diff --git a/backend/tests/test_chat_agent.py b/backend/tests/test_chat_agent.py index c720c7b..36666ea 100644 --- a/backend/tests/test_chat_agent.py +++ b/backend/tests/test_chat_agent.py @@ -136,7 +136,7 @@ def test_scoring_uses_the_same_page_image_as_its_saved_evidence(db, environment) sent_images.extend(images or []) return ScoreResult(criterion_id=criterion["criterion_id"], score=1, max_score=criterion["max_score"], reason="测试评分", - evidence=[{"chunk_id": chunk.id, "page": 2, "quote": "原文证据"}], confidence=.9) + evidence=[{"chunk_id": chunk.id, "page": 2, "quote": "原文证据", "target_id": "t0"}], confidence=.9) # The evidence page must be a substantive body page: a contents page cannot support # a content judgement, so the source page is classified accordingly during ingest. @@ -214,6 +214,12 @@ def test_duplicate_scan_requires_teacher_file_choice(environment, db): generated_action = db.get(PendingAction, generated["pending_action"]["id"]) assert generated_action.preview["requires_edit"] is True assert all(file["status"] == "ignored" for file in generated_action.payload["projects"][0]["files"]) + with pytest.raises(HTTPException) as blocked: + confirm(db, conversation, generated_action.id) + assert blocked.value.status_code == 409 + with pytest.raises(HTTPException) as invalid_selection: + prepare(db, conversation, "import", generated_action.payload, teacher_edited=True) + assert invalid_selection.value.status_code == 400 proposed = prepare(db, conversation, "import", payload) assert proposed.preview["requires_edit"] is True with pytest.raises(HTTPException): diff --git a/backend/tests/test_chat_sessions.py b/backend/tests/test_chat_sessions.py index 1a236eb..374843d 100644 --- a/backend/tests/test_chat_sessions.py +++ b/backend/tests/test_chat_sessions.py @@ -388,7 +388,8 @@ def test_upgrade_from_legacy_0003_migrates_pages_and_citations(session_factory, config.set_main_option('script_location', str(root / 'migrations')) command.upgrade(config, 'head') with session_factory() as db: - assert db.scalar(text('SELECT version_num FROM alembic_version')) == '0006' + from alembic.script import ScriptDirectory + assert db.scalar(text('SELECT version_num FROM alembic_version')) == ScriptDirectory.from_config(config).get_current_head() assert db.get(Conversation, conversation_id).title == '旧记录' assert db.scalar(select(ChatMessage.content).where(ChatMessage.conversation_id == conversation_id)) == '旧问题' run = db.scalar(select(AgentRun).where(AgentRun.conversation_id == conversation_id)) diff --git a/backend/tests/test_evidence_pages.py b/backend/tests/test_evidence_pages.py index 7ddbcfd..8442eee 100644 --- a/backend/tests/test_evidence_pages.py +++ b/backend/tests/test_evidence_pages.py @@ -428,13 +428,14 @@ def test_project_and_version_isolation_is_unchanged(db): def test_batch_snapshot_records_the_evidence_policy(db): from types import SimpleNamespace from wisegrade import services + from wisegrade.evidence import POLICY_VERSION from wisegrade.models import ScoringBatch created = services.create_batch(SimpleNamespace(course_version_id=1, rubric_version_id=1), db) assert created["id"] batch = db.get(ScoringBatch, created["id"]) plan = batch.rubric_snapshot["evidence_plan"] - assert plan["version"] == 1 + assert plan["version"] == POLICY_VERSION assert plan["content_allowed_page_types"] == ["body", "appendix"] targets = plan["criteria"]["stage3.data"] assert [target["use"] for target in targets] == ["content", "content"] diff --git a/backend/tests/test_import_and_model_errors.py b/backend/tests/test_import_and_model_errors.py index d0b1e1a..1e63df6 100644 --- a/backend/tests/test_import_and_model_errors.py +++ b/backend/tests/test_import_and_model_errors.py @@ -24,6 +24,8 @@ def test_scan_selected_student_folder_prefills_cover_identity(tmp_path, monkeypa assert project["status"] == "needs_confirmation" assert len(project["files"]) == 5 assert project["missing"] == [] + assert project["folder_identity"] == {"student_no": "", "name": "", "topic": ""} + assert all(file["detected_identity"]["student_no"] == "9000000001" for file in project["files"]) assert scan("test", root=tmp_path)["projects"] == result["projects"] @@ -39,6 +41,26 @@ def test_scan_conflicting_covers_need_confirmation(tmp_path, monkeypatch): assert project["student_no"] == "" assert project["warnings"] assert project["status"] == "needs_confirmation" + assert "学号" in project["warnings"][0] + assert "student_no" not in project["warnings"][0] + assert {file["detected_identity"]["student_no"] for file in project["files"]} == {"9000000001", "9000000002"} + + +def test_scan_unreadable_cover_has_no_stale_identity(tmp_path, monkeypatch): + from wisegrade import storage + folder = tmp_path / "9000000001_学生甲_机器人系统" + folder.mkdir() + (folder / "阶段报告1.pdf").write_bytes(b"fixture") + (folder / "阶段报告2.pdf").write_bytes(b"broken") + def read(path): + if path.name.endswith("2.pdf"): + raise ValueError("损坏PDF") + return "学号：9000000001\n姓名：学生甲\n题目：机器人系统" + monkeypatch.setattr(storage, "first_text", read) + project = scan(".", root=tmp_path)["projects"][0] + assert project["folder_identity"]["student_no"] == "9000000001" + assert project["files"][0]["detected_identity"]["student_no"] == "9000000001" + assert project["files"][1]["detected_identity"] == {} def test_scan_parent_skips_empty_folders(tmp_path, monkeypatch): diff --git a/backend/tests/test_import_validation.py b/backend/tests/test_import_validation.py new file mode 100644 index 0000000..19487d9 --- /dev/null +++ b/backend/tests/test_import_validation.py @@ -0,0 +1,36 @@ +import pytest +from pydantic import ValidationError + +from wisegrade.api import ImportConfirm, ProjectConfirm + + +def project(**changes): + return {"relative_dir": "test/student", "student_no": "9000000001", "name": "学生甲", "topic": "机器人项目", + "files": [{"relative_path": "test/student/最终报告.pdf", "category": "final_report", "sha256": "a" * 64}], **changes} + + +def test_confirm_normalizes_identity_before_validating(): + result = ProjectConfirm.model_validate(project(student_no=" 9000000001 ", name=" 学生甲 ", topic=" 机器人项目 ")) + assert (result.student_no, result.name, result.topic) == ("9000000001", "学生甲", "机器人项目") + + +@pytest.mark.parametrize("changes", [ + {"name": " "}, {"topic": " "}, {"student_no": "９００００００００１"}, {"files": []}, + {"files": [{"relative_path": "test/student/a.pdf", "category": "unknown", "sha256": "a" * 64}]}, + {"files": [{"relative_path": "test/student/a.pdf", "category": "final_report", "sha256": "a" * 64, "status": "ignored"}]}, + {"files": [{"relative_path": f"test/student/{i}.pdf", "category": "final_report", "sha256": "a" * 64} for i in range(2)]}, +]) +def test_confirm_rejects_invalid_identity_or_selection(changes): + with pytest.raises(ValidationError): + ProjectConfirm.model_validate(project(**changes)) + + +def test_empty_import_is_rejected(): + with pytest.raises(ValidationError): + ImportConfirm(course_version_id=1, projects=[]) + + +def test_ignored_unknown_material_does_not_block_valid_selection(): + data = project() + data["files"].append({"relative_path": "test/student/foreign.pdf", "category": "unknown", "sha256": "b" * 64, "status": "ignored"}) + assert len(ProjectConfirm.model_validate(data).files) == 2 diff --git a/backend/tests/test_optimization.py b/backend/tests/test_optimization.py index 95a58d8..75cf4e8 100644 --- a/backend/tests/test_optimization.py +++ b/backend/tests/test_optimization.py @@ -26,7 +26,7 @@ from wisegrade.pages import classify_document from wisegrade.parsing import text_quality from wisegrade.rag import INDEX_VERSION, build_chunks, ingest, keyword_overlap, query_terms, retrieve, retrieve_targets from wisegrade.rules import Criterion, EvidenceTarget -from wisegrade.scoring import _attach_page_types, _score_criterion +from wisegrade.scoring import _attach_page_types, _body_navigator, _score_criterion from wisegrade.seed import seed_initial from wisegrade.storage import extract_identity, sha256 from wisegrade.submissions import file_index_lock @@ -149,6 +149,38 @@ def test_shared_chunk_keeps_all_target_ids(): assert selection["evidence"][0]["target_ids"] == ["t0", "t1"] +def test_plan_navigation_finds_future_work_body_despite_nonempty_semantic_hits(sqlite_db): + db = sqlite_db + item = project(db) + doc = document(db, item, pages=[ + {"page_number": 3, "text": "目录\n4.2 后续改进方向.......... 