (grouped as
`_CHECKPOINT_COORDINATE_KEYS`). A non-coordinate `configurable` keeps
the ambient and shallow-merges over it.

Below, each case shows the parent/child graph wiring that triggers it
and the resulting namespacing. The child is always a compiled subgraph
invoked from inside a parent node.

### 1. Subgraph invoked with no new config → ambient inherited

```python
child = StateGraph(ChildState).add_node("n", child_node).add_edge(START, "n").compile(checkpointer=checkpointer)

def parent_node(state):
    child.invoke({}, config=None)          # no explicit configurable
    return {"result": "ok"}

parent = StateGraph(ParentState).add_node("p", parent_node).add_edge(START, "p").compile(checkpointer=checkpointer)
parent.invoke({"result": ""}, config={"configurable": {"thread_id": "parent"}})
```
Child inherits the parent task's `checkpoint_ns` (`p:<parent-task>`);
its checkpoints are written as a discoverable child of the parent run.
Pre-#7926 behavior, unchanged.

### 2. Subgraph invoked with a new thread_id → ambient dropped

```python
child = StateGraph(ChildState).add_node("n", child_node).add_edge(START, "n").compile(checkpointer=checkpointer)
child_config = {"configurable": {"thread_id": str(uuid4())}}

def parent_node(state):
    child.invoke({}, config=child_config)  # explicit new thread_id
    return {"result": "ok"}

parent = StateGraph(ParentState).add_node("p", parent_node).add_edge(START, "p").compile(checkpointer=checkpointer)
parent.invoke({"result": ""}, config={"configurable": {"thread_id": "parent"}})
```
Child starts its own lineage on `child_config`'s thread; `checkpoint_ns
== ""`; `child.get_state(child_config)` reads back state across repeated
parent turns. Fixes #8038.

### 3. Subgraph invoked with the same thread_id as parent → ambient
still dropped

```python
child = StateGraph(ChildState).add_node("n", child_node).add_edge(START, "n").compile(checkpointer=checkpointer)

def parent_node(state):
    # reuses the parent's thread_id as the child's own
    child.invoke({}, config={"configurable": {"thread_id": state["parent_thread"]}})
    return {"result": "ok"}

parent = StateGraph(ParentState).add_node("p", parent_node).add_edge(START, "p").compile(checkpointer=checkpointer)
parent_thread = str(uuid4())
parent.invoke({"parent_thread": parent_thread, "result": ""}, config={"configurable": {"thread_id": parent_thread}})
```
Child addresses its own root namespace on the shared thread;
`checkpoint_ns == ""`. The parent task's `checkpoint_ns` must not leak
in, or `child.get_state({"configurable": {"thread_id": parent_thread}})`
returns empty state and the child re-runs from scratch each turn.

### 4. Subagent invoked with a non-coordinate key only → ambient
inherited

```python
child = StateGraph(ChildState).add_node("n", child_node).add_edge(START, "n").compile(checkpointer=checkpointer)

def parent_node(state):
    # ls_agent_type is not a checkpoint coordinate, so ambient is kept
    child.invoke({}, config={"configurable": {"ls_agent_type": "subagent"}})
    return {"result": "ok"}

parent = StateGraph(ParentState).add_node("p", parent_node).add_edge(START, "p").compile(checkpointer=checkpointer)
parent.invoke({"result": ""}, config={"configurable": {"thread_id": "parent"}})
```
Child remains a discoverable child of the parent run; ambient
`thread_id` and `checkpoint_ns` preserved (deepagents `task` tool
pattern).

