feat: 分析报告 v2 富报告(Phase B)——SSE生成+预览写回+JD逐条比对+打印水印, 修复风险条目JSON外漏

README.md

@@ -171,6 +171,7 @@ backend\.venv\Scripts\python.exe backend/scripts/start_server.py --with-frontend
 | GET | `/api/ingest-queue` | 查看待确认采集队列 |
 | POST | `/api/ingest-queue/clear` | 确认并清空采集队列 |
 | POST | `/api/reports` | 生成候选人分析报告（冻结 AI 评估快照 + 公网分享链接） |
+| POST | `/api/reports/generate-v2/stream` | v2 富报告 SSE 流式生成（预览 → 仅快照 / 写回） |
 | GET | `/api/reports?candidateId=` | 列出候选人分析报告历史 |
 | PATCH | `/api/reports/{reportId}` | 停用 / 重新启用分析报告分享 |
 | GET | `/api/share/report/{reportId}` | 分析报告公网匿名只读（需链接内 token） |

backend/app/repositories/report_repository.py

@@ -31,10 +31,14 @@ class ReportError(Exception):
         super().__init__(detail)
 
 
-def build_snapshot(candidate: Candidate) -> dict[str, Any]:
-    """从候选人行 data JSON 冻结分享快照，只取对外安全字段。"""
+def build_snapshot(candidate: Candidate, analysis_override: dict[str, Any] | None = None) -> dict[str, Any]:
+    """从候选人行 data JSON 冻结分享快照，只取对外安全字段。
+
+    analysis_override：v2「仅快照（不写回）」分支直接传入新生成的富分析对象 R；
+    否则复用候选行里已持久化的 analysis（v1 路径）。
+    """
     data = parse_json(candidate.data, {})
-    analysis = data.get("analysis")
+    analysis = analysis_override if analysis_override is not None else data.get("analysis")
     if not isinstance(analysis, dict) or analysis.get("score") in (None, ""):
         raise ReportError(400, "候选人尚未生成有效的 AI 评估报告，请先生成后再分享")
     return {
@@ -80,8 +84,9 @@ def create_report(
     author_username: str,
     author_name: str,
     base_url: str,
+    analysis_override: dict[str, Any] | None = None,
 ) -> dict[str, Any]:
-    snapshot = build_snapshot(candidate)
+    snapshot = build_snapshot(candidate, analysis_override)
     snapshot["author"] = {"username": author_username, "name": author_name}
     snapshot["generatedAt"] = utc_now_iso()
     now = utc_now_iso()

backend/app/routers/reports.py

-"""分析报告（候选人报告）路由（ADR 0002）。
-
-- POST   /api/reports                    为当前工作区候选人冻结一份分享快照
-- GET    /api/reports?candidateId=...    列出该候选人报告历史（附分享开关状态）
-- PATCH  /api/reports/{report_id}        停用 / 重新启用（{status: active|disabled}）
-- GET    /api/share/report/{report_id}   公网匿名只读（带不可猜 token，security.py 放行）
-
-管理端点一律登录态 + resolve_owner 工作区隔离；公开只读端点独立于账号体系，
-路由内部校验 report_id + token，失败统一按「不存在/已失效」处理。
+"""分析报告（候选人报告）路由（ADR 0002 / Phase B v2）。
+
+管理端点（登录态 + resolve_owner 隔离）：
+- POST   /api/reports                           冻结快照并创建分享（可选 analysis 覆盖 → v2「仅快照」）
+- GET    /api/reports?candidateId=...           列出候选人报告历史（附分享开关状态）
+- PATCH  /api/reports/{report_id}               停用 / 重新启用
+- POST   /api/reports/generate-v2/stream        v2 富报告 SSE 流式生成（LLM → 本地回退）
+公开只读（匿名，security.py 放行 GET /api/share/）：
+- GET    /api/share/report/{report_id}          带不可猜 token 读取快照
 """
 
+import json
+from collections.abc import AsyncIterator
 from typing import Any
 
 from fastapi import APIRouter, Depends, HTTPException, Request
+from fastapi.responses import StreamingResponse
 from sqlalchemy import select
 from sqlalchemy.orm import Session
 
 from backend.app.config import get_settings
 from backend.app.db import get_db
 from backend.app.dependencies import CurrentUser, get_current_user, resolve_owner
-from backend.app.models import Candidate
+from backend.app.models import Candidate, Job
+from backend.app.repositories.json_utils import parse_json
 from backend.app.repositories.pool_repository import POOL_OWNER
 from backend.app.repositories.report_repository import (
     ReportError,
@@ -27,6 +31,7 @@ from backend.app.repositories.report_repository import (
     read_public_report,
     set_report_status,
 )
+from backend.app.services.candidate_report_v2 import generate_v2_analysis
 
 router = APIRouter()
 
@@ -44,6 +49,15 @@ def _candidate_id(payload: dict[str, Any]) -> str:
     return value
 
