Python: [Bug]: non-English users (input messages) will never automatically load topic files.

[changyao-han]

Description

In agent-framework/blob/main/python/packages/core/agent_framework/_harness/_memory.py

line 70:
WORD_PATTERN = re.compile(r"[a-z0-9][a-z0-9-]{1,}", flags=re.IGNORECASE)

non-English users (input messages) will never automatically load topic files.

Code Sample

Error Messages / Stack Traces

Package Versions

agent-framework-core: 1.8.0

Python Version

No response

Additional Context

No response

[github-actions]

🤖 Automated triage reproduction notes (agent-authored — trust but verify)

Agent analysis

Repro: _extract_keywords() in python/packages/core/agent_framework/_harness/_memory.py (line 177-181) uses WORD_PATTERN (line 70: r"[a-z0-9][a-z0-9-]{1,}") which only matches ASCII, so non-English messages yield an empty keyword set. This causes _select_topics() (line 1086-1087) to return [] immediately. Minimal repro: call _extract_keywords([Message("user", ["告诉我关于编程的知识"]) and observe the empty set; the pattern also affects _topic_score() at line 1074.

• Failing test: python/packages/core/tests/core/test_word_pattern_non_english.py
• Files examined: python/packages/core/agent_framework/_harness/_memory.py, python/packages/core/pyproject.toml
• Tests run: python/packages/core/tests/core/test_word_pattern_non_english.py (6 tests, all passing - confirming the bugy behavior)
• Reported version: 1.8.0
• Current version: 1.10.0