deer-flow/backend/tests/test_checkpoint_state.py
Vanzeren 42baed8c8c
feat(checkpoint): dual-mode checkpoint storage with LangGraph DeltaChannel (#4292)
* feat(checkpoint): dual-mode checkpoint storage with LangGraph DeltaChannel

Add a restart-required database.checkpoint_channel_mode ("full" default,
"delta") that stores the messages channel via LangGraph 1.2 DeltaChannel,
cutting checkpoint storage from O(n^2) to O(n) for append-only history.
Existing full checkpoints seed delta state transparently; no data migration.

- config: mode schema + freeze-on-first-use with
  CheckpointModeReconfigurationError; mode marker persisted in checkpoint
  metadata; unsafe delta->full downgrade rejected fail-closed with
  CheckpointModeMismatchError (run-level error, failed state read)
- state: delta message state schema; CheckpointStateAccessor centralizes
  materialized reads for all consumers (threads API, branches,
  regeneration, compaction, state updates, memory, goal workers)
- runtime: raw writers (run durations, interrupted title, thread goal)
  parent their checkpoints to the checkpoint they derive from, preserving
  delta ancestry; rollback forks the pre-run lineage through a state
  mutation graph with Overwrite restores; InMemorySaver delta-history
  override delegates to the base walk (fixes dropped first write after
  migration, also present upstream)
- tests: conformance suite over {memory, sqlite, postgres} covering
  migration replay, stable message IDs, storage shape and writer
  preservation; conftest fixture isolates the frozen mode between tests;
  stale config fakes refreshed
- ci: backend unit tests gain a postgres service

* fix(checkpoint): close materialization gaps in goal flow, guard public factory

- Route goal-continuation message reads through CheckpointStateAccessor:
  raw channel_values reads see the delta sentinel in delta mode, which
  disabled goal continuation (stand_down=no_durable_end_of_turn) after
  durable assistant turns. Raw tuples remain for tuple-only metadata
  (checkpoint id, pending_writes).
- Reject checkpoint_channel_mode='delta' + checkpointer in
  create_deerflow_agent at construction: factory-built persisted graphs
  bypass mode-marker injection and the fail-closed gate, reproducing
  silent mixed-mode state loss. Delta without persistence stays allowed.
- Import the postgres saver lazily (pytest.importorskip in the fixture)
  so the documented default install collects the suite; add a CI job
  running pytest --collect-only on uv sync --group dev without extras.
- Fix test_checkpointer fallback test to patch get_app_config at its
  use site (provider module), making it deterministic when a local
  config.yaml selects a persistent backend.

* fix(gateway): preserve extension-owned channels in state mutations, bump config version

- build_state_mutation_graph / build_checkpoint_state_mutation_accessor
  accept an explicit state_schema; branch and POST /state now compile the
  mutation graph from the thread's effective schema
  (graph_state_schema on the assistant graph). The base-ThreadState
  fallback silently discarded channels contributed by custom
  AgentMiddleware.state_schema on branch (data loss) and returned a
  false-success 200 on POST /state.
- POST /state validates values keys against the mutation graph's
  channels and rejects unknown fields with 422 instead of ignoring
  them; reducer detection covers extension channels
  (BinaryOperatorAggregate or DeltaChannel) so Overwrite replace
  semantics work for middleware reducers in both modes.
- Endpoint regression: custom AgentMiddleware.state_schema value
  survives branch, updates through POST /state, and an unknown field
  receives 422.
- config_version 26 -> 27 for the new database.checkpoint_channel_mode
  (example, Helm chart values + README, support-bundle fixture), so
  existing installs get the outdated-config warning and
  make config-upgrade merges the field; covered by a test driving the
  real example file and the real config-upgrade script.

* fix(gateway): resolve assistant schema via one boundary, copy branch reducer values with Overwrite

GET /threads/{id}/state now resolves the thread's assistant_id through a
single reusable boundary (thread metadata -> assistant_id -> effective
graph), so channels contributed by a custom AgentMiddleware.state_schema
are materialized instead of dropped by the default lead schema. POST
/state uses the same boundary instead of resolving the schema ad hoc.