 
+def _candidate_row(db: Session, candidate_id: str, owner: str) -> Candidate:
+    if owner == POOL_OWNER:
+        raise HTTPException(status_code=400, detail="人才池候选人的分析报告暂不支持，请先认领到工作区")
+    candidate = db.scalar(select(Candidate).where(Candidate.id == candidate_id, Candidate.owner_id == owner))
+    if candidate is None:
+        raise HTTPException(status_code=404, detail="候选人不存在或不属于当前工作区")
+    return candidate
+
+
 @router.post("/reports")
 def generate_report(
     payload: dict[str, Any],
@@ -52,12 +66,11 @@ def generate_report(
     user: CurrentUser = Depends(get_current_user),
     owner: str = Depends(resolve_owner),
 ) -> dict[str, Any]:
-    if owner == POOL_OWNER:
-        raise HTTPException(status_code=400, detail="人才池候选人的分析报告暂不支持，请先认领到工作区")
     candidate_id = _candidate_id(payload)
-    candidate = db.scalar(select(Candidate).where(Candidate.id == candidate_id, Candidate.owner_id == owner))
-    if candidate is None:
-        raise HTTPException(status_code=404, detail="候选人不存在或不属于当前工作区")
+    candidate = _candidate_row(db, candidate_id, owner)
+    analysis_override = payload.get("analysis")
+    if analysis_override is not None and not isinstance(analysis_override, dict):
+        raise HTTPException(status_code=400, detail="analysis 参数必须是对象")
     try:
         report = create_report(
             db,
@@ -66,6 +79,7 @@ def generate_report(
             author_username=user.username,
             author_name=user.name,
             base_url=_base_url(request),
+            analysis_override=analysis_override,
         )
     except ReportError as exc:
         raise HTTPException(status_code=exc.status_code, detail=exc.detail) from exc
@@ -113,6 +127,52 @@ def update_report_status(
     return {"ok": True, "report": report}
 
 
+@router.post("/reports/generate-v2/stream")
+async def generate_v2_report_stream(
+    payload: dict[str, Any],
+    db: Session = Depends(get_db),
+    user: CurrentUser = Depends(get_current_user),
+    owner: str = Depends(resolve_owner),
+) -> StreamingResponse:
+    """v2 富报告 SSE 流式生成（Phase B）。
+
+    在同步段先锁定候选/岗位数据，事件流只做异步 LLM/本地组装：
+    - {"type":"stage","stage":"...","message":"..."}  进度
+    - {"type":"done","analysis":{...}}                 完成（v2 富对象 R）
+    - {"type":"error","message":"..."}                 失败
+    """
+    candidate_id = _candidate_id(payload)
+    candidate = _candidate_row(db, candidate_id, owner)
+    candidate_data = parse_json(candidate.data, {})
+    job_id = candidate.primary_job_id or candidate_data.get("jobId")
+    job_row = db.scalar(select(Job).where(Job.id == job_id, Job.owner_id == owner)) if job_id else None
+    job_data = parse_json(job_row.data, {}) if job_row else {}
+    if job_row is not None:
+        job_data = {**job_data, "id": job_row.id, "title": job_data.get("title") or job_row.title or ""}
+
+    async def event_stream() -> AsyncIterator[str]:
+        def emit(event: dict[str, Any]) -> str:
+            return f"data: {json.dumps(event, ensure_ascii=False)}\n\n"
+
+        yield emit({"type": "stage", "stage": "preparing", "message": "正在读取候选人与岗位资料…"})
+        if job_row is None or not str(job_data.get("title") or "").strip():
+            yield emit({"type": "error", "message": "请先为候选人选择并确认关联岗位，再生成报告"})
+            return
+        yield emit({"type": "stage", "stage": "analyzing", "message": "AI 正在做初筛评估与逐条比对…"})
+        try:
+            analysis = await generate_v2_analysis(candidate_data, job_data)
+        except Exception as exc:  # pragma: no cover - defensive
+            yield emit({"type": "error", "message": f"报告生成失败: {exc}"})
+            return
+        yield emit({"type": "done", "analysis": analysis})
+
+    return StreamingResponse(
+        event_stream(),
+        media_type="text/event-stream",
+        headers={"Cache-Control": "no-cache", "X-Accel-Buffering": "no"},
+    )
+
+
 @router.get("/share/report/{report_id}")
 def public_share_report(report_id: str, request: Request, db: Session = Depends(get_db)) -> dict[str, Any]:
     """公网匿名只读：不依赖登录态，报告路由在 security.py 已放行 GET。"""