11\n4.3 参考文献.......... 12"}, + {"page_number": 6, "text": "2.3 系统控制\n控制器通过闭环反馈调整电机转速，实现稳定的运动控制。"}, + {"page_number": 14, "text": "4.2 后续改进方向\n后续将探索深度学习检测算法，提高复杂光照场景下的鲁棒性；" + "引入模型预测控制优化动态跟踪，并完善状态监控和数据回放功能。"}, + ]) + noise = db.scalar(select(KnowledgeChunk).where(KnowledgeChunk.document_id == doc.id, + KnowledgeChunk.page == 6)) + navigator = _body_navigator(db, course_version_id=1, project_id=item.id, category="stage3") + targets = normalize_targets(["计划", "章节"]) + selected = retrieve_targets(db, course_version_id=1, project_id=item.id, category="stage3", + targets=targets, navigator=navigator, limit=2, + retrieve_fn=lambda *args, **kwargs: [{ + "chunk_id": noise.id, "document_id": doc.id, "page": 6, + "page_type": "body", "text": noise.text, "rank": .03}]) + assert any(hit["page"] == 14 and "t0" in hit["target_ids"] for hit in selected["evidence"]) + assert all(hit["page_type"] == "body" for hit in selected["evidence"]) + assert navigator({**targets[0], "allowed_page_types": ["appendix"]}) == [] + + +def test_heading_navigation_supplements_existing_hits_for_custom_labels(): + target = normalize_targets(["实验数据"]) + raw = {"chunk_id": 1, "document_id": 1, "page": 2, "text": "仅描述装置。", "page_type": "body", "rank": .03} + body = {"chunk_id": 2, "document_id": 1, "page": 5, "text": "实测响应时间80毫秒。", "page_type": "body", "rank": .5} + result = retrieve_targets(None, course_version_id=1, project_id=1, targets=target, limit=1, + retrieve_fn=lambda *args, **kwargs: [raw], navigator=lambda _: [body]) + assert result["evidence"][0]["chunk_id"] == 2 + + def test_document_filter_and_index_readiness_precede_limit(sqlite_db): db = sqlite_db item = project(db) @@ -341,6 +373,100 @@ def test_model_declining_score_records_original_gap(sqlite_db): assert score.evidence_gaps[0]["reason"] == "只有装置介绍" +@pytest.mark.parametrize("missing", [["误差分析（t1）：缺少误差来源和统计数据"], + ["尚缺 t1 对应的正文"], ["当前资料无法判断"]]) +def test_declined_score_attributes_descriptive_missing_evidence_to_its_target(sqlite_db, missing): + db = sqlite_db + item = project(db) + file = SubmissionFile(project_id=item.id, category="stage3", relative_path="test.pdf", mime_type="application/pdf", + size=1, mtime=1, sha256="test", recognition_status="recognized", parse_status="ready") + db.add(file) + db.commit() + ingest(db, course_version_id=1, project_id=item.id, submission_file_id=file.id, title="test.pdf", + category="stage3", source_hash="test", pages=[{"page_number": 1, "text": "实验测得响应时间80毫秒，但未解释误差来源。"}], embedder=FLAT) + criterion = {"criterion_id": "stage3.data", "category": "stage3", "max_score": 20, + "description": "实验数据与误差分析", "required_evidence": ["实验数据", "误差分析"]} + created = services.create_batch(SimpleNamespace(course_version_id=1, rubric_version_id=1), db) + class Client: + def json_call(self, *args, **kwargs): + return ScoreResult(criterion_id="stage3.data", score=None, max_score=20, + reason="误差分析尚不足以判断", missing_evidence=missing, confidence=.2) + def retriever(session, **kwargs): + if kwargs.get("public_only"): + return [] + return retrieve(session, **{**kwargs, "embedder": FLAT}) + score = _score_criterion(db, db.get(ScoringBatch, created["id"]), item, criterion, Client(), retriever=retriever) + expected = ["t0", "t1"] if missing == ["当前资料无法判断"] else ["t1"] + assert [gap["target_id"] for gap in score.evidence_gaps] == expected + assert score.status == "needs_review" and score.model_score is None + + +def test_failed_model_call_cannot_overwrite_a_concurrent_teacher_review(sqlite_db, monkeypatch): + from wisegrade import scoring + db = sqlite_db + item = project(db) + created = services.create_batch(SimpleNamespace(course_version_id=1, rubric_version_id=1), db) + batch = db.get(ScoringBatch, created["id"]) + criterion = batch.rubric_snapshot["criteria"][0] + batch.rubric_snapshot = {**batch.rubric_snapshot, "criteria": [criterion]} + batch.status, batch.progress_total, batch.progress_done = "running", 1, 0 + score = CriterionScore(batch_id=batch.id, project_id=item.id, criterion_id=criterion["criterion_id"], status="pending") + db.add(score) + db.commit() + factory = sessionmaker(bind=db.get_bind(), expire_on_commit=False) + def teacher_then_failure(*args, **kwargs): + with factory() as teacher_db: + services.review_score(score.id, SimpleNamespace(score=0, comment="教师核定零分", reason="已核对"), teacher_db) + raise ModelError("模型响应超时") + monkeypatch.setattr(scoring, "_project_facts", lambda *args, **kwargs: {}) + monkeypatch.setattr(scoring, "_score_criterion", teacher_then_failure) + failures, reviews = scoring._run_project(batch.id, item.id, factory, lambda *args: None, FLAT, lambda *a, **k: []) + db.expire_all() + saved = db.get(CriterionScore, score.id) + assert saved.status == "reviewed" + assert saved.teacher_score == 0 and saved.teacher_comment == "教师核定零分" + assert saved.error == "" + assert failures == reviews == 0 + assert db.get(ScoringBatch, batch.id).progress_done == 1 + + +@pytest.mark.parametrize("complete", [True, False]) +def test_retry_skips_facts_only_for_fully_graded_projects(sqlite_db, monkeypatch, complete): + from wisegrade import scoring + db = sqlite_db + item = project(db) + created = services.create_batch(SimpleNamespace(course_version_id=1, rubric_version_id=1), db) + batch = db.get(ScoringBatch, created["id"]) + criteria = batch.rubric_snapshot["criteria"][:2] + batch.rubric_snapshot = {**batch.rubric_snapshot, "criteria": criteria} + batch.status, batch.progress_total, batch.progress_done = "running", 2, 0 + item.facts = {"conclusions": ["已抽取并确认的事实"]} + db.add_all([ + CriterionScore(batch_id=batch.id, project_id=item.id, criterion_id=criteria[0]["criterion_id"], status="scored", model_score=0), + CriterionScore(batch_id=batch.id, project_id=item.id, criterion_id=criteria[1]["criterion_id"], + status="reviewed" if complete else "needs_review", teacher_score=0 if complete else None), + ]) + db.commit() + calls = [] + def client_factory(*args): + calls.append("client") + def facts(*args, **kwargs): + calls.append("facts") + return {} + def scored(session, batch, item, criterion, *args, **kwargs): + return session.scalar(select(CriterionScore).where(CriterionScore.batch_id == batch.id, + CriterionScore.project_id == item.id, CriterionScore.criterion_id == criterion["criterion_id"])) + monkeypatch.setattr(scoring, "_project_facts", facts) + monkeypatch.setattr(scoring, "_score_criterion", scored) + factory = sessionmaker(bind=db.get_bind(), expire_on_commit=False) + failures, reviews = scoring._run_project(batch.id, item.id, factory, client_factory, FLAT, lambda *a, **k: []) + assert calls == ([] if complete else ["client", "facts"]) + assert failures == 0 and reviews == (0 if complete else 1) + db.expire_all() + assert db.get(ScoringBatch, batch.id).progress_done == 2 + assert db.get(Project, item.id).facts == {"conclusions": ["已抽取并确认的事实"]} + + def test_unrebuilt_course_document_does_not_complete_task(sqlite_db): db = sqlite_db doc = KnowledgeDocument(course_version_id=1, title="无页面资料", document_type="course", @@ -532,6 +658,75 @@ def test_valid_target_cannot_be_attached_to_another_targets_chunk(): validate_score(result, criterion, [chunk]) +def test_legacy_missing_target_uses_unique_chunk_association(): + criterion = {"criterion_id": "x", "max_score": 10, "required_evidence": ["实验数据", "实现方法"]} + chunk = {"chunk_id": 1, "page": 1, "page_type": "body", "text": "采用视觉传感器实现位置反馈。", "target_ids": ["t1"]} + result = ScoreResult(criterion_id="x", score=8, max_score=10, reason="有实现方法", confidence=.8, + evidence=[{"chunk_id": 1, "page": 1, "quote": chunk["text"]}]) + assert validate_score(result, criterion, [chunk]).evidence[0].target_id == "t1" + + +def test_target_mismatch_feedback_identifies_the_legal_chunk_target(): + criterion = {"criterion_id": "x", "max_score": 10, "required_evidence": ["实验数据", "实现方法"]} + chunk = {"chunk_id": 42, "page": 1, "page_type": "body", "text": "采用视觉传感器实现位置反馈。", "target_ids": ["t1"]} + result = ScoreResult(criterion_id="x", score=8, max_score=10, reason="错误关联", confidence=.8, + evidence=[{"chunk_id": 42, "page": 1, "quote": chunk["text"], "target_id": "t0"}]) + with pytest.raises(ModelError) as failure: + validate_score(result, criterion, [chunk]) + assert "chunk_id=42" in str(failure.value) + assert "允许 target_id=t1" in str(failure.value) + + +@pytest.mark.parametrize("rich", [False, True]) +def test_missing_target_is_not_guessed_when_ambiguous_or_required(rich): + required = [{"label": "实验数据"}, {"label": "实现方法"}] if rich else ["实验数据", "实现方法"] + criterion = {"criterion_id": "x", "max_score": 10, "required_evidence": required} + chunk = {"chunk_id": 1, "page": 1, "page_type": "body", "text": "采用视觉传感器实现位置反馈。", "target_ids": ["t0", "t1"]} + result = ScoreResult(criterion_id="x", score=8, max_score=10, reason="需要明确目标", confidence=.8, + evidence=[{"chunk_id": 1, "page": 1, "quote": chunk["text"]}]) + with pytest.raises(ModelError, match="缺少 target_id"): + validate_score(result, criterion, [chunk]) + + +def test_scoring_retry_carries_target_contract_and_specific_repair_feedback(sqlite_db): + db = sqlite_db + item = project(db) + file = SubmissionFile(project_id=item.id, category="stage3", relative_path="test.pdf", mime_type="application/pdf", + size=1, mtime=1, sha256="test", recognition_status="recognized", parse_status="ready") + db.add(file) + db.commit() + doc = ingest(db, course_version_id=1, project_id=item.id, submission_file_id=file.id, title="test.pdf", + category="stage3", source_hash="test", pages=[{"page_number": 1, "text": "采用视觉传感器实现位置反馈。"}], embedder=FLAT) + chunk = db.scalar(select(KnowledgeChunk).where(KnowledgeChunk.document_id == doc.id)) + criterion = {"criterion_id": "stage3.design", "category": "stage3", "max_score": 20, + "description": "实现方法", "required_evidence": ["实验数据", "实现方法"]} + created = services.create_batch(SimpleNamespace(course_version_id=1, rubric_version_id=1), db) + batch = db.get(ScoringBatch, created["id"]) + db.add(CriterionScore(batch_id=batch.id, project_id=item.id, criterion_id=criterion["criterion_id"], + status="needs_review", evidence_gaps=[{"target_id": "t1", "reason": "上次未命中正文"}])) + db.commit() + class Client: + calls = 0 + def json_call(self, prompt, payload, *args, **kwargs): + self.calls += 1 + assert "每条引文的 target_id 必须属于该片段的 target_ids" in prompt + if self.calls == 2: + assert payload["validation_feedback"]["error"].endswith("允许 target_id=t1") + return ScoreResult(criterion_id=criterion["criterion_id"], score=15, max_score=20, + reason="有实现方法", confidence=.8, evidence=[{ + "chunk_id": chunk.id, "page": 1, "quote": chunk.text, + "target_id": "t0" if self.calls == 1 else "t1"}]) + client = Client() + def selected(*args, **kwargs): + return {"evidence": [{"chunk_id": chunk.id, "document_id": doc.id, "page": 1, + "text": chunk.text, "page_type": "body", "target_ids": ["t1"]}], + "uncovered_targets": []} + score = _score_criterion(db, batch, item, criterion, client, + retriever=lambda *args, **kwargs: [], multi_retriever=selected) + assert score.status == "scored" and client.calls == 2 + assert score.evidence_gaps == [] + + @pytest.mark.parametrize("version", [None, 1]) def test_worker_rejects_old_policy_without_touching_scores(sqlite_db, version): from wisegrade.scoring import run_batch diff --git a/backend/wisegrade/actions.py b/backend/wisegrade/actions.py index bd46d95..c8542b9 100644 --- a/backend/wisegrade/actions.py +++ b/backend/wisegrade/actions.py @@ -6,6 +6,7 @@ from pathlib import Path from uuid import uuid4 from fastapi import HTTPException +from pydantic import ValidationError from sqlalchemy import select from . import services @@ -149,7 +150,12 @@ def prepare(db, conversation: Conversation, kind: str, payload: dict, *, teacher raise HTTPException(400, "整批 Excel 导出须在未限定学生项目的会话中准备") if kind == "import": from .api import ImportConfirm - data = ImportConfirm.model_validate(payload) + try: + data = ImportConfirm.model_validate(payload, context={ + "allow_unselected_import_draft": not teacher_edited}) + except ValidationError as exc: + messages = [error["msg"] for error in exc.errors(include_input=False, include_context=False)] + raise HTTPException(400, "导入资料尚未满足确认条件：" + "；".join(messages)) from exc payload = data.model_dump() require_scope(db, conversation, version_id=data.course_version_id) if teacher_edited and any(item.student_no == "00000" or item.name == "待确认" or item.topic == "待确认" for item in data.projects): @@ -164,9 +170,10 @@ def prepare(db, conversation: Conversation, kind: str, payload: dict, *, teacher scanned.append(match) if match["status"] != "recognized" or match["warnings"] or any(f["status"] != "recognized" for f in match["files"]): ambiguous.append(item.relative_dir) + unselected = any(not any(file.status != "ignored" for file in item.files) for item in data.projects) preview = {"course_version_id": data.course_version_id, "projects": scanned, "selection": payload["projects"], "ambiguous": ambiguous, - "requires_edit": bool(ambiguous) and not teacher_edited} + "requires_edit": (bool(ambiguous) or unselected) and not teacher_edited} elif kind == "course_ingest": require_scope(db, conversation, version_id=payload["course_version_id"]) preview = {"course_version_id": payload["course_version_id"], "templates": diff --git a/backend/wisegrade/api.py b/backend/wisegrade/api.py index 0c65344..7341ba2 100644 --- a/backend/wisegrade/api.py +++ b/backend/wisegrade/api.py @@ -3,7 +3,7 @@ from pathlib import Path from fastapi import APIRouter, Depends, HTTPException from fastapi.responses import FileResponse -from pydantic import BaseModel, Field +from pydantic import BaseModel, Field, ValidationInfo, field_validator, model_validator from sqlalchemy import func, select from sqlalchemy.orm import Session @@ -67,16 +67,37 @@ class FileConfirm(BaseModel): class ProjectConfirm(BaseModel): relative_dir: str - student_no: str = Field(pattern=r"^\d{5,20}$") + student_no: str = Field(pattern=r"^[0-9]{5,20}$") name: str = Field(min_length=1) topic: str = Field(min_length=1) direction: str = "" - files: list[FileConfirm] + files: list[FileConfirm] = Field(min_length=1) + + @field_validator("student_no", "name", "topic", "direction", mode="before") + @classmethod + def trim_identity(cls, value): + return value.strip() if isinstance(value, str) else value + + @model_validator(mode="after") + def validate_selection(self, info: ValidationInfo): + selected = [file for file in self.files if file.status != "ignored"] + if not selected: + # Only the chat preview builder supplies this context. HTTP imports, + # edited previews and final execution always use strict validation. + if (info.context or {}).get("allow_unselected_import_draft"): + return self + raise ValueError("至少选择1份资料") + categories = [file.category for file in selected] + if any(category not in CATEGORIES for category in categories): + raise ValueError("请选择有效材料类型") + if len(set(categories)) != len(categories): + raise ValueError("每类材料只允许选择1份，其余请设为忽略") + return self class ImportConfirm(BaseModel): course_version_id: int - projects: list[ProjectConfirm] + projects: list[ProjectConfirm] = Field(min_length=1) class BatchInput(BaseModel): diff --git a/backend/wisegrade/llm.py b/backend/wisegrade/llm.py index 7c32162..7691dbd 100644 --- a/backend/wisegrade/llm.py +++ b/backend/wisegrade/llm.py @@ -238,11 +238,19 @@ def