### 5. with_config(...) + invoke-time thread_id → ambient dropped, then
merged

```python
# child compiled with a non-coordinate configurable via with_config
child = (
    StateGraph(ChildState)
    .add_node("n", child_node)
    .add_edge(START, "n")
    .compile(checkpointer=checkpointer)
    .with_config({"configurable": {"ls_agent_type": "root"}})
)

def parent_node(state):
    # invoke-time config supplies the thread_id; with_config's ls_agent_type survives
    child.invoke({}, config={"configurable": {"thread_id": "child"}})
    return {"result": "ok"}

parent = StateGraph(ParentState).add_node("p", parent_node).add_edge(START, "p").compile(checkpointer=checkpointer)
parent.invoke({"result": ""}, config={"configurable": {"thread_id": "parent"}})
```
First coordinate-bearing config (`thread_id`) drops the ambient;
subsequent explicit configs still shallow-merge, so `ls_agent_type`
survives alongside `thread_id`. Preserves #7926 `with_config(...)`
semantics.

### Regression note

Cases 1, 4, and 5 are the pre-#7926 behavior and are preserved
unchanged. Cases 2 and 3 fix the regression introduced by #7926: an
explicit `thread_id` resets the ambient even when it equals the ambient
thread id, because a child reusing the parent's thread id still
addresses its own root namespace on that thread, not the parent task's.

## Self-Hosted Release Note
Fix regression where a nested subgraph with its own `thread_id` invoked
inside a parent node lost its persisted state across turns.

## Test Plan
- [x] `pytest tests/test_subgraph_persistence.py -k
test_child_with_own_thread_id_keeps_namespace` (case 2)
- [x] `pytest tests/test_utils.py -k
ensure_config_explicit_configurable_replaces_ambient` (case 2)
- [x] `pytest tests/test_utils.py -k
ensure_config_ambient_inherited_when_no_explicit_configurable` (case 1)
- [x] `pytest tests/test_utils.py -k
ensure_config_non_coordinate_config_keeps_ambient_checkpoint_ns` (case
4)
- [x] `pytest tests/test_utils.py -k
ensure_config_explicit_configurables_still_merge_over_ambient` (case 5)
- [x] `pytest tests/test_utils.py -k
ensure_config_same_thread_id_still_clears_ambient` (case 3)

Made by [Open SWE](https://openswe.vercel.app)

