fix(memory): serialize cursor allocation in append_history (#4081)

nanobot/agent/memory.py

@@ -6,6 +6,7 @@ import asyncio
 import json
 import os
 import re
+import threading
 import weakref
 from contextlib import suppress
 from datetime import datetime
@@ -61,6 +62,7 @@ class MemoryStore:
         self._dream_cursor_file = self.memory_dir / ".dream_cursor"
         self._corruption_logged = False  # rate-limit non-int cursor warning
         self._oversize_logged = False  # rate-limit oversized-entry warning
+        self._append_lock = threading.Lock()  # serialize cursor allocation + append
         self._git = GitStore(workspace, tracked_files=[
             "SOUL.md", "USER.md", "memory/MEMORY.md", "memory/.dream_cursor",
         ])
@@ -248,7 +250,6 @@ class MemoryStore:
         large writes (e.g. an LLM echoing its input back as a "summary").
         """
         limit = max_chars if max_chars is not None else _HISTORY_ENTRY_HARD_CAP
-        cursor = self._next_cursor()
         ts = datetime.now().strftime("%Y-%m-%d %H:%M")
         raw = entry.rstrip()
         if len(raw) > limit:
@@ -262,16 +263,20 @@ class MemoryStore:
                 )
             raw = truncate_text(raw, limit)
         content = strip_think(raw)
-        if raw and not content:
-            logger.debug(
-                "history entry {} stripped to empty (likely template leak); "
-                "persisting empty content to avoid re-polluting context",
-                cursor,
-            )
-        record = {"cursor": cursor, "timestamp": ts, "content": content}
-        with open(self.history_file, "a", encoding="utf-8") as f:
-            f.write(json.dumps(record, ensure_ascii=False) + "\n")
-        self._cursor_file.write_text(str(cursor), encoding="utf-8")
+        # Cursor allocation and the append must be atomic: concurrent writers
+        # could otherwise read the same current cursor and emit duplicates.
+        with self._append_lock:
+            cursor = self._next_cursor()
+            if raw and not content:
+                logger.debug(
+                    "history entry {} stripped to empty (likely template leak); "
+                    "persisting empty content to avoid re-polluting context",
+                    cursor,
+                )
+            record = {"cursor": cursor, "timestamp": ts, "content": content}
+            with open(self.history_file, "a", encoding="utf-8") as f:
+                f.write(json.dumps(record, ensure_ascii=False) + "\n")
+            self._cursor_file.write_text(str(cursor), encoding="utf-8")
         return cursor
 
     @staticmethod

tests/agent/test_memory_store.py

@@ -129,6 +129,33 @@ class TestHistoryWithCursor:
         cursor = store.append_history("new event")
         assert cursor == 1
 
+    def test_append_history_allocates_unique_cursors_under_concurrent_writes(self, store):
+        """Regression: concurrent appends must not allocate duplicate cursors."""
+        import threading
+
+        writers = 16
+        start = threading.Barrier(writers)
+        cursors: list[int] = []
+        lock = threading.Lock()
+
+        def worker(i):
+            start.wait()
+            c = store.append_history(f"event {i}")
+            with lock:
+                cursors.append(c)
+
+        threads = [threading.Thread(target=worker, args=(i,)) for i in range(writers)]
+        for t in threads:
+            t.start()
+        for t in threads:
+            t.join()
+
+        assert len(cursors) == writers
+        assert len(set(cursors)) == writers, f"duplicate cursors: {sorted(cursors)}"
+        assert sorted(cursors) == list(range(1, writers + 1))
+        persisted = store.read_unprocessed_history(since_cursor=0)
+        assert sorted(e["cursor"] for e in persisted) == list(range(1, writers + 1))
+
     def test_compact_history_drops_oldest(self, tmp_path):
         store = MemoryStore(tmp_path, max_history_entries=2)
         store.append_history("event 1")