validate_score(result: ScoreResult, criterion: dict, chunks: list[dict], *, # Empty identifiers are accepted only for pre-object legacy rubrics. if any(isinstance(value, dict) for value in criterion.get("required_evidence", [])): raise ModelError("模型证据缺少 target_id") - target = default_content or (targets[0] if targets else None) + associated = [by_target[key] for key in chunk.get("target_ids", []) if key in by_target] + if associated: + if len(associated) != 1: + raise ModelError(f"模型证据缺少 target_id：chunk_id={chunk['chunk_id']} 对应多个检索目标，请明确选择") + target = associated[0] + else: + target = default_content or (targets[0] if targets else None) if target is None: raise ModelError("模型证据未关联任何必备证据目标") if chunk.get("target_ids") and target["target_id"] not in chunk["target_ids"]: - raise ModelError("模型证据未关联该片段的检索目标") + allowed = ",".join(str(key) for key in chunk["target_ids"] if key in by_target) + raise ModelError(f"模型证据未关联该片段的检索目标：chunk_id={chunk['chunk_id']}，" + f"target_id={target['target_id']}；允许 target_id={allowed}") if not page_type_allowed(page_type, target["allowed_page_types"]): raise ModelError(f"模型证据来自不允许的页面类型（{page_type}），不能支持「{target['label']}」") if target["use"] == USE_CONTENT and not content_evidence_allowed(page_type): diff --git a/backend/wisegrade/rag.py b/backend/wisegrade/rag.py index 7de5191..88e5397 100644 --- a/backend/wisegrade/rag.py +++ b/backend/wisegrade/rag.py @@ -391,11 +391,16 @@ def retrieve_targets(db, *, course_version_id: int, project_id: int, targets: li substantive_only=target.get("use") != "structure", semantic_pool=SEMANTIC_POOL, keyword_pool=KEYWORD_POOL) hits = [normalize_hit(hit) for hit in (raw_hits or [])] - if target.get("use") == "content" and not hits and navigator is not None: + if target.get("use") == "content" and navigator is not None: # The contents list may only be used to navigate: its headings are matched # against headings that really occur in body text, and those sections and # their neighbouring pages are searched. The quotation still lands on body. - hits = [dict(hit, via="toc_navigation") for hit in (navigator(target) or [])] + # Semantic recall always returns neighbours, even for a short vague + # query. Supplement it with matching body sections rather than waiting + # for zero hits, and keep those sections ahead of incidental matches. + navigated = [normalize_hit(hit) for hit in (navigator(target) or [])] + seen_ids = {hit["chunk_id"] for hit in navigated} + hits = navigated + [hit for hit in hits if hit["chunk_id"] not in seen_ids] per_target[target["target_id"]] = hits for hit in hits: if hit["chunk_id"] not in seen: diff --git a/backend/wisegrade/scoring.py b/backend/wisegrade/scoring.py index c675f0a..ce8b240 100644 --- a/backend/wisegrade/scoring.py +++ b/backend/wisegrade/scoring.py @@ -1,4 +1,5 @@ from concurrent.futures import ThreadPoolExecutor, as_completed +import re from sqlalchemy import and_, or_, select, update @@ -97,14 +98,22 @@ def _body_navigator(db, *, course_version_id, project_id, category): return None def navigate(target): + if "body" not in target.get("allowed_page_types", ["body"]): + return [] label = normalize_heading(target.get("label") or "") if not label: return [] + # A rubric often says simply "计划", while the actual section is future + # work. Expand only this concept; headings still have to exist in body text. + labels = [label] + plan_terms = ("计划", "后续改进", "后续工作", "未来工作", "下一阶段", "下一步") + if any(term in label for term in plan_terms): + labels.extend(plan_terms) candidates = [] for heading, places in headings.items(): if not heading: continue - if heading.startswith(label) or label.startswith(heading) or label in heading or heading in label: + if any(term in heading or heading in term for term in labels): candidates.extend(places) if not candidates: return [] @@ -221,6 +230,8 @@ def _score_criterion(db, batch, project, criterion, client, *, retriever, multi_ prompt = (batch.course_snapshot["prompts"]["score"] + " 证据quote必须从所引用chunk_id的text逐字复制连续片段，保留原文标点和换行；page必须等于该片段的page；" "target_id必须取自required_evidence_targets。不要改写、缩写或拼接引文。" + "每条引文的 target_id 必须属于该片段的 target_ids；同一片段可对应多个目标，" + "请按实际内容选择，不要按证据列表顺序分配编号。" "不得把目录、摘要或章节标题当作内容证据；无法据此判断内容时把score设为null。") error = "" result = None @@ -233,7 +244,11 @@ def _score_criterion(db, batch, project, criterion, client, *, retriever, multi_ error = str(exc) if attempt: raise - prompt += f" 上次回答未通过校验（{error}）；请改用允许页面类型中的实质正文片段，并逐字复制quote。" + payload["validation_feedback"] = { + "error": error, + "instruction": "根据具体错误修正输出；引用目标只能从对应片段的 target_ids 选择，" + "quote须为该片段原文。没有足够依据时保留空分，不要编造引用。"} + prompt += " 上次回答未通过校验；请按 validation_feedback 修正，不要重复错误引用。" score = db.scalar(select(CriterionScore).where(CriterionScore.id == score.id) .with_for_update().execution_options(populate_existing=True)) if score.status == STATUS_REVIEWED or score.teacher_score is not None: @@ -244,13 +259,19 @@ def _score_criterion(db, batch, project, criterion, client, *, retriever, multi_ if result.score is None: # The retrieval found body pages but the model judged them insufficient. score.status, score.error = STATUS_NEEDS_REVIEW, "模型认为现有正文证据不足以判断该项，转教师复核。" + # Models normally write "误差分析（t1）：缺少…", not just the bare + # label. Attribute those explanations to their target instead of marking + # every target as missing. Unknown explanations remain conservative. declined = [target for target in targets if target["use"] == USE_CONTENT and - (not result.missing_evidence or target["label"] in result.missing_evidence)] + any(target["label"] in note or re.search( + r"(? Path: @@ -114,6 +115,7 @@ def scan(relative: str, *, root: Path | None = None): if item.is_dir() and not item.is_symlink() and has_materials(item)] for folder in folders: info = parse_student_folder(folder.name) + folder_identity = {key: info[key] for key in IDENTITY_LABELS} files = [] extracted = {} identity_values = {} @@ -124,6 +126,7 @@ def scan(relative: str, *, root: Path | None = None): if candidate.suffix.lower() not in SUPPORTED: continue error = "" + identity = {} try: text = first_text(candidate) category, status = classify(candidate, text) @@ -137,6 +140,7 @@ def scan(relative: str, *, root: Path | None = None): files.append({"relative_path": candidate.relative_to(base).as_posix(), "category": category, "status": status, "size": stat.st_size, "mtime": stat.st_mtime, "sha256": sha256(candidate), "mime_type": mimetypes.guess_type(candidate.name)[0] or "application/octet-stream", "error": error, + "detected_identity": identity, "first_text": text[:500] if not error else ""}) for category in CATEGORIES: matches = [file for file in files if file["category"] == category] @@ -145,17 +149,18 @@ def scan(relative: str, *, root: Path | None = None): file["status"] = "duplicate" for key, values in identity_values.items(): if len(values) > 1: - warnings.append(f"不同材料封面的{key}不一致，请核对后填写") + warnings.append(f"不同材料封面的{IDENTITY_LABELS[key]}不一致，请核对后填写") else: extracted[key] = next(iter(values)) for key, value in extracted.items(): if info.get(key) and value != info[key]: - warnings.append(f"封面{key}“{value}”与目录名“{info[key]}”不一致") + warnings.append(f"封面{IDENTITY_LABELS[key]}“{value}”与目录名“{info[key]}”不一致") elif not info.get(key): info[key] = value if warnings: info["status"] = "needs_confirmation" info.update({"relative_dir": folder.relative_to(base).as_posix(), "files": files, + "folder_identity": folder_identity, "detected_identity": extracted, "warnings": warnings, "missing": [category for category in CATEGORIES if not any(file["category"] == category for file in files)]}) projects.append(info) diff --git a/docs/IMPORT_MODEL_FIX_2026-10-14.md b/docs/IMPORT_MODEL_FIX_2026-10-14.md index a828d7b..fdd4b47 100644 --- a/docs/IMPORT_MODEL_FIX_2026-10-14.md +++ b/docs/IMPORT_MODEL_FIX_2026-10-14.md @@ -25,3 +25,15 @@ `test/示例/汇报PPT.pdf` 首页实际显示“课程导学／自动化专业综合实践”。应在评分前确认它是否属于学生汇报稿，而不是课程说明。本次只导入和索引该示例资料包，未创建或启动它的新评分批次。 