Branch writes wrap every copied reducer channel in Overwrite (derived
from the effective mutation graph: BinaryOperatorAggregate + DeltaChannel),
not just messages, so already-aggregated values are never re-merged.

Regression tests use a real AgentMiddleware.state_schema with a
non-identity reducer in both full and delta modes: GET /state returns the
extension value, POST /state replaces it, branch preserves it
byte-for-byte; the unknown-field 422 is a separate assertion.

* refactor(checkpoint): collapse read-path round-trips and ship dual-mode parity tests

Address review round 4 on PR #4292:

- Push ahistory/history limit through Pregel into checkpointer.alist
  (SQL LIMIT) instead of materializing all rows and breaking in Python
- Fold the read-side mode-compat gate onto the returned snapshot's
  metadata; only writes keep the pre-write tuple fetch (fail-closed)
- Cache factory-built accessor graphs per (assistant_id, mode) with
  factory-identity revalidation; state reads no longer build a lead
  agent per request
- get_thread: one snapshot fetch + one raw pending_writes fetch on the
  resolved checkpoint (post-checkpoint __error__ writes never surface
  in snapshot.tasks; verified empirically)
- DeerFlowClient.get_thread: single checkpointer.list walk collects
  pending_writes per checkpoint instead of N get_tuple calls
- InMemorySaver delta-history patch: stand-down when the upstream
  override disappears, try/except guard, validated-version warning,
  guard tests
- make_lead_agent mode precedence: first freeze is owned by app_config
  (client-supplied configurable key ignored); once frozen, injected
  key/app_config must match or fail closed
- Rollback: lock in non-message channel restoration via fork
  inheritance with a dedicated reducer-channel test
- Add tests/test_threads_checkpoint_mode.py and
  tests/test_gateway_checkpoint_mode.py referenced by AGENTS.md and
  the PR validation section: lifecycle parity (memory + sqlite),
  per-step blob-count storage guard, gateway endpoint parity

Counted-saver tests pin checkpoint round-trips for aget/ahistory so
these regressions cannot silently return.

* fix(checkpoint): precise mode-mismatch HTTP mapping, gate E2E, and accessor resilience

- threads router: map CheckpointModeMismatchError to 409 (with cause and
  thread id) and CheckpointModeReconfigurationError to 503 across all state
  endpoints instead of swallowing both into a generic 500
- gate coverage: seed a real delta checkpoint into AsyncSqliteSaver and
  assert aget/aupdate/ahistory fail closed; assert 409 at the HTTP boundary
  through the real route stack
- rollback: compile the restore mutation graph with the thread's effective
  state schema per the build_state_mutation_graph contract
- inheritance contract locks: rollback and manual compaction preserve
  middleware-contributed channels via checkpoint fork cloning
- services: revalidate the accessor-graph cache against app_config identity
  so config.yaml hot-reloads never serve a stale compiled graph
- services: degrade full-mode state reads to raw checkpointer reads when the
  agent factory is unavailable (delta gate still applies; delta mode has no
  fallback)
- deps: override websockets==16.0 (langgraph-sdk 0.4.2's <16 pin silently
  downgraded 16.0 -> 15.0.1; pin is not grounded in any API incompatibility)
  and bump the langchain lower bound to what the lockfile actually resolves

* fix(checkpoint): include anchor checkpoint in degraded history walk + cover get_thread

- _RawCheckpointReadAccessor.ahistory: alist(before=...) is exclusive while
  pregel's get_state_history treats config.checkpoint_id as the inclusive
  start; fetch the anchor explicitly so both read paths paginate identically
- extend the degraded-path gateway test: GET /thread returns raw values, and
  POST /history with before starts at the anchor checkpoint