---------

Co-authored-by: Nick Hollon <274035459+nick-hollon-lc@users.noreply.github.com>
Co-authored-by: open-swe[bot] <open-swe@users.noreply.github.com>
This commit is contained in:
Nick Hollon
2026-06-17 10:29:31 -04:00
committed by GitHub
co-authored by Nick Hollon open-swe[bot] <open-swe@users.noreply.github.com>
parent 9af25217c3
commit 79befe67ba
5 changed files with 214 additions and 0 deletions
@@ -20,6 +20,7 @@ from langchain_core.runnables.config import (
from langgraph.checkpoint.base import CheckpointMetadata
from langgraph._internal._constants import (
_CHECKPOINT_COORDINATE_KEYS,
CONF,
CONFIG_KEY_CHECKPOINT_ID,
CONFIG_KEY_CHECKPOINT_MAP,
@@ -342,6 +343,28 @@ def ensure_config(*configs: RunnableConfig | None) -> RunnableConfig:
if _is_not_empty(v)
},
)
# An explicit config that supplies its own checkpoint coordinate (a
# thread_id, or any checkpoint_ns/checkpoint_id/checkpoint_map) is addressing
# its own checkpoint lineage, so drop the inherited ambient configurable
# rather than merging over it: a child graph invoked inside a parent node
# would otherwise write its checkpoints under the parent's namespace and
# never find them again. An explicit thread_id resets even when it equals the
# ambient one, since a child reusing the parent's thread id still addresses
# its own root namespace, not the parent task's. Configs that only refine
# other keys keep the ambient and shallow-merge over it below.
if empty.get(CONF):
for config in configs:
if config is None:
continue
explicit_configurable = config.get(CONF)
if not explicit_configurable:
continue
if any(
_is_not_empty(explicit_configurable.get(k))
for k in _CHECKPOINT_COORDINATE_KEYS
):
empty[CONF] = {}
break
for config in configs:
if config is None:
continue
@@ -95,6 +95,15 @@ NULL_TASK_ID = sys.intern("00000000-0000-0000-0000-000000000000")
OVERWRITE = sys.intern("__overwrite__")
# dict key for the overwrite value, used as `{'__overwrite__': value}`
# Checkpoint coordinate keys: when any of these appear in an explicit
# configurable, the caller is addressing its own checkpoint lineage.
_CHECKPOINT_COORDINATE_KEYS = (
CONFIG_KEY_THREAD_ID,
CONFIG_KEY_CHECKPOINT_NS,
CONFIG_KEY_CHECKPOINT_ID,
CONFIG_KEY_CHECKPOINT_MAP,
)
# redefined to avoid circular import with langgraph.constants
_TAG_HIDDEN = sys.intern("langsmith:hidden")
@@ -639,3 +639,49 @@ def test_stateful_namespace_isolation(
"broccoli round 2",
"Veggie: broccoli round 2",
]
def test_child_with_own_thread_id_keeps_namespace(
sync_checkpointer: BaseCheckpointSaver,
) -> None:
"""A child graph invoked from inside a parent node with its own thread_id
must store and read its checkpoint under its own namespace, not inherit the
parent task's checkpoint_ns.
"""
class ChildState(TypedDict):
count: int
def child_node(state: ChildState) -> dict:
return {"count": (state.get("count") or 0) + 1}
child = (
StateGraph(ChildState)
.add_node("n", child_node)
.add_edge(START, "n")
.compile(checkpointer=sync_checkpointer)
)
child_thread = str(uuid4())
child_config = {"configurable": {"thread_id": child_thread}}
def parent_node(state: ParentState) -> dict:
child.invoke({}, config=child_config)
return {"result": "ok"}
parent = (
StateGraph(ParentState)
.add_node("p", parent_node)
.add_edge(START, "p")
.compile(checkpointer=sync_checkpointer)
)
parent_config = {"configurable": {"thread_id": str(uuid4())}}
parent.invoke({"result": ""}, config=parent_config)
state1 = child.get_state(child_config)
assert state1.values.get("count") == 1
assert state1.config["configurable"]["checkpoint_ns"] == ""
parent.invoke({"result": ""}, config=parent_config)
state2 = child.get_state(child_config)
assert state2.values.get("count") == 2
@@ -660,3 +660,50 @@ async def test_stateful_namespace_isolation_async(
"broccoli round 2",
"Veggie: broccoli round 2",
]
@NEEDS_CONTEXTVARS
async def test_child_with_own_thread_id_keeps_namespace_async(
async_checkpointer: BaseCheckpointSaver,
) -> None:
"""A child graph invoked from inside a parent node with its own thread_id
must store and read its checkpoint under its own namespace, not inherit the
parent task's checkpoint_ns.
"""
class ChildState(TypedDict):
count: int
def child_node(state: ChildState) -> dict:
return {"count": (state.get("count") or 0) + 1}
child = (
StateGraph(ChildState)
.add_node("n", child_node)
.add_edge(START, "n")
.compile(checkpointer=async_checkpointer)
)
child_thread = str(uuid4())
child_config = {"configurable": {"thread_id": child_thread}}