运行证据保存在工作区 `.wisegrade-testing/docker-preview/`，包括 `import-model-live.json`、`import-test-sample-2026-10-14.json`、浏览器截图和构建/部署日志。实际入库与批次重试的最终状态见该目录 `import-model-final-2026-10-14.json`。 + +## 后续：确认导入的校验与封面来源 + +按钮被禁用的原因原来隐藏在网页布尔表达式中；扫描只返回合并后的身份，冲突时清空字段，但未提供逐文件来源，忽略文件后也不重新计算。这使用户无法从页面判断如何完成合法导入。 + +- `d7bd152`（2 个文件）：扫描返回每份文件的封面身份和原目录身份，警告使用中文字段名；读取失败的文件不沿用上一份身份。 +- `b24930b`（3 个文件）：显示具体阻碍原因和逐文件身份；只根据当前选中文件重算身份和冲突，忽略异项目文件后可重新预填。保留人工修改，提供“使用所选资料的封面信息”入口。材料类别未知、同类多份或没有选中材料时直接提示。 +- `36bf473`（2 个文件）：API 在写入前统一去除身份字段首尾空白，要求 5–20 位 ASCII 数字学号，拒绝空白姓名/选题、空项目、全部忽略及无效/重复材料类型。忽略的未知类型文件不阻碍合法选择。 +- 后端隔离测试 **74 passed**，网页测试 **17 passed**，TypeScript/Vite 构建通过；新增用例在修复前失败。API、网页和空闲 worker 已更新，本轮没有新增数据库迁移。 +- 实际网页无接口模拟验证：输入 `1` 显示学号规则；忽略异项目 PPT 并采用剩余报告封面信息后，身份冲突清除且确认按钮可用。没有提交真实导入或调用评分模型。空项目 API 请求返回 422。证据见工作区 `docker-preview/import-root-live.json` 和对应截图。 + +本轮修复不放宽学号格式要求，也不把不同学生的材料自动改成同一身份。已有项目若先前被人工填写为测试身份，需要核对所选资料并重新确认导入才能更新数据库中的身份。 diff --git a/docs/SCORING_REVIEW_FIX_2026-10-17.md b/docs/SCORING_REVIEW_FIX_2026-10-17.md new file mode 100644 index 0000000..99e7f1a --- /dev/null +++ b/docs/SCORING_REVIEW_FIX_2026-10-17.md @@ -0,0 +1,46 @@ +# 评分待复核与引用校验修复（2026-10-17） + +## 触发场景与根因 + +- 批次 #8 的 `stage3.plan` 目标为“计划”。实际阶段报告3的 PDF 第14页有“4.2 后续改进方向”正文，但旧检索选出的14个片段没有这一页。短查询仍会得到语义近邻，导致原先仅在零命中时启用的章节导航不执行；“计划”与“后续改进方向”也没有按同一概念匹配。 +- `stage2.experiment` 记录的错误为“模型证据未关联该片段的检索目标”。模型提示只要求使用规则中的目标编号，没有明确要求同时属于片段的 `target_ids`；重试也未指出具体可用的片段与目标关联。此外，旧字符串规则中省略编号时会错误地默认绑定第一个内容目标。 +- 页面将所有 `failed` 统一标成“模型调用失败”，重试提示也一律指向网络；批次进度只显示待复核数量，没有显示失败数量。100% 表示处理结束，并不表示所有评分都成功。 + +## 修复 + +- 内容检索始终补充匹配的真实正文章节，兼容计划/后续改进/后续工作等表述，保留14片段预算、项目/材料范围、正文许可与引用校验。 +- 只有旧规则省略编号且片段唯一关联某个目标时才补全；多目标歧义、显式错误编号、目录冒充内容证据仍然拒绝。重试带具体校验错误及允许编号，不通过忽略错误来获得分数。 +- 成功重试清理旧证据缺口；教师复核记录仍按原逻辑保护。 +- 页面区分输出校验失败、连接失败和一般评分失败，分别显示待复核/失败/材料异常数量，并解释处理完成不等于全部已评分。 + +## 验证 + +- 新增/扩展的导航、目标关联、重试反馈、旧缺口清理和页面测试均先观察到失败，再验证修复后通过。 +- 后端4个相关测试文件：81 passed（隔离 SQLite 业务测试，不代表全量 PostgreSQL 测试通过）。 +- Playwright：19 passed；TypeScript/Vite 构建通过。 +- Docker API/Worker/Web 已更新并健康；更新前无运行中批次，队列和执行中任务均为0。 +- 生产数据只读 PostgreSQL + 真实离线 BGE-M3 检索：相同规则修复前未选第14页，修复后选中该页正文，片段数量仍为14。 +- 真实网页验证：批次 #8 显示“待复核2”“评分失败1”和未评分项提示；失败行显示“模型输出校验失败”；复核弹窗解释证据尚不足以自动评分。无页面错误、无业务写请求、无实际模型调用。 + +## 边界与使用 + +历史批次分数未自动改写。刷新网页后可重试 #8 的未成功项；已成功评分和教师复核项不重新评分。此次未实际调用收费模型重新给学生打分，因此不声称重试一定消除全部待复核项。 + +课程导学PPT缺少学生项目汇报内容、报告安全论证不足等情况仍应提示复核。修复检索遗漏和校验反馈，不将证据不足强制变成有效分数。 + +## 后续优化：重试开销、人工复核保护与快速筛选 + +- `050aa4e`：复现“教师保存零分后模型超时”的交错场景，修复异常路径覆盖人工复核状态的问题。异常处理重新加锁读取评分，保留教师分与评语。 +- `14cef5c`：多学生批次重试时，已全部评分完成的学生直接跳过模型初始化、事实抽取和一致性检查，批次进度仍正确增加；有待复核项的学生仍正常处理。 +- `5a6e5d4`：模型返回“误差分析（t1）：缺少……”等描述时，证据缺口归到对应目标；无法识别目标的说明仍保守保留所有相关内容目标，分数继续为空。 +- `4acc756`：评分页增加全部/待处理/待教师复核/失败/材料异常筛选，以及已评分、教师已复核、待处理数量汇总。仅在浏览器内筛选，不重新评分；零分正常保留，无结果时说明当前筛选下没有评分项。 +- `a5cb40f`：聊天导入遇到重复文件时允许生成待编辑草稿；所有资料都被忽略的草稿强制要求编辑，正式导入与编辑后预览仍严格校验。非法编辑返回400，未修改草稿不能执行（409）。 +- `f7db0a4`：修正扩展测试暴露的旧假模型缺目标编号、迁移头版本和证据政策版本断言，保留真实引用、迁移数据与证据权限的检查。 + +验证： + +- Docker 独立 PostgreSQL/Redis 测试服务内执行完整非 `integration` 后端套件：146 passed，2 deselected；未运行在线模型调用测试。测试服务已停止，未使用正式数据库写入测试数据。 +- Playwright：20 passed；TypeScript/Vite 构建通过。 +- 已更新 API/Worker/Web，服务健康。更新前确认无排队和执行中的任务。 +- 实际网页：批次8显示22项已评分，筛选待处理得到0项与明确空态；批次7筛选出1项待复核。无页面异常、无业务写请求、无模型调用。 +- 批次8完整评分响应在更新前后规范化 JSON 的 SHA-256 一致，22项成绩未被改写。本轮无需重新运行已有评分。 diff --git a/frontend/src/App.tsx b/frontend/src/App.tsx index 0bc2a96..8cbb27a 100644 --- a/frontend/src/App.tsx +++ b/frontend/src/App.tsx @@ -5,6 +5,7 @@ import { DashboardOutlined, BookOutlined, FileSearchOutlined, FolderOpenOutlined import { api, post, put } from './api'; import { Batches } from './Batches'; import { projectLabel, projectsInBatch } from './batchProjects'; +import { fillImportIdentity, importErrors, reviewImport } from './importReview'; const Chat = lazy(() => import('./Chat')); const { Header, Sider, Content } = Layout; @@ -30,8 +31,14 @@ const evidencePlace = (evidence: any) => `第 ${evidence.page} ${isSlideEvidence const evidencePreviewLabel = (evidence: any) => isSlideEvidence(evidence) ? '查看该张幻灯片的完整图像' : '查看该页的完整图像'; const pageTypeLabels: Record = { cover: '封面', abstract: '摘要', toc: '目录', body: '正文', appendix: '附录', references: '参考文献', unknown: '识别不确定', mixed: '目录与正文混合页' }; const pageTypeLabel = (value: string) => pageTypeLabels[value] || '未分类页面'; -const statusLabels: Record = { pending: '待评分', scored: '模型已评分', reviewed: '教师已复核', needs_review: '待教师复核', unscorable: '材料缺失或解析失败', failed: '模型调用失败' }; +const statusLabels: Record = { pending: '待评分', scored: '模型已评分', reviewed: '教师已复核', needs_review: '待教师复核', unscorable: '材料缺失或解析失败', failed: '评分失败' }; const statusLabel = (value: string) => statusLabels[value] || value; +const scoreStatusLabel = (row: any) => { + if (row.status !== 'failed') return statusLabel(row.status); + if (/模型证据|target_id|评分项或满分|JSON校验/.test(row.error || '')) return '模型输出校验失败'; + if (/ConnectError|Timeout|Network|连接|超时|HTTP/i.test(row.error || '')) return '模型连接失败'; + return '评分失败'; +}; const statusColor = (value: string) => ({ scored: 'green', reviewed: 'green', needs_review: 'gold', unscorable: 'red', failed: 'red' }[value] || 'blue'); const batchStatusLabels: Record = { created: '待启动', queued: '排队中', running: '运行中', completed: '已完成', review_required: '已完成，有待复核项', partial: '部分完成', failed: '失败' }; const batchStatusLabel = (value: string) => batchStatusLabels[value] || value; @@ -51,7 +58,7 @@ function JsonField({ value, onChange, disabled, rows = 4, resetKey }: { value: a function Dashboard() { const [data, setData] = useState({}); useEffect(() => { load('/dashboard', setData); }, []); - return <>工作概览先确认课程版本与规则，再扫描挂载目录，创建评分批次并逐项复核。{[['课程', data.courses], ['学生项目', data.projects], ['评分批次', data.batches], ['待处理评分项', data.needs_attention], ['待教师复核', data.needs_review]].map(([label, count]) => )}; + return <>工作概览先确认课程版本与规则，再扫描挂载目录，创建评分批次并逐项复核。{[['课程', data.courses], ['学生项目', data.projects], ['评分批次', data.batches], ['待处理评分项', data.needs_attention], ['待教师复核', data.needs_review]].map(([label, count]) => )}; } function Courses() { @@ -103,14 +110,26 @@ function Import({ openImported, scoreImported }: { openImported: (ids: number[]) useEffect(() => { load('/courses', setCourses); }, []); useEffect(() => { setScanData(undefined); setImported(undefined); load(`/storage/directories?relative=${encodeURIComponent(path)}`, setDirs); }, [path]); useEffect(() => { if (courseId) load(`/courses/${courseId}/versions`, setVersions); }, [courseId]); - const editProject = (index: number, key: string, value: any) => { const next = structuredClone(scanData); next.projects[index][key] = value; setScanData(next); }; - const editFile = (projectIndex: number, fileIndex: number, key: string, value: any) => { const next = structuredClone(scanData); next.projects[projectIndex].files[fileIndex][key] = value; setScanData(next); }; + const errors = importErrors(scanData?.projects, versionId); + const editProject = (index: number, key: string, value: any) => { + const next = structuredClone(scanData), project = next.projects[index]; + project[key] = value; + if (['student_no', 'name', 'topic'].includes(key)) project.identity_edits = { ...project.identity_edits, [key]: true }; + setScanData(next); + }; + const editFile = (projectIndex: number, fileIndex: number, key: string, value: any) => { + const next = structuredClone(scanData), project = next.projects[projectIndex]; + project.files[fileIndex][key] = value; fillImportIdentity(project); setScanData(next); + }; + const useCover = (index: number) => { + const next = structuredClone(scanData); fillImportIdentity(next.projects[index], true); setScanData(next); + }; const confirm = async () => { - if (!versionId || !scanData?.projects?.length) return; + if (errors.length || importing) { message.error(errors[0] || '正在导入，请稍候'); return; } setImporting(true); try { - const projects = scanData.projects.map((p: any) => ({ relative_dir: p.relative_dir, student_no: p.student_no, - name: p.name, topic: p.topic, direction: p.direction || '', + const projects = scanData.projects.map((p: any) => ({ relative_dir: p.relative_dir, student_no: p.student_no.trim(), + name: p.name.trim(), topic: p.topic.trim(), direction: (p.direction || '').trim(), files: p.files.map((f: any) => ({ relative_path: f.relative_path, category: f.category, status: f.status === 'ignored' ? 'ignored' : 'recognized', sha256: f.sha256 })) })); const result = await post('/import/confirm', { course_version_id: versionId, projects }); @@ -119,14 +138,60 @@ function Import({ openImported, scoreImported }: { openImported: (ids: number[]) } catch (error) { fail(error); } finally { setImporting(false); } }; - return <>挂载目录扫描与导入确认浏览 /data/inbox 下的目录，可选择包含学生子文件夹的总目录，也可直接进入放有 PDF、DOCX、PPTX 的资料目录扫描。封面信息会预填，请核对学号、姓名和选题后确认导入。{path}