* fix(gateway): preserve degraded checkpoint timestamps

* fix(gateway): harden degraded checkpoint access

* fix(gateway): resolve assistants for checkpoint reads

---------

Co-authored-by: Willem Jiang <willem.jiang@gmail.com>
2026-07-22 08:33:29 +08:00

307 lines
12 KiB
Python

from __future__ import annotations
from copy import deepcopy
from types import SimpleNamespace
from typing import Annotated, Any, TypedDict
import pytest
from langchain_core.messages import AnyMessage
from langgraph.checkpoint.memory import InMemorySaver
from langgraph.graph.message import add_messages
from deerflow.runtime import CheckpointStateAccessor
from deerflow.runtime.checkpoint_mode import CHECKPOINT_MODE_METADATA_KEY, INTERNAL_CHECKPOINT_MODE_KEY
class FakeCheckpointer:
def __init__(self) -> None:
self.sync_configs: list[dict[str, Any]] = []
self.async_configs: list[dict[str, Any]] = []
def get_tuple(self, config: dict[str, Any]) -> None:
self.sync_configs.append(config)
return None
async def aget_tuple(self, config: dict[str, Any]) -> None:
self.async_configs.append(config)
return None
class FakeGraph:
def __init__(self) -> None:
self.checkpointer: Any = None
self.store: Any = None
self.calls: list[tuple[Any, ...]] = []
self.sync_history_yields = 0
self.async_history_yields = 0
def get_state(self, config: dict[str, Any]) -> SimpleNamespace:
self.calls.append(("get", config))
return SimpleNamespace(values={"messages": ["sync"]})
def get_state_history(self, config: dict[str, Any], *, limit: int | None = None):
self.calls.append(("history", config, limit))
for index in range(4):
if limit is not None and self.sync_history_yields >= limit:
return
self.sync_history_yields += 1
yield SimpleNamespace(values={"index": index})
def update_state(self, config: dict[str, Any], values: dict[str, Any], *, as_node: str | None = None) -> dict[str, Any]:
self.calls.append(("update", config, values, as_node))
return {"updated": values, "as_node": as_node}
async def aget_state(self, config: dict[str, Any]) -> SimpleNamespace:
self.calls.append(("aget", config))
return SimpleNamespace(values={"messages": ["async"]})
async def aget_state_history(self, config: dict[str, Any], *, limit: int | None = None):
self.calls.append(("ahistory", config, limit))
for index in range(4):
if limit is not None and self.async_history_yields >= limit:
return
self.async_history_yields += 1
yield SimpleNamespace(values={"index": index})
async def aupdate_state(
self,
config: dict[str, Any],
values: dict[str, Any],
*,
as_node: str | None = None,
) -> dict[str, Any]:
self.calls.append(("aupdate", config, values, as_node))
return {"updated": values, "as_node": as_node}
def _assert_delta_config_is_copied(original: dict[str, Any], forwarded: dict[str, Any]) -> None:
assert forwarded is not original
assert forwarded["configurable"] is not original["configurable"]
assert forwarded["metadata"] is not original["metadata"]
assert forwarded["configurable"][INTERNAL_CHECKPOINT_MODE_KEY] == "delta"
assert forwarded["metadata"][CHECKPOINT_MODE_METADATA_KEY] == "delta"
def test_sync_accessor_binds_persistence_guards_operations_and_preserves_input() -> None:
graph = FakeGraph()
saver = FakeCheckpointer()
store = object()
accessor = CheckpointStateAccessor.bind(graph, saver, store=store, mode="delta")
config = {
"configurable": {"thread_id": "thread-sync", "checkpoint_ns": ""},
"metadata": {"caller": "test"},
"tags": ["preserved"],
}
original = deepcopy(config)
snapshot = accessor.get(config)
history = accessor.history(config, limit=2)
update = accessor.update(config, {"messages": ["changed"]}, as_node="tools")
assert snapshot.values == {"messages": ["sync"]}
assert [item.values for item in history] == [{"index": 0}, {"index": 1}]
