fix(job_agent): keep multi-word tags when parsing LLM tag:score output

Author: he-yufeng

job_agent.py

@@ -180,7 +180,10 @@
     "技术": ["工程能力", "系统设计"],
 }
 
-_TAG_SCORE_RE = re.compile(r"([^:：\s]+?)\s*[:：]\s*([1-5])(?=\s|$)")
+# Allow spaces inside a tag so a multi-word skill like "Machine Learning" is
+# captured whole, but stop at a newline / comma so we don't run across the
+# boundary between two tags. The captured tag is .strip()-ed by the caller.
+_TAG_SCORE_RE = re.compile(r"([^:：\n,，]+?)\s*[:：]\s*([1-5])(?=\s|$)")
 
 
 def parse_llm_response(reply: str, valid_tags: Set[str]) -> List[Tuple[str, int]]:

tests/test_job_agent.py

@@ -0,0 +1,61 @@
+from job_agent import (
+    deduplicate,
+    format_skill_string,
+    normalize_text,
+    parse_llm_response,
+)
+
+_VALID = {"Python", "Machine Learning", "Data Analysis", "SQL"}
+
+
+def test_parse_keeps_multi_word_tags():
+    # A multi-word skill like "Machine Learning" must be captured whole, not
+    # truncated to its last word (which wouldn't be in the tag set).
+    parsed = parse_llm_response(
+        "Python:5\nMachine Learning:4\nData Analysis:3", _VALID
+    )
+    assert parsed == [("Python", 5), ("Machine Learning", 4), ("Data Analysis", 3)]
+
+
+def test_parse_handles_multiple_tags_on_one_line():
+    parsed = parse_llm_response("Python:5 Machine Learning:4", _VALID)
+    assert parsed == [("Python", 5), ("Machine Learning", 4)]
+
+
+def test_parse_accepts_fullwidth_colon():
+    parsed = parse_llm_response("SQL：2", _VALID)
+    assert parsed == [("SQL", 2)]
+
+
+def test_parse_drops_unknown_and_out_of_range():
+    # Unknown tags and scores outside 1-5 are ignored.
+    parsed = parse_llm_response("Unknown:5\nPython:9\nPython:4", _VALID)
+    assert parsed == [("Python", 4)]
+
+
+def test_parse_dedupes_keeping_first():
+    parsed = parse_llm_response("Python:5\nPython:2", _VALID)
+    assert parsed == [("Python", 5)]
+
+
+def test_parse_strips_markdown_emphasis():
+    parsed = parse_llm_response("**Python**:5", _VALID)
+    assert parsed == [("Python", 5)]
+
+
+def test_parse_empty_returns_empty():
+    assert parse_llm_response("", _VALID) == []
+
+
+def test_normalize_text_strips_whitespace_and_lowercases():
+    assert normalize_text("  Hello   World  ") == "helloworld"
+    assert normalize_text(None) == ""
+    assert normalize_text(123) == "123"
+
+
+def test_deduplicate_preserves_order_and_skips_empty():
+    assert deduplicate(["a", "b", "a", "", "c", "b"]) == ["a", "b", "c"]
+
+
+def test_format_skill_string_matches_csv_shape():
+    assert format_skill_string([("Python", 5), ("SQL", 3)]) == "Python , 5 , AI | SQL , 3 , AI"