{dirs.map(dir => )}
+ const projectIndex = (project: any) => scanData.projects.findIndex((item: any) => item.relative_dir === project.relative_dir); + const identityInput = (project: any, key: string, label: string) => editProject(projectIndex(project), key, event.target.value)} />; + return <>挂载目录扫描与导入确认 + 浏览 /data/inbox 下的目录，可选择学生总目录或直接存放资料的目录。请核对封面身份；不属于当前项目的文件可设为“忽略”。 + {path} + +
{dirs.map(dir => )}
{imported && imported.projectIds.length > 0 && } />} - {scanData && <>{!scanData.projects.length && } v.published), v => `v${v.version} ${v.title}`)} value={versionId} onChange={setVersionId} /> editProject(scanData.projects.findIndex((x: any) => x.relative_dir === p.relative_dir), 'student_no', e.target.value)} /> }, { title: '姓名', render: (_, p, index) => editProject(scanData.projects.findIndex((x: any) => x.relative_dir === p.relative_dir), 'name', e.target.value)} /> }, { title: '选题', render: (_, p, index) => editProject(scanData.projects.findIndex((x: any) => x.relative_dir === p.relative_dir), 'topic', e.target.value)} /> }, { title: '选题方向', render: (_, p) => editProject(scanData.projects.findIndex((x: any) => x.relative_dir === p.relative_dir), 'direction', e.target.value)} /> }, { title: '待处理', render: (_, p) => {p.status !== 'recognized' && 身份信息需确认}{p.missing.map((c: string) => 缺{categories[c]})}{p.warnings?.map((w: string) => {w})} }]} - expandable={{ defaultExpandAllRows: true, expandedRowRender: (p: any) => { const pi = scanData.projects.findIndex((item: any) => item.relative_dir === p.relative_dir); return size="small" rowKey="relative_path" pagination={false} dataSource={p.files} columns={[{ title: '文件', dataIndex: 'relative_path' }, { title: '类型', render: (_, f, fi) => editFile(pi, fi, 'status', value)} /> }, { title: '扫描结果', render: (_, f) => {f.status} }, { title: '异常', dataIndex: 'error' }]} />; } }} />} + {scanData && <>{errors.length > 0 && {errors.map((error, index) =>
  • {error}
    • )}} />} + + v.published), v => `v${v.version} ${v.title}`)} value={versionId} onChange={setVersionId} /> + + +
    identityInput(project, 'student_no', '学号') }, + { title: '姓名', render: (_, project) => identityInput(project, 'name', '姓名') }, + { title: '选题', render: (_, project) => identityInput(project, 'topic', '选题') }, + { title: '选题方向', render: (_, project) => editProject(projectIndex(project), 'direction', event.target.value)} /> }, + { title: '待处理', render: (_, project) => { const review = reviewImport(project); return + {review.errors.length > 0 ? 请修正导入信息 : 导入信息已填写} + {review.missing.map(category => 缺{categories[category]})} + {review.warnings.map(warning => {warning})} + {review.metadata && Object.values(review.identity).some(Boolean) && } + ; } }, + ]} expandable={{ defaultExpandAllRows: true, expandedRowRender: (project: any) => { + const pi = projectIndex(project); return size="small" rowKey="relative_path" pagination={false} dataSource={project.files} columns={[ + { title: '文件', dataIndex: 'relative_path' }, + { title: '封面身份', render: (_, file) => + {file.detected_identity?.name ? `${file.detected_identity.name}（${file.detected_identity.student_no || '未识别学号'}）` : '首页未识别身份'} + {file.detected_identity?.topic || ''} }, + { title: '类型', render: (_, file, fi) => editFile(pi, fi, 'status', value)} /> }, + { title: '扫描结果', render: (_, file) => + {({ recognized: '已识别', needs_confirmation: '需确认类型', duplicate: '同类多份', unparseable: '无法读取', ignored: '已忽略' } as Record)[file.status] || file.status} }, + { title: '异常', dataIndex: 'error' }, + ]} />; + } }} /> + 封面冲突请核对上表来源。允许缺少某类材料，相关评分项会标记为待处理。人工填写的信息会保留；点击“使用所选资料的封面信息”可重新采用识别值。 + } ; } @@ -140,6 +205,7 @@ function Projects({ openScores }: { openScores: (id: number) => void }) { function Scores({ batchId, projectId }: { batchId?: number; projectId?: number }) { const [batches, setBatches] = useState([]), [projects, setProjects] = useState([]), [batch, setBatch] = useState(batchId), [project, setProject] = useState(projectId), [result, setResult] = useState(), [review, setReview] = useState(), [revisions, setRevisions] = useState([]); + const [scoreFilter, setScoreFilter] = useState('all'); const [form] = Form.useForm(); const [sourcesReady, setSourcesReady] = useState(false); useEffect(() => { @@ -154,6 +220,7 @@ function Scores({ batchId, projectId }: { batchId?: number; projectId?: number } const scopedProjects = projectsInBatch(selectedBatch, projects); useEffect(() => { setResult(undefined); setReview(undefined); setRevisions([]); + setScoreFilter('all'); if (!sourcesReady || !selectedBatch) return; if (!scopedProjects.some(item => item.id === project)) { setProject(scopedProjects.length === 1 ? scopedProjects[0].id : undefined); @@ -165,6 +232,12 @@ function Scores({ batchId, projectId }: { batchId?: number; projectId?: number } return () => { cancelled = true; }; }, [batch, project, batches, projects, sourcesReady]); const refresh = () => batch && project && load(`/projects/${project}/scores?batch_id=${batch}`, setResult); + const scoreRows: any[] = result?.rows || []; + const count = (status: string) => scoreRows.filter(row => row.status === status).length; + const unfinished = (row: any) => !['scored', 'reviewed'].includes(row.status); + const attentionCount = scoreRows.filter(unfinished).length; + const filteredScores = scoreRows.filter(row => scoreFilter === 'all' || + (scoreFilter === 'attention' ? unfinished(row) : row.status === scoreFilter)); return <>逐项评分与教师复核
    row.criterion.criterion_id} dataSource={result.rows} pagination={false} columns={[{ title: '评分项', render: (_, r) => <>{categories[r.criterion.category] || r.criterion.category}{r.criterion.name} }, { title: '模型原始分', render: (_, r) => `${r.model_score ?? '—'} / ${r.criterion.max_score}` }, { title: '教师分', render: (_, r) => r.teacher_score ?? '—' }, { title: '最终分', render: (_, r) => r.final_score ?? '待处理' }, { title: '状态 / 失败原因', render: (_, r) => {statusLabel(r.status)}{r.error && {r.error}} }, { title: '操作', render: (_, r) => }]} />} - setReview(undefined)} onOk={async () => { try { const values = await form.validateFields(); await put(`/criterion-scores/${review.score_id}/review`, values); setReview(undefined); refresh(); message.success('复核记录已保存'); } catch (e) { fail(e); } }} okText="保存教师复核" okButtonProps={{ disabled: !review?.score_id }}>{review && <>{review.status === 'needs_review' && gap.target_label).join('、') || review.missing_evidence?.join('、') || '见缺失证据'}。系统未给出模型分，也没有按低分处理；请核对材料后直接填写教师评分。`} />} `${gap.target_label}（检索范围：${(gap.searched_page_types || []).map(pageTypeLabel).join('、') || '正文'}）`).join('；') || '无' }, { key: 'suggestion', label: '改进建议', children: review.suggestion || '无' }, { key: 'error', label: '异常', children: review.error || '无' }]} />{review.evidence?.map((ev: any, index: number) => {ev.document} · {evidencePlace(ev)}{pageTypeLabel(ev.page_type)}{ev.historical && 历史评分引用}{ev.target_label && {ev.target_label}}{ev.quote}{ev.url && 查看页面文字}{ev.image_url && {evidencePreviewLabel(ev)}（核对引文）})}
    {review.score_id && }{revisions.length > 0 && 
    {JSON.stringify(revisions, null, 2)}
    }}     + {result && +
