diff --git a/DEMO.md b/DEMO.md index 969d96b..17f96e7 100644 --- a/DEMO.md +++ b/DEMO.md @@ -4,6 +4,11 @@ demo_video_url: "" 演示视频录制后请把地址填入上方 `demo_video_url`。以下是无视频时的图文演示脚本。 +## 演示数据准备 + +- 可直接导入 `tests/fixtures/` 三份脱敏样例：教师画像、指南赛道、期望匹配结论 +- 缺省匹配取赛道库**最新年度**赛道：演示前确认"指南赛道"表已录入当年赛道，往年赛道不会混入结果 + ## 演示一：教师问推荐（对话） 1. 对话任务："我是张教授，研究钙钛矿、光伏、半导体和界面调控，SCI 一区 3 篇，今年适合报什么？" @@ -15,3 +20,8 @@ demo_video_url: "" 1. "指南赛道"表批量导入 2027 年各学部赛道；"教师画像"表维护本院教师。 2. 顶栏「赛道匹配」一键匹配；批量模式出汇总页，强推荐名单一目了然。 3. 「生成建议书」重建 HTML 供打印分发。 + +## 演示三：批量容错（可选） + +1. input 目录放入多位教师画像 JSON，混入一个格式损坏的文件。 +2. 发起批量匹配：坏文件被跳过并逐条告警说明，其余教师正常出报告与汇总页。 diff --git a/scripts/service/match_service.py b/scripts/service/match_service.py index bde7b8b..fb26226 100644 --- a/scripts/service/match_service.py +++ b/scripts/service/match_service.py @@ -28,19 +28,28 @@ TIER_NOT = "不建议" _TIER_RANK = {TIER_NOT: 0, TIER_TRY: 1, TIER_STRONG: 2} -# 职称档位：数值越大越高职称；门槛 min_title 使用"X及以上"档位 +# 职称档位：数值越大越高职称；门槛 min_title 使用"X及以上"档位。 +# 助理教授按中级档从审处理（预聘制档次各校不一，宁缺勿滥）； +# 未识别职称遇门槛赛道一律剔除并说明原因（审慎非激进）。 _TITLE_RANK = { "助教": 0, + "初级": 0, "讲师": 1, + "中级": 1, "助理研究员": 1, + "助理教授": 1, + "工程师": 1, "博士后": 1, "副高及以上": 2, "副教授": 2, "副研究员": 2, "高级实验师": 2, + "高级工程师": 2, "正高及以上": 3, "教授": 3, "研究员": 3, + "教授级高级工程师": 3, + "正高级工程师": 3, } _STRONG_THRESHOLD = 80 @@ -197,7 +206,7 @@ def match_teacher(teacher: dict[str, Any], tracks: list[dict[str, Any]], season: if min_sci > 0: reasons.append(f"SCI一区余量：现有 {sci_q1} 篇 / 要求 {min_sci} 篇") if not reasons: - reasons.append("与赛道方向存在部分相关性") + reasons.append("现有研究标签与该赛道需求标签暂无重合，补充研究标签后可能改善匹配") if match_pct >= _STRONG_THRESHOLD: tier = TIER_STRONG @@ -346,14 +355,17 @@ def run_match( def load_tracks(year: Optional[int] = None) -> list[dict[str, Any]]: - """读取赛道库（缺省全部年度；可按年度筛选）。""" + """读取赛道库（缺省取赛道库最新年度，与 actions.json 契约一致；显式传 year 按年度筛选）。""" conn = get_conn() try: cur = conn.cursor() - if year: - cur.execute("SELECT * FROM guide_tracks WHERE year = ? ORDER BY id", (year,)) - else: - cur.execute("SELECT * FROM guide_tracks ORDER BY year DESC, id") + if year is None: + cur.execute("SELECT MAX(year) FROM guide_tracks") + row = cur.fetchone() + if not row or row[0] is None: + return [] + year = int(row[0]) + cur.execute("SELECT * FROM guide_tracks WHERE year = ? ORDER BY id", (year,)) columns = [d[0] for d in cur.description] rows = [dict(zip(columns, row)) for row in cur.fetchall()] finally: @@ -361,6 +373,35 @@ def load_tracks(year: Optional[int] = None) -> list[dict[str, Any]]: return rows +def collect_input_teachers(input_dir: str) -> tuple[list[dict[str, Any]], list[dict[str, str]]]: + """读取 input/ 目录下的教师画像 JSON；坏文件跳过并记录原因，不中断整批。 + + 返回 (教师列表, 跳过文件列表)；跳过文件形如 {"file": 名称, "reason": 原因}。 + """ + import os + + files = sorted( + os.path.join(input_dir, name) + for name in os.listdir(input_dir) + if name.lower().endswith(".json") + ) + teachers: list[dict[str, Any]] = [] + skipped_files: list[dict[str, str]] = [] + for path in files: + name = os.path.basename(path) + try: + with open(path, encoding="utf-8") as f: + payload = json.load(f) + if isinstance(payload, dict): + payload = payload.get("teachers") or [payload] + if not isinstance(payload, list): + raise ValueError("内容不是教师对象数组") + teachers.extend(item for item in payload if isinstance(item, dict)) + except (OSError, ValueError) as exc: + skipped_files.append({"file": name, "reason": str(exc)}) + return teachers, skipped_files + + def load_teachers(teacher_name: str = "") -> list[dict[str, Any]]: """读取教师画像（可按姓名精确筛选）。""" conn = get_conn() diff --git a/scripts/service/task_service.py b/scripts/service/task_service.py index db0e66c..60db9aa 100644 --- a/scripts/service/task_service.py +++ b/scripts/service/task_service.py @@ -478,7 +478,10 @@ def cmd_match(*, teacher: str = "", year: Optional[int] = None) -> int: print(f"ERROR:TEACHER_NOT_FOUND {message}") return 1 if not tracks: - message = "指南赛道库为空：请先录入赛道（add-track / import-tracks）" + if year: + message = f"指南赛道库中没有 {year} 年度赛道：请核对年度或先录入（add-track / import-tracks）" + else: + message = "指南赛道库为空：请先录入赛道（add-track / import-tracks）" log.warning("task_failed task_type=match error_code=TRACK_EMPTY stage=match") finish(status="failed", message=message, skill=SKILL_SLUG, error_code="TRACK_EMPTY", stage="match") print(f"ERROR:TRACK_EMPTY {message}") @@ -529,9 +532,13 @@ def cmd_match(*, teacher: str = "", year: Optional[int] = None) -> int: def cmd_match_batch_input() -> int: - """批量匹配（async Action 入口）：匹配 input/ 目录中的教师画像 JSON 与库内赛道。""" + """批量匹配（async Action 入口）：匹配 input/ 目录中的教师画像 JSON 与库内赛道。 + + 坏文件（解析失败/非对象数组）跳过并告警，不中断整批；画像不完整教师同样跳过并记录（E1）。 + 缺省匹配赛道库最新年度赛道。 + """ log = _get_task_logger() - from service.match_service import load_tracks, run_match + from service.match_service import collect_input_teachers, load_tracks, run_match from util.runtime_paths import get_input_dir with job_context(skill=SKILL_SLUG): @@ -546,26 +553,17 @@ def cmd_match_batch_input() -> int: return 1 input_dir = get_input_dir() - files = sorted( - os.path.join(input_dir, name) - for name in os.listdir(input_dir) - if name.lower().endswith(".json") - ) - if not files: - message = f"input 目录没有教师画像 JSON：{input_dir}" + teachers, skipped_files = collect_input_teachers(input_dir) + for entry in skipped_files: + log.warning("batch_input_file_skipped file=%s reason=%s stage=batch", entry["file"], entry["reason"]) + emit(f"{entry['file']}：{entry['reason']}，该文件已跳过", type="warn", skill=SKILL_SLUG, stage="batch") + if not teachers: + message = f"input 目录没有可用的教师画像 JSON（{len(skipped_files)} 个文件无法解析）：{input_dir}" log.warning("task_failed task_type=match_batch error_code=TEACHER_NOT_FOUND stage=batch") finish(status="failed", message=message, skill=SKILL_SLUG, error_code="TEACHER_NOT_FOUND", stage="batch") print(f"ERROR:TEACHER_NOT_FOUND {message}") return 1 - teachers = [] - for path in files: - with open(path, encoding="utf-8") as f: - payload = json.load(f) - if isinstance(payload, dict): - payload = payload.get("teachers") or [payload] - teachers.extend(payload) - tracks = load_tracks() if not tracks: message = "指南赛道库为空：请先录入赛道（add-track / import-tracks）" @@ -597,6 +595,7 @@ def cmd_match_batch_input() -> int: "batch_no": result["batch_no"], "count": result["count"], "skipped": result.get("skipped") or [], + "skipped_files": skipped_files, "tier_counts": result["tier_counts"], "brief_file": result["brief_file"], } diff --git a/tests/test_match_rules.py b/tests/test_match_rules.py index 4daa124..ccfc40c 100644 --- a/tests/test_match_rules.py +++ b/tests/test_match_rules.py @@ -2,6 +2,9 @@ """赛道匹配内核金标准用例（离线、确定性）。""" from __future__ import annotations +import json +import os +import tempfile import unittest from _support import IsolatedDataRoot @@ -80,13 +83,29 @@ class TestScoring(MatchRuleTestBase): self.assertEqual(top["match_pct"], 100) self.assertEqual(top["tier"], "强烈推荐") - def test_t1_try_band(self) -> None: + def test_coverage_only_below_try_band(self) -> None: result = self._match(_teacher(researchTags="钙钛矿,光伏", recentPapers="")) entry = next(e for e in result["top3"] if e["track"] == "面上项目·光伏材料") - # coverage 2/3 → 0.7*0.667=0.4667 → 47% 不建议？ + 无 priority → 47%；二维材料 coverage 1/3=0.333→23% + # 仅覆盖率 2/3 → 0.7×0.6667=46.7% → 47%，低于 50% 可尝试档 → 不建议 self.assertEqual(entry["match_pct"], 47) self.assertEqual(entry["tier"], "不建议") + def test_t1_try_band_with_supplement_hint(self) -> None: + result = self._match(_teacher(researchTags="钙钛矿,光伏,界面调控")) + entry = next(e for e in result["top3"] if e["track"] == "面上项目·光伏材料") + # 覆盖率 2/3 → 0.7×0.6667=46.7% + 优先方向 1 命中 0.15 = 61.7% → 62%，落在可尝试档 + self.assertEqual(entry["match_pct"], 62) + self.assertEqual(entry["tier"], "可尝试") + self.assertTrue(any("建议补充" in r for r in entry["reasons"])) + + def test_zero_match_reason_is_factual(self) -> None: + result = self._match(_teacher(researchTags="量子计算", recentPapers="", projectHistory="")) + entry = next(e for e in result["top3"] if e["track"] == "面上项目·光伏材料") + reasons_text = "".join(entry["reasons"]) + self.assertEqual(entry["match_pct"], 0) + self.assertIn("暂无重合", reasons_text) + self.assertNotIn("部分相关性", reasons_text) + def test_t1_strong_boundary(self) -> None: # 恰好 80% 属强烈推荐 result = self._match(_teacher(researchTags="钙钛矿,光伏,半导体", recentPapers="")) @@ -210,5 +229,77 @@ class TestBatchSkipIncomplete(MatchRuleTestBase): run_match([_teacher(researchTags="", recentPapers="")], _tracks(), "2027", skip_incomplete=False) +class TestTrackYearScope(MatchRuleTestBase): + """缺省取赛道库最新年度（actions.json 契约）；显式年度不受影响。""" + + def _seed_two_years(self) -> None: + from service.profile_service import upsert_track + + upsert_track({"year": 2026, "track_name": "旧方向", "program_name": "面上项目", "track_tags": "钙钛矿,光伏"}) + upsert_track({"year": 2027, "track_name": "新方向", "program_name": "面上项目", "track_tags": "量子计算"}) + + def test_load_tracks_defaults_to_latest_year(self) -> None: + from service.match_service import load_tracks + + self._seed_two_years() + self.assertEqual({t["year"] for t in load_tracks()}, {2027}) + + def test_load_tracks_explicit_year(self) -> None: + from service.match_service import load_tracks + + self._seed_two_years() + self.assertEqual({t["year"] for t in load_tracks(year=2026)}, {2026}) + + def test_load_tracks_empty_library(self) -> None: + from service.match_service import load_tracks + + self.assertEqual(load_tracks(), []) + + def test_kernel_default_matches_latest_year_only(self) -> None: + from service.match_service import load_teachers, load_tracks, run_match + from service.profile_service import upsert_teacher + + self._seed_two_years() + upsert_teacher({"teacher_name": "张教授", "title": "教授", "research_tags": "量子计算,量子信息"}) + payload = run_match(load_teachers(), load_tracks(), "2027") + report = payload["reports"][0] + all_tracks = [e["track"] for entries in report["tiers"].values() for e in entries] + all_tracks += [e["track"] for e in report["top3"]] + self.assertIn("面上项目·新方向", all_tracks) + self.assertNotIn("面上项目·旧方向", all_tracks) + + +class TestBatchInputCollection(MatchRuleTestBase): + """input/ 目录收集：坏文件跳过并记录原因，不中断整批。""" + + def test_collect_input_teachers_skips_bad_files(self) -> None: + from service.match_service import collect_input_teachers + + with tempfile.TemporaryDirectory() as tmp: + with open(os.path.join(tmp, "a.json"), "w", encoding="utf-8") as f: + json.dump([_teacher(), _teacher(teacherName="李教授")], f, ensure_ascii=False) + with open(os.path.join(tmp, "single.json"), "w", encoding="utf-8") as f: + json.dump({"teacherName": "王教授"}, f, ensure_ascii=False) + with open(os.path.join(tmp, "bad.json"), "w", encoding="utf-8") as f: + f.write("{not-json") + with open(os.path.join(tmp, "scalar.json"), "w", encoding="utf-8") as f: + json.dump("just-a-string", f, ensure_ascii=False) + + teachers, skipped = collect_input_teachers(tmp) + + names = [t.get("teacherName") for t in teachers] + self.assertEqual(names, ["张教授", "李教授", "王教授"]) + self.assertEqual({e["file"] for e in skipped}, {"bad.json", "scalar.json"}) + self.assertTrue(all(e["reason"] for e in skipped)) + + def test_collect_input_teachers_empty_dir(self) -> None: + from service.match_service import collect_input_teachers + + with tempfile.TemporaryDirectory() as tmp: + teachers, skipped = collect_input_teachers(tmp) + self.assertEqual(teachers, []) + self.assertEqual(skipped, []) + + if __name__ == "__main__": unittest.main()