assert graph.sync_history_yields == 2
assert update == {"updated": {"messages": ["changed"]}, "as_node": "tools"}
assert graph.checkpointer is saver
assert graph.store is store
assert config == original
for call in graph.calls:
_assert_delta_config_is_copied(config, call[1])
assert graph.calls[-1][2:] == ({"messages": ["changed"]}, "tools")
@pytest.mark.anyio
async def test_async_accessor_binds_persistence_guards_operations_and_preserves_input() -> None:
graph = FakeGraph()
saver = FakeCheckpointer()
store = object()
accessor = CheckpointStateAccessor.bind(graph, saver, store=store, mode="delta")
config = {
"configurable": {"thread_id": "thread-async", "checkpoint_ns": ""},
"metadata": {"caller": "test"},
"tags": ["preserved"],
}
original = deepcopy(config)
snapshot = await accessor.aget(config)
history = await accessor.ahistory(config, limit=2)
update = await accessor.aupdate(config, {"messages": ["changed"]}, as_node="agent")
assert snapshot.values == {"messages": ["async"]}
assert [item.values for item in history] == [{"index": 0}, {"index": 1}]
assert graph.async_history_yields == 2
assert update == {"updated": {"messages": ["changed"]}, "as_node": "agent"}
assert graph.checkpointer is saver
assert graph.store is store
assert config == original
for call in graph.calls:
_assert_delta_config_is_copied(config, call[1])
assert graph.calls[-1][2:] == ({"messages": ["changed"]}, "agent")
def test_sync_history_zero_limit_guards_without_consuming_a_snapshot() -> None:
graph = FakeGraph()
saver = FakeCheckpointer()
accessor = CheckpointStateAccessor.bind(graph, saver)
config = {"configurable": {"thread_id": "thread-sync-zero"}}
assert accessor.history(config, limit=0) == []
# The read-side gate folds onto the returned snapshot: no standalone tuple fetch.
assert len(saver.sync_configs) == 0
assert graph.sync_history_yields == 0
@pytest.mark.anyio
async def test_async_history_zero_limit_guards_without_consuming_a_snapshot() -> None:
graph = FakeGraph()
saver = FakeCheckpointer()
accessor = CheckpointStateAccessor.bind(graph, saver)
config = {"configurable": {"thread_id": "thread-async-zero"}}
assert await accessor.ahistory(config, limit=0) == []
assert len(saver.async_configs) == 0
assert graph.async_history_yields == 0
@pytest.mark.anyio
async def test_full_accessor_gates_writes_and_checks_reads_on_the_returned_snapshot() -> None:
"""Full mode: only writes pay the pre-write tuple fetch; reads check the
marker on the materialized snapshot's metadata instead."""
graph = FakeGraph()
saver = FakeCheckpointer()
accessor = CheckpointStateAccessor.bind(graph, saver)
config = {"configurable": {"thread_id": "thread-full"}}
accessor.get(config)
accessor.history(config, limit=1)
accessor.update(config, {}, as_node=None)
await accessor.aget(config)
await accessor.ahistory(config, limit=1)
await accessor.aupdate(config, {}, as_node=None)
assert len(saver.sync_configs) == 1
assert len(saver.async_configs) == 1
for prepared in [*saver.sync_configs, *saver.async_configs]:
assert prepared["configurable"][INTERNAL_CHECKPOINT_MODE_KEY] == "full"
assert CHECKPOINT_MODE_METADATA_KEY not in prepared["metadata"]
assert config == {"configurable": {"thread_id": "thread-full"}}
@pytest.mark.anyio
async def test_full_accessor_raises_when_the_returned_snapshot_is_delta_marked() -> None:
"""A full-mode accessor must fail closed on a delta checkpoint, detected
via the returned snapshot metadata (no pre-read tuple fetch)."""
from deerflow.runtime.checkpoint_mode import CheckpointModeMismatchError
graph = FakeGraph()
saver = FakeCheckpointer()
accessor = CheckpointStateAccessor.bind(graph, saver, mode="full")
config = {"configurable": {"thread_id": "thread-delta-marked"}}