    row.criterion.criterion_id} dataSource={filteredScores} locale={{ emptyText: '当前筛选下没有评分项' }} pagination={false} columns={[{ title: '评分项', render: (_, r) => <>{categories[r.criterion.category] || r.criterion.category}{r.criterion.name} }, { title: '模型原始分', render: (_, r) => `${r.model_score ?? '—'} / ${r.criterion.max_score}` }, { title: '教师分', render: (_, r) => r.teacher_score ?? '—' }, { title: '最终分', render: (_, r) => r.final_score ?? '待处理' }, { title: '状态 / 失败原因', render: (_, r) => {scoreStatusLabel(r)}{r.error && {r.error}} }, { title: '操作', render: (_, r) => }]} />} + setReview(undefined)} onOk={async () => { try { const values = await form.validateFields(); await put(`/criterion-scores/${review.score_id}/review`, values); setReview(undefined); refresh(); message.success('复核记录已保存'); } catch (e) { fail(e); } }} okText="保存教师复核" okButtonProps={{ disabled: !review?.score_id }}>{review && <>{review.status === 'needs_review' && gap.target_label).join('、') || review.missing_evidence?.join('、') || '见缺失证据'}。系统未给出模型分，也没有按低分处理；请核对下方证据，补充或更正材料后重试，或直接填写教师评分。`} />} `${gap.target_label}（检索范围：${(gap.searched_page_types || []).map(pageTypeLabel).join('、') || '正文'}）`).join('；') || '无' }, { key: 'suggestion', label: '改进建议', children: review.suggestion || '无' }, { key: 'error', label: '异常', children: review.error || '无' }]} />{review.evidence?.map((ev: any, index: number) => {ev.document} · {evidencePlace(ev)}{pageTypeLabel(ev.page_type)}{ev.historical && 历史评分引用}{ev.target_label && {ev.target_label}}{ev.quote}{ev.url && 查看页面文字}{ev.image_url && {evidencePreviewLabel(ev)}（核对引文）})}
    {review.score_id && }{revisions.length > 0 && 
    {JSON.stringify(revisions, null, 2)}
    }}     ; } diff --git a/frontend/src/Batches.tsx b/frontend/src/Batches.tsx index 1e2d304..440beb0 100644 --- a/frontend/src/Batches.tsx +++ b/frontend/src/Batches.tsx @@ -148,10 +148,15 @@ export function Batches({ openBatch }: { openBatch: (id: number, projectId?: num { title: '状态', width: 140, render: (_, batch) => {labels[batch.status] || batch.status} {batch.needs_review > 0 && 待复核 {batch.needs_review}} + {batch.failed > 0 && 评分失败 {batch.failed}} + {batch.unscorable > 0 && 材料异常 {batch.unscorable}} }, - { title: '进度', width: 160, render: (_, batch) => <> + { title: '处理进度', width: 190, render: (_, batch) => <> {batch.progress_done} / {batch.progress_total} 项 + {batch.progress_total > 0 && batch.progress_done >= batch.progress_total && + (batch.needs_review > 0 || batch.failed > 0 || batch.unscorable > 0) && + 处理结束，仍有未评分项} }, { title: '异常', dataIndex: 'error', width: 150 }, { title: '操作', width: 180, render: (_, batch) => diff --git a/frontend/src/importReview.ts b/frontend/src/importReview.ts new file mode 100644 index 0000000..5ef134d --- /dev/null +++ b/frontend/src/importReview.ts @@ -0,0 +1,66 @@ +const fields = ['student_no', 'name', 'topic'] as const; +type IdentityField = typeof fields[number]; +const labels = { student_no: '学号', name: '姓名', topic: '选题' }; +const categories: Record = { stage1: '阶段报告1', stage2: '阶段报告2', stage3: '阶段报告3', + final_report: '最终报告', presentation: '汇报PPT/PDF' }; +const text = (value: unknown) => String(value ?? '').trim(); +const fileName = (file: any) => text(file.relative_path).split('/').pop() || '文件'; + +export function reviewImport(project: any) { + const selected = (project.files || []).filter((file: any) => file.status !== 'ignored'); + const metadata = (project.files || []).some((file: any) => file.detected_identity != null); + const identity: Record = { student_no: '', name: '', topic: '' }; + const warnings: string[] = []; + for (const field of fields) { + const sources = new Map(); + for (const file of selected) { + const value = text(file.detected_identity?.[field]); + if (value) sources.set(value, [...(sources.get(value) || []), fileName(file)]); + } + const folderValue = text(project.folder_identity?.[field]); + const values = [...sources.keys()]; + identity[field] = folderValue || (values.length === 1 ? values[0] : ''); + if (values.length > 1) { + warnings.push(`${labels[field]}冲突：${values.map(value => `${value}（${sources.get(value)!.join('、')}）`).join('；')}`); + } else if (folderValue && values.length === 1 && folderValue !== values[0]) { + warnings.push(`${labels[field]}与目录不一致：封面为${values[0]}（${sources.get(values[0])!.join('、')}），目录为${folderValue}`); + } + } + if (!metadata) warnings.push(...(project.warnings || [])); + + const errors: string[] = []; + if (!/^[0-9]{5,20}$/.test(text(project.student_no))) errors.push('学号必须为5–20位数字'); + if (!text(project.name)) errors.push('请填写姓名'); + if (!text(project.topic)) errors.push('请填写选题'); + if (!selected.length) errors.push('至少选择1份资料'); + const counts = new Map(); + for (const file of selected) { + if (!categories[file.category]) errors.push(`${fileName(file)}：请选择材料类型`); + else counts.set(file.category, (counts.get(file.category) || 0) + 1); + if (file.error) errors.push(`${fileName(file)}：无法读取，请修复文件或设为忽略`); + } + for (const [category, count] of counts) { + if (count > 1) errors.push(`${categories[category]}选了${count}份，请只保留1份，其余设为忽略`); + } + const missing = Object.keys(categories).filter(category => !selected.some((file: any) => file.category === category)); + return { selected, metadata, identity, warnings, errors, missing }; +} + +export function fillImportIdentity(project: any, force = false) { + const review = reviewImport(project); + if (!review.metadata) return project; + for (const field of fields) { + if (force || !project.identity_edits?.[field]) project[field] = review.identity[field]; + } + if (force) project.identity_edits = {}; + return project; +} + +export function importErrors(projects: any[] | undefined, versionId: number | undefined) { + const errors = versionId ? [] : ['请选择已发布课程版本']; + if (!projects?.length) return [...errors, '当前目录没有PDF、DOCX或PPTX资料，请重新扫描资料目录']; + for (const project of projects) { + errors.push(...reviewImport(project).errors.map(error => `${project.relative_dir}：${error}`)); + } + return errors; +} diff --git a/frontend/tests/batches.spec.ts b/frontend/tests/batches.spec.ts index 0e6eb37..18f404a 100644 --- a/frontend/tests/batches.spec.ts +++ b/frontend/tests/batches.spec.ts @@ -149,6 +149,72 @@ test('failed scoring shows the network error and retry guidance on the page', as await expect(page.getByText(/已经成功的评分和教师复核会保留/)).toBeVisible(); }); +test('batch progress exposes failures separately from teacher review', async ({ page }) => { + await setup(page); + await page.route('**/api/scoring-batches', route => route.fulfill({ json: [{ + id: 71, course_version_id: 1, project_ids: [11], status: 'partial', + progress_done: 22, progress_total: 22, needs_review: 2, failed: 1, unscorable: 0, + }] })); + await page.goto('/?page=batches'); + const row = page.getByRole('row').filter({ hasText: '#71' }); + await expect(row).toContainText('评分失败 1'); + await expect(row).toContainText('待复核 2'); + await expect(row).toContainText('处理结束，仍有未评分项'); +}); + +test('output validation and insufficient evidence are not presented as network failures', async ({ page }) => { + await setup(page); + await page.route('**/api/projects/11/scores?*', route => route.fulfill({ json: { + student_no: '9000000001', name: '学生甲', topic: '项目', totals: { overall: null }, facts: {}, + rows: [ + { criterion: { criterion_id: 'stage2.experiment', category: 'stage2', name: '实验方案', max_score: 20 }, + status: 'failed', error: '模型证据未关联该片段的检索目标', evidence: [] }, + { criterion: { criterion_id: 'stage3.plan', category: 'stage3', name: '下一阶段计划', max_score: 10 }, + status: 'needs_review', error: '模型认为现有正文证据不足以判断该项，转教师复核。', + evidence: [], missing_evidence: ['计划'], evidence_gaps: [], model_comment: '当前片段不足。' }, + ], + } })); + await page.goto('/?page=scores&batch_id=71&project_id=11'); + await expect(page.getByText('模型输出校验失败', { exact: true })).toBeVisible(); + await expect(page.getByText('模型调用失败', { exact: true })).toHaveCount(0); + await expect(page.getByText(/输出校验失败可重试/)).toBeVisible(); + await page.getByRole('row').filter({ hasText: '下一阶段计划' }).getByRole('button', { name: '查看证据/复核' }).click(); + await expect(page.getByText('本项证据不足以自动评分，需要教师复核', { exact: true })).toBeVisible(); + await expect(page.getByText(/缺少或尚不足以支撑判断的目标：计划/)).toBeVisible(); +}); + +test('score filters isolate actionable rows without hiding zero grades or changing data', async ({ page }) => { + await setup(page); + const writes: string[] = []; + page.on('request', request => { if (request.method() !