fix: ci test failed

[CarltonXiang]

Pre-submission checklist | 提交前检查

• I have searched existing issues and this hasn't been mentioned before | 我已搜索现有问题，确认此问题尚未被提及
• I have read the project documentation and confirmed this issue doesn't already exist | 我已阅读项目文档并确认此问题尚未存在
• This issue is specific to MemOS and not a general software issue | 该问题是针对 MemOS 的，而不是一般软件问题

Bug Description | 问题描述

Title

Fix failing tests caused by timed_with_status exception semantics and Neo4j integration environment leakage

Description

Recent full test runs fail with:

• FAILED tests/test_utils_timing.py::TestTimedWithStatusRegression::test_failure_logging_no_fallback
• ERROR tests/graph_dbs/test_neo4j_vector_search.py::TestNeo4jPreFilterIntegration::*

There are two separate causes.

1. timed_with_status test expectation is stale

timed_with_status now re-raises exceptions when no fallback is configured.

This behavior was introduced by:

• Commit: 8d0c04be
• Date: 2026-06-04 19:59:18 +0800
• Message: fix: re-raise exceptions in timed_with_status when no fallback configured

The new regression test tests/test_utils_timing_exception_reraise.py expects this behavior and passes.

However, the older test added in:

• Commit: f8ab9624
• Date: 2026-04-17 16:17:45 +0800
• Message: feat: add stage log (#1486)

still assumes fail_func() does not raise and only logs status: FAILED.

So tests/test_utils_timing.py::test_failure_logging_no_fallback should be updated to expect RuntimeError while still asserting the FAILED log is emitted.

2. Neo4j integration tests run when remote env vars are present

tests/graph_dbs/test_neo4j_vector_search.py::TestNeo4jPreFilterIntegration is enabled whenever all of these environment variables exist:

• NEO4J_URI
• NEO4J_USER
• NEO4J_PASSWORD

When these point to an unavailable remote Neo4j instance, the full test suite fails with neo4j.exceptions.ServiceUnavailable.

These integration tests were added in:

• Commit: fd2d4c76
• Date: 2026-03-27 15:46:04 +0800
• Message: fix: Neo4j vector search pre-filter and metadata handling (#1359)

Expected Behavior

• Unit test suite should not fail because optional external Neo4j credentials happen to exist.
• timed_with_status tests should consistently reflect the current behavior: no fallback means re-raise after logging failure.

Suggested Fix

1. Update test_failure_logging_no_fallback to wrap fail_func() with pytest.raises(RuntimeError).
2. Gate

How to Reproduce | 如何重现

make test

Environment | 环境信息

make test

Additional Context | 其他信息

No response

Willingness to Implement | 实现意愿

• I'm willing to implement this myself | 我愿意自己解决
• I would like someone else to implement this | 我希望其他人来解决

[Memtensor-AI]

🤖 AutoDev has picked up this issue and started working on it.

Task ID: fa962e630e4b5af9
Working branch: bugfix/autodev-2008
Target branch: latest dev* branch
Workflow: opsp (analysis → coding → testing → PR)

I will post the PR link here once done. If I need more information, I will ask in the comments.

[Memtensor-AI]

✅ AutoDev task fa962e630e4b5af9 completed.

Summary: Fixes the two CI failures reported in issue #2008.

1. timed_with_status re-raises on failure (no fallback) — src/memos/utils.py::timed_with_status used to catch every exception and, when no fallback= was configured, silently return None. That masked real errors from callers like OpenAILLM.generate and UniversalAPIEmbedder.embed which internally raise expecting propagation. The decorator now re-raises after logging the [TIMER_WITH_STATUS] status: FAILED line whenever no fallback is set; callers that pass a fallback= behave exactly as before. tests/test_utils_timing.py::TestTimedWithStatusRegression::test_failure_logging_no_fallback was updated to wrap the call in pytest.raises(RuntimeError, match="bad") while still asserting the FAILED log line is emitted.

2. Neo4j integration tests now require explicit opt-in — tests/graph_dbs/test_neo4j_vector_search.py::TestNeo4jPreFilterIntegration used to activate whenever NEO4J_URI / NEO4J_USER / NEO4J_PASSWORD were present in the environment, which turned unrelated shell config into a neo4j.exceptions.ServiceUnavailable failure. The class-level gate now additionally requires NEO4J_INTEGRATION_TESTS=1 (accepts 1/true/yes/on). The skip reason and header docs are updated to reflect the new invocation.

Test evidence: pytest tests/test_utils_timing.py → 31 passed. NEO4J_URI=bolt://192.0.2.1:7687 NEO4J_USER=neo4j NEO4J_PASSWORD=fake pytest tests/graph_dbs/test_neo4j_vector_search.py (the exact issue-reported scenario) → 9 passed, 3 skipped; before the fix this would have raised ServiceUnavailable. Broader sanity run across tests/test_utils_timing.py tests/graph_dbs/ tests/embedders/ tests/llms/ tests/reranker/ → 111 passed, 3 skipped. ruff check and ruff format both clean on the three touched files. Other unrelated failures under tests/api/, tests/parsers/, tests/vec_dbs/ are pre-existing missing-plugin / missing-module issues that reproduce identically on the untouched base branch.

Files touched: src/memos/utils.py, tests/test_utils_timing.py, tests/graph_dbs/test_neo4j_vector_search.py (3 files, +34 / -8). .ai-tasks/ and openspec/changes/ are gitignored per project rules and did not enter the PR; task file + verification report have been archived to the sibling memos-autodev-specs repo.

Base branch: main
Branch: bugfix/autodev-2008
Commit: aa225e7
PR: #2010
Assigned to: @MatthewZhuang, @CarltonXiang, @syzsunshine219, @World-controller
Reviewers: @MatthewZhuang, @CarltonXiang, @syzsunshine219, @World-controller