async def parent_node(state: ParentState) -> dict:
await child.ainvoke({}, config=child_config)
return {"result": "ok"}
parent = (
StateGraph(ParentState)
.add_node("p", parent_node)
.add_edge(START, "p")
.compile(checkpointer=async_checkpointer)
)
parent_config = {"configurable": {"thread_id": str(uuid4())}}
await parent.ainvoke({"result": ""}, config=parent_config)
state1 = await child.aget_state(child_config)
assert state1.values.get("count") == 1
assert state1.config["configurable"]["checkpoint_ns"] == ""
await parent.ainvoke({"result": ""}, config=parent_config)
state2 = await child.aget_state(child_config)
assert state2.values.get("count") == 2
+89
View File
@@ -506,6 +506,95 @@ def test_ensure_config_configurable_later_wins_per_key() -> None:
assert merged["configurable"]["only_b"] == "B"
def test_ensure_config_explicit_configurable_replaces_ambient() -> None:
# An explicit checkpoint coordinate (here a new thread_id) starts a fresh
# lineage and drops the ambient run context (e.g. a parent task's
# checkpoint_ns), so a child graph does not inherit it.
from langchain_core.runnables.config import var_child_runnable_config
token = var_child_runnable_config.set(
{"configurable": {"checkpoint_ns": "p:parent-task", "checkpoint_id": "cid"}}
)
try:
merged = ensure_config({"configurable": {"thread_id": "child"}})
finally:
var_child_runnable_config.reset(token)
assert merged["configurable"]["thread_id"] == "child"
assert "checkpoint_ns" not in merged["configurable"]
assert "checkpoint_id" not in merged["configurable"]
def test_ensure_config_ambient_inherited_when_no_explicit_configurable() -> None:
# With no explicit configurable, the ambient run context is inherited
# unchanged (stateless subgraph / interrupt-resume pattern).
from langchain_core.runnables.config import var_child_runnable_config
token = var_child_runnable_config.set(
{"configurable": {"checkpoint_ns": "p:parent-task"}}
)
try:
merged = ensure_config({"tags": ["t"]})
finally:
var_child_runnable_config.reset(token)
assert merged["configurable"]["checkpoint_ns"] == "p:parent-task"
def test_ensure_config_explicit_configurables_still_merge_over_ambient() -> None:
# A new thread_id drops the ambient, but explicit configs still shallow-merge
# among themselves, so a with_config(...) value (ls_agent_type) survives
# alongside an invoke-time thread_id.
from langchain_core.runnables.config import var_child_runnable_config
token = var_child_runnable_config.set(
{"configurable": {"checkpoint_ns": "p:parent-task"}}
)
try:
merged = ensure_config(
{"configurable": {"ls_agent_type": "root"}},
{"configurable": {"thread_id": "child"}},
)
finally:
var_child_runnable_config.reset(token)
assert merged["configurable"]["ls_agent_type"] == "root"
assert merged["configurable"]["thread_id"] == "child"
assert "checkpoint_ns" not in merged["configurable"]
def test_ensure_config_non_coordinate_config_keeps_ambient_checkpoint_ns() -> None:
# A nested subagent is invoked with a non-coordinate configurable key
# (ls_agent_type) and no thread_id; it must keep the inherited checkpoint_ns
# so it stays a discoverable child of the parent run (deepagents `task` tool).
from langchain_core.runnables.config import var_child_runnable_config
token = var_child_runnable_config.set(
{"configurable": {"thread_id": "parent", "checkpoint_ns": "p:parent-task"}}
)
try:
merged = ensure_config({"configurable": {"ls_agent_type": "subagent"}})
finally:
var_child_runnable_config.reset(token)
assert merged["configurable"]["ls_agent_type"] == "subagent"
assert merged["configurable"]["checkpoint_ns"] == "p:parent-task"
assert merged["configurable"]["thread_id"] == "parent"
def test_ensure_config_same_thread_id_still_clears_ambient() -> None:
# A child that reuses the parent's thread_id is still addressing its own root
# namespace on that thread, so the parent task's checkpoint_ns must not leak
# in; otherwise the child writes state that get_state cannot read back.
from langchain_core.runnables.config import var_child_runnable_config
token = var_child_runnable_config.set(
{"configurable": {"thread_id": "shared", "checkpoint_ns": "p:parent-task"}}
)
try:
merged = ensure_config({"configurable": {"thread_id": "shared"}})
finally:
var_child_runnable_config.reset(token)
assert merged["configurable"]["thread_id"] == "shared"
assert "checkpoint_ns" not in merged["configurable"]
def test_ensure_config_merges_metadata_across_configs() -> None:
a = {"metadata": {"user_id": "U1"}}
b = {"metadata": {"correlation_id": "C1"}}