fix(memory): sign DeerMem reload before reading the document (#6238)

* fix(memory): sign DeerMem reload before reading the document

reload() read the document before computing its cache signature, so a
write committed in between cached the old document under the new
signature and later load() calls served stale memory until the next
write. Compute the signature first, matching load().

* test(memory): cover markdown storage and agent scope in reload race test

* docs(changelog): reference #6238 for the DeerMem reload fix

CHANGELOG.md

@@ -466,6 +466,13 @@ This release closes that milestone with **301 merged pull requests**.
 
 ### Fixed
 
+- **memory:** A DeerMem memory reload no longer pins an older document in the
+  cache. `reload()` read the document before computing its cache signature, so
+  a write committed in between (for example by the background memory updater)
+  cached the old document under the new signature, and every later `load()`
+  returned the outdated memory until the next write. `reload()` now computes
+  the signature first, as `load()` already did, so a racing write forces a
+  re-read instead. ([#6238])
 - **channels:** Discord now runs its channel-connection database work on the
   Gateway event loop. discord.py delivers messages on a private loop in the
   client thread, and the Discord adapter awaited the connection repository there
@@ -7679,3 +7686,4 @@ with **180 merged pull requests** since the first 2.0 milestone tag.
 [#6214]: https://github.com/bytedance/deer-flow/pull/6214
 [#6229]: https://github.com/bytedance/deer-flow/pull/6229
 [#6230]: https://github.com/bytedance/deer-flow/pull/6230
+[#6238]: https://github.com/bytedance/deer-flow/pull/6238

CHANGELOG_zh.md

@@ -423,6 +423,11 @@
 
 ### 修复
 
+- **记忆：** DeerMem 记忆重新加载不再把旧文档固定在缓存中。`reload()`
+  此前先读取文档、后计算缓存签名，若两者之间有写入提交（例如后台记忆更
+  新器），旧文档就会以新签名写入缓存，之后每次 `load()` 都返回过时的记
+  忆，直到下一次写入。`reload()` 现在与 `load()` 一样先计算签名，竞争写
+  入只会触发重新读取。([#6238])
 - **渠道：** Discord 的渠道连接数据库操作现在在 Gateway 事件循环上执行。
   discord.py 在客户端线程的私有事件循环上投递消息，而 Discord 适配器此前就在该
   循环上 await 连接仓库，但仓库的 SQLAlchemy 引擎与连接池属于 Gateway 循环。在
@@ -6414,3 +6419,4 @@ DeerFlow 2.0 是围绕"超级智能体"框架的彻底重写，核心包含子
 [#6214]: https://github.com/bytedance/deer-flow/pull/6214
 [#6229]: https://github.com/bytedance/deer-flow/pull/6229
 [#6230]: https://github.com/bytedance/deer-flow/pull/6230
+[#6238]: https://github.com/bytedance/deer-flow/pull/6238

backend/packages/harness/deerflow/agents/memory/AGENTS.md

@@ -134,6 +134,7 @@ Use the typed conflict classes instead of matching exception text.
 
 The weak lock cache must not retain inactive user scopes.
 Cache validation uses the manifest metadata and persisted revision.
+Unlocked `load()`/`reload()` compute that signature before reading the document.
 Out-of-band Markdown edits require `reload()`.
 POSIX atomic replacement must sync the parent directory.
 

backend/packages/harness/deerflow/agents/memory/backends/deermem/deermem/core/storage.py

@@ -1396,8 +1396,11 @@ class FileMemoryStorage(MemoryStorage):
                 migration_notifications = self._run_read_migrations_locked(path, agent_name, user_id=user_id)
         for notification_agent, notifications in migration_notifications:
             self._dispatch_retrieval_notifications(notifications, user_id=user_id, agent_name=notification_agent)
-        document = self._read_document(path, agent_name, user_id=user_id)
+        # Sign before reading, as load() does: a write landing in between then
+        # leaves a stale signature that forces a re-read, never a stale
+        # document cached under the new signature.
         signature = self._scope_signature(path, agent_name)
+        document = self._read_document(path, agent_name, user_id=user_id)
         with self._cache_lock:
             self._memory_cache[key] = (copy.deepcopy(document), signature)
         if _rebuild_retrieval and agent_name is not None and self._retrieval is not None:

backend/tests/test_memory_storage.py

@@ -191,6 +191,45 @@ class TestFileMemoryStorage:
         memory2 = storage.reload()
         assert memory2["user"]["workContext"]["summary"] == "updated"
 
+    @pytest.mark.parametrize("storage_class", [FileMemoryStorage, MarkdownMemoryStorage])
+    @pytest.mark.parametrize("agent_name", [None, "test-agent"])
+    def test_reload_does_not_pin_a_document_older_than_its_signature(self, tmp_path, storage_class, agent_name):
+        """A write committed while reload() reads must not be hidden from later loads."""
+        storage = storage_class(DeerMemConfig(storage_path=str(tmp_path)))
+
+        def memory_with(text: str) -> dict:
+            memory = create_empty_memory()
+            if agent_name is None:
+                memory["user"]["workContext"]["summary"] = text
+            else:
+                memory["facts"] = [{"id": "fact_1", "content": text, "category": "context", "confidence": 0.9}]
+            return memory
+
+        def text_of(memory: dict) -> str:
+            if agent_name is None:
+                return memory["user"]["workContext"]["summary"]
+            return memory["facts"][0]["content"]
+
+        assert storage.save(memory_with("old"), agent_name, user_id="alice")
+        read_document = storage._read_document
+        raced = False
+
+        def read_then_concurrent_write(*args, **kwargs):
+            nonlocal raced
+            document = read_document(*args, **kwargs)
+            if not raced:
+                raced = True
+                # A background updater commits after reload() has read the
+                # document but before it caches the result.
+                assert storage.save(memory_with("new"), agent_name, user_id="alice")
+            return document
+
+        with patch.object(storage, "_read_document", side_effect=read_then_concurrent_write):
+            storage.reload(agent_name, user_id="alice")
+
+        assert raced
+        assert text_of(storage.load(agent_name, user_id="alice")) == "new"
+
 
 class TestCreateStorage:
     """Test create_storage(config) (replaces the old get_memory_storage() singleton)."""