This is an automated email from the ASF dual-hosted git repository.
wenjin272 pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/flink-agents.git
The following commit(s) were added to refs/heads/main by this push:
new a28336fa [test][python] Assert long-term memory content instead of an
exact LLM item count (#984)
a28336fa is described below
commit a28336fa9388a4d0c8176052525ac3508dc7ac40
Author: Weiqing Yang <[email protected]>
AuthorDate: Fri Aug 14 05:53:27 2026 -0700
[test][python] Assert long-term memory content instead of an exact LLM item
count (#984)
Generated-by: Claude Code 2.1.228
---
.../e2e_tests_integration/long_term_memory_test.py | 47 +++++++++++++++++++---
1 file changed, 41 insertions(+), 6 deletions(-)
diff --git
a/python/flink_agents/e2e_tests/e2e_tests_integration/long_term_memory_test.py
b/python/flink_agents/e2e_tests/e2e_tests_integration/long_term_memory_test.py
index b2d7bfe4..33f9c2db 100644
---
a/python/flink_agents/e2e_tests/e2e_tests_integration/long_term_memory_test.py
+++
b/python/flink_agents/e2e_tests/e2e_tests_integration/long_term_memory_test.py
@@ -303,13 +303,48 @@ def check_result(*, result_dir: Path) -> None:
for record in actual_result:
records[f"{record.name}.{record.count}"] = record
+ assert "alice.2" in records, f"missing alice.2; got {sorted(records)}"
+ assert "bob.2" in records, f"missing bob.2; got {sorted(records)}"
+ # The extraction model decides whether each key's two facts collapse into a
+ # single item or stay separate, so the item count is not fixed. Each set is
+ # reported alongside the other's failure to tell a per-key miss from a
+ # store-wide one.
items = records["alice.2"].items
- # LLMs may treat different review comments as updates to the same
- # fact or as distinct facts.
- assert len(items) == 1
- item: MemorySetItem = items[-1]
- assert item.created_at < item.updated_at
- assert "bananas" in item.value
+ values = [item.value for item in items or []]
+ bob_values = [item.value for item in records["bob.2"].items or []]
+ # A scoping break misattributes in either direction. alice's facts landing
+ # in bob's set leaves alice's set short rather than inflated, so check that
+ # direction ahead of the emptiness assertion below, which would otherwise
+ # report the symptom in place of the cause. bob's set has to be populated
+ # for that scan to carry any weight, since an empty one satisfies it
+ # without ever being examined.
+ assert bob_values, f"bob's memory set is empty (alice's set: {values})"
+ bob_leaked = [
+ value
+ for value in bob_values
+ if "watermelon" in value.lower() or "bananas" in value.lower()
+ ]
+ assert not bob_leaked, f"bob's set contains alice's facts: {bob_values}"
+ assert values, f"alice's memory set is empty (bob's set: {bob_values})"
+
+ # The stored text is the model's paraphrase of the input, so match loosely.
+ assert any("bananas" in value.lower() for value in values), (
+ f"no stored item carries the updated fact: {values}"
+ )
+ # Each partition key is scoped to its own memories, so bob's facts must
+ # never surface in alice's set.
+ leaked = [
+ value
+ for value in values
+ if "swimming" in value.lower() or "vegetarian" in value.lower()
+ ]
+ assert not leaked, f"alice's set contains bob's facts: {values}"
+ # Both timestamps are populated whether an item was created or updated, and
+ # the parser turns an unrecognized format into None rather than raising.
+ assert all(item.created_at and item.updated_at for item in items), (
+ f"stored items are missing timestamps: "
+ f"{[(item.created_at, item.updated_at) for item in items]}"
+ )
# verify async add doesn't block process other key
assert datetime.fromisoformat(