== 'GET') writes.push(request.url()); }); + await page.route('**/api/projects/11/scores?*', route => route.fulfill({ json: { + student_no: '9000000001', name: '学生甲', topic: '项目', totals: { overall: null }, facts: {}, + rows: ['scored', 'reviewed', 'needs_review', 'failed', 'unscorable'].map((status, i) => ({ + criterion: { criterion_id: `stage1.${i}`, category: 'stage1', name: `测试项目${i}`, max_score: 20 }, + status, model_score: status === 'scored' ? 0 : null, teacher_score: status === 'reviewed' ? 0 : null, + final_score: ['scored', 'reviewed'].includes(status) ? 0 : null, + error: status === 'failed' ? '模型证据未能匹配检索原文' : '', evidence: [], + })), + } })); + await page.goto('/?page=scores&batch_id=71&project_id=11'); + await expect(page.getByText('已评分 1 项 · 教师已复核 1 项 · 待处理 3 项', { exact: true })).toBeVisible(); + const rows = page.getByRole('row').filter({ hasText: '测试项目' }); + await expect(rows).toHaveCount(5); + for (const [option, count, name] of [ + ['仅待处理（3）', 3, '测试项目2'], ['仅待教师复核（1）', 1, '测试项目2'], + ['仅评分失败（1）', 1, '测试项目3'], ['仅材料异常（1）', 1, '测试项目4'], + ['全部评分项（5）', 5, '测试项目0'], + ] as const) { + await page.keyboard.press('Escape'); + await expect(page.locator('.ant-select-dropdown:visible')).toHaveCount(0); + await choose(page, '评分项筛选', option); + await expect(rows).toHaveCount(count); + await expect(page.getByRole('row').filter({ hasText: name })).toBeVisible(); + } + await expect(page.getByRole('row').filter({ hasText: '测试项目0' })).toContainText('0 / 20'); + expect(writes).toEqual([]); +}); + test('a slow project list keeps the student specified in a multi-project batch link', async ({ page }) => { const state = await setup(page); await page.route('**/api/scoring-batches', route => route.fulfill({ json: [{ diff --git a/frontend/tests/import.spec.ts b/frontend/tests/import.spec.ts new file mode 100644 index 0000000..4b83550 --- /dev/null +++ b/frontend/tests/import.spec.ts @@ -0,0 +1,98 @@ +import { expect, test, type Page } from '@playwright/test'; + +const identity = { student_no: '9000000001', name: '学生甲', topic: '机器人项目' }; +const mixed = () => ({ relative_dir: 'test/student', student_no: '', name: '', topic: '', + status: 'needs_confirmation', folder_identity: {}, detected_identity: {}, missing: [], + warnings: ['不同材料封面的学号不一致，请核对后填写'], files: [ + { relative_path: 'test/student/最终报告.pdf', category: 'final_report', status: 'recognized', + sha256: 'a'.repeat(64), detected_identity: identity }, + { relative_path: 'test/student/汇报PPT.pdf', category: 'presentation', status: 'needs_confirmation', + sha256: 'b'.repeat(64), detected_identity: { student_no: '9000000002', name: '学生乙', topic: '另一项目' } }, + ] }); + +async function setup(page: Page, project: any) { + const imports: any[] = []; + await page.route('**/api/**', async route => { + const request = route.request(); + const path = new URL(request.url()).pathname.replace(/^\/api/, ''); + let data: any = []; + if (path === '/courses') data = [{ id: 1, name: '课程甲' }]; + if (path === '/courses/1/versions') data = [{ id: 1, published: true, version: 1, title: '初始版本' }]; + if (path === '/import/scan') data = { projects: [project] }; + if (path === '/import/confirm') { imports.push(request.postDataJSON()); data = { project_ids: [12] }; } + await route.fulfill({ json: data }); + }); + await page.goto('/?page=import'); + await page.getByRole('button', { name: '扫描此目录', exact: true }).click(); + await choose(page, '导入课程', '课程甲'); + await choose(page, '导入课程版本', /初始版本/); + return imports; +} + +async function choose(page: Page, label: string, option: string | RegExp) { + await page.keyboard.press('Escape'); + await expect(page.locator('.ant-select-dropdown:visible')).toHaveCount(0); + const input = page.getByRole('combobox', { name: label, exact: true }); + await input.focus(); await input.press('ArrowDown'); + const listId = await input.getAttribute('aria-controls'); + const dropdown = page.locator('.ant-select-dropdown').filter({ has: page.locator(`[id="${listId}"]`) }); + await expect(dropdown).toBeVisible(); + await dropdown.getByText(option, { exact: typeof option === 'string' }).click(); +} + +test('invalid student number explains why confirm is disabled and recovers after editing', async ({ page }) => { + const project = { ...mixed(), student_no: '1', name: '1', topic: '1' }; + delete (project as any).folder_identity; + for (const file of project.files) delete (file as any).detected_identity; + await setup(page, project); + await expect(page.getByText(/学号必须为5–20位数字/)).toBeVisible(); + await expect(page.getByRole('button', { name: '确认导入', exact: true })).toBeDisabled(); + await page.getByRole('textbox', { name: '学号 test/student', exact: true }).fill('9000000001'); + await expect(page.getByRole('button', { name: '确认导入', exact: true })).toBeEnabled(); +}); + +test('ignoring a foreign PPT recalculates identity and submits the intended project', async ({ page }) => { + const imports = await setup(page, mixed()); + await expect(page.getByText('学生乙（9000000002）', { exact: true })).toBeVisible(); + await expect(page.getByText(/学号冲突：/)).toBeVisible(); + await choose(page, '处理 test/student/汇报PPT.pdf', '忽略'); + await expect(page.getByRole('textbox', { name: '学号 test/student', exact: true })).toHaveValue(identity.student_no); + await expect(page.getByRole('textbox', { name: '姓名 test/student', exact: true })).toHaveValue(identity.name); + await expect(page.getByText(/学号冲突：/)).toHaveCount(0); + await expect(page.getByRole('button', { name: '确认导入', exact: true })).toBeEnabled(); + await page.getByRole('button', { name: '确认导入', exact: true }).click(); + await expect.poll(() => imports.length).toBe(1); + expect(imports[0].projects[0].student_no).toBe(identity.student_no); + expect(imports[0].projects[0].files[1].status).toBe('ignored'); +}); + +test('file selection preserves manual corrections until the user chooses cover values', async ({ page }) => { + await setup(page, mixed()); + const name = page.getByRole('textbox', { name: '姓名 test/student', exact: true }); + await name.fill('人工修正姓名'); + await choose(page, '处理 test/student/汇报PPT.pdf', '忽略'); + await expect(name).toHaveValue('人工修正姓名'); + await page.getByRole('button', { name: '使用所选资料的封面信息', exact: true }).click(); + await expect(name).toHaveValue(identity.name); +}); + +test('ignoring all files gives an explicit selected-material error', async ({ page }) => { + await setup(page, mixed()); + await choose(page, '处理 test/student/汇报PPT.pdf', '忽略'); + await choose(page, '处理 test/student/最终报告.pdf', '忽略'); + await expect(page.getByText(/至少选择1份资料/)).toBeVisible(); + await expect(page.getByRole('button', { name: '确认导入', exact: true })).toBeDisabled(); +}); + +test('unknown and duplicate material types explain the block before sending a request', async ({ page }) => { + const project = mixed(); + project.files[1].category = 'unknown'; + project.files[1].detected_identity = identity; + await setup(page, project); + await expect(page.getByText(/请选择材料类型/)).toBeVisible(); + await expect(page.getByRole('button', { name: '确认导入', exact: true })).toBeDisabled(); + await choose(page, '材料类型 test/student/汇报PPT.pdf', '最终报告'); + await expect(page.getByText(/最终报告选了2份/)).toBeVisible(); + await choose(page, '处理 test/student/汇报PPT.pdf', '忽略'); + await expect(page.getByRole('button', { name: '确认导入', exact: true })).toBeEnabled(); +});