graph.get_state = lambda _config: SimpleNamespace(values={}, metadata={CHECKPOINT_MODE_METADATA_KEY: "delta"})
with pytest.raises(CheckpointModeMismatchError, match="requires delta mode"):
accessor.get(config)
assert len(saver.sync_configs) == 0
async def _delta_history(_config, *, limit=None):
yield SimpleNamespace(values={"index": 0}, metadata={})
yield SimpleNamespace(values={"index": 1}, metadata={CHECKPOINT_MODE_METADATA_KEY: "delta"})
graph.aget_state_history = _delta_history
with pytest.raises(CheckpointModeMismatchError, match="requires delta mode"):
await accessor.ahistory(config)
assert len(saver.async_configs) == 0
@pytest.mark.anyio
async def test_full_accessor_writes_still_check_compatibility_before_writing() -> None:
"""Writes cannot be un-applied, so the pre-write tuple fetch stays."""
from deerflow.runtime.checkpoint_mode import CheckpointModeMismatchError
graph = FakeGraph()
class DeltaMarkedSaver(FakeCheckpointer):
async def aget_tuple(self, config):
self.async_configs.append(config)
return SimpleNamespace(metadata={CHECKPOINT_MODE_METADATA_KEY: "delta"})
saver = DeltaMarkedSaver()
accessor = CheckpointStateAccessor.bind(graph, saver, mode="full")
config = {"configurable": {"thread_id": "thread-delta-write"}}
with pytest.raises(CheckpointModeMismatchError, match="requires delta mode"):
await accessor.aupdate(config, {"messages": []}, as_node=None)
assert len(saver.async_configs) == 1
assert graph.calls == []
class _CountingSaver(InMemorySaver):
"""InMemorySaver that counts checkpoint round-trips."""
def __init__(self) -> None:
super().__init__()
self.aget_tuple_calls = 0
self.alist_limits: list[int | None] = []
async def aget_tuple(self, config):
self.aget_tuple_calls += 1
return await super().aget_tuple(config)
async def alist(self, config, *, filter=None, before=None, limit=None):
self.alist_limits.append(limit)
async for item in super().alist(config, filter=filter, before=before, limit=limit):
yield item
def _build_counting_graph(saver):
from langchain_core.messages import HumanMessage
from langgraph.graph import StateGraph
class _State(TypedDict):
messages: Annotated[list[AnyMessage], add_messages]
async def _append(state):
return {"messages": [HumanMessage(content=f"turn-{len(state.get('messages') or [])}")]}
builder = StateGraph(_State)
builder.add_node("append", _append)
builder.set_entry_point("append")
builder.set_finish_point("append")
return builder.compile(checkpointer=saver)
@pytest.mark.anyio
async def test_ahistory_pushes_limit_into_alist_and_reads_each_snapshot_once() -> None:
"""The history limit must reach ``checkpointer.alist`` (SQL LIMIT), and the
read-side compat gate must not add a standalone tuple fetch per call."""
saver = _CountingSaver()
graph = _build_counting_graph(saver)
accessor = CheckpointStateAccessor.bind(graph, saver, mode="full")
config = {"configurable": {"thread_id": "thread-counted"}}
for _ in range(4):
await graph.ainvoke({}, config)
saver.aget_tuple_calls = 0
history = await accessor.ahistory(config, limit=2)
assert len(history) == 2
assert saver.alist_limits[-1] == 2
# get_state_history walks via alist only; no aget_tuple in the read path.
assert saver.aget_tuple_calls == 0
@pytest.mark.anyio
async def test_aget_fetches_the_checkpoint_exactly_once_in_full_mode() -> None:
"""Full-mode reads: one fetch inside aget_state; the folded gate adds none."""
saver = _CountingSaver()
graph = _build_counting_graph(saver)
accessor = CheckpointStateAccessor.bind(graph, saver, mode="full")
config = {"configurable": {"thread_id": "thread-counted-get"}}
await graph.ainvoke({}, config)
saver.aget_tuple_calls = 0
snapshot = await accessor.aget(config)
assert [message.content for message in snapshot.values["messages"]] == ["turn-0"]
assert saver.aget_tuple_calls == 1