backend/app/services/candidate_report_v2.py

+"""v2 分析报告生成（Phase B，ADR 0002 追加）。
+
+设计（与决策一致）：
+1. 先生成 v1 兼容基础评估（复用 analyze_resume 的 LLM → local-structured 回退链）；
+2. 再生成「对外富报告」字段（conclusion/grade/highlights/jdCompare/dimensionNotes），
+   LLM 优先、无 Key 时本地规则组装，绝不 500；
+3. 输出单一富对象 R = 基础评估 + v2 字段 + version:2，供「预览 → 仅快照 / 写回」两分支复用。
+
+铁律对齐 llm.py：候选人在发送前脱敏；岗位内部字段（_sanitize_job_for_external）不进 prompt；
+自述优势不作为亮点/JD 比对证据。
+"""
+
+import json
+from typing import Any
+
+from backend.app.config import get_settings
+from backend.app.services.llm import (
+    _chat_json,
+    _sanitize_job_for_external,
+    analyze_resume,
+    job_criteria,
+)
+
+_GRADE_RULES = ((85, "A"), (72, "B"), (60, "C"))
+
+
+def _list(value: Any) -> list[str]:
+    if isinstance(value, list):
+        return [str(item).strip() for item in value if str(item).strip()]
+    return []
+
+
+def grade_for_score(score: Any) -> str:
+    try:
+        value = int(round(float(score)))
+    except (TypeError, ValueError):
+        return "C"
+    for threshold, grade in _GRADE_RULES:
+        if value >= threshold:
+            return grade
+    return "D"
+
+
+def _find_sentence(keyword: str, sentences: list[str]) -> str:
+    needle = str(keyword or "").strip().lower()
+    for sentence in sentences:
+        if needle and needle in str(sentence).lower():
+            return str(sentence)
+    return ""
+
+
+def local_jd_compare(base: dict[str, Any], criteria: dict[str, list[str]]) -> list[dict[str, str]]:
+    hit_pool = [
+        *_list(base.get("hardMatch")),
+        *_list(base.get("strengths")),
+        *_list(base.get("projectEvidence")),
+        *_list(base.get("evidence")),
+    ]
+    miss_pool = [*_list(base.get("risks")), *_list(base.get("interviewQuestions"))]
+    matched_terms = [str(item) for item in _list(base.get("evidence"))]
+    rows: list[dict[str, str]] = []
+
+    def add_row(requirement: str) -> None:
+        evidence = _find_sentence(requirement, hit_pool)
+        if evidence and any(token in evidence for token in ("待确认", "缺口", "未体现", "未在简历", "建议面试")):
+            conclusion = "待确认"
+        elif evidence:
+            conclusion = "匹配"
+        elif _find_sentence(requirement, miss_pool) or requirement in matched_terms or not requirement:
+            conclusion = "未体现"
+        else:
+            conclusion = "待确认"
+        rows.append({"requirement": requirement, "evidence": evidence, "conclusion": conclusion})
+
+    for requirement in criteria.get("mustHave") or []:
+        add_row(requirement)
+    for requirement in (criteria.get("niceToHave") or [])[:6]:
+        add_row(requirement)
+    for requirement in (criteria.get("knockout") or [])[:4]:
+        add_row(requirement)
+    return rows
+
+
+def local_v2_fields(base: dict[str, Any], job: dict[str, Any], criteria: dict[str, list[str]]) -> dict[str, Any]:
+    score = base.get("score")
+    decision = str(base.get("decision") or "建议进入初筛")
+    summary = str(base.get("summary") or "")
+    highlights = _list(base.get("strengths"))[:4] or ["暂无突出亮点，建议结合简历与面试进一步核实。"]
+    risks = _list(base.get("risks"))
+    risk_text = "；".join(risks[:3]) if risks else "暂未发现明显风险"
+    dimension_notes = {
+        "hard": "；".join(_list(base.get("hardMatch"))[:3]) or "见评估明细",
+        "experience": "；".join(_list(base.get("experienceMatch"))[:3]) or "见评估明细",
+        "evidence": "；".join(_list(base.get("projectEvidence"))[:3]) or "见评估明细",
+        "competency": "；".join(_list(base.get("competencyFrame"))[:3]) or "见评估明细",
+        "risk": risk_text,
+    }
+    return {
+        "version": 2,
+        "conclusion": f"{decision}（{score} 分）。{summary}" if summary else f"{decision}（{score} 分）。",
+        "grade": grade_for_score(score),
+        "highlights": highlights,
+        "jdCompare": local_jd_compare(base, criteria),
+        "dimensionNotes": dimension_notes,
+        "enrichProvider": "local",
+        "targetJob": str(job.get("title") or base.get("jobTitle") or ""),
+    }
+
+
+async def llm_v2_fields(payload: dict[str, Any], base: dict[str, Any]) -> dict[str, Any] | None:
+    settings = get_settings()
+    if not settings.effective_llm_api_key:
+        return None
+    safe_job = _sanitize_job_for_external(payload.get("job") or {})
+    criteria = job_criteria(safe_job)
+    criteria_json = json.dumps(criteria, ensure_ascii=False, indent=2)
+    job_json = json.dumps({k: v for k, v in safe_job.items() if k not in ("criteria", "jdHistory")}, ensure_ascii=False, indent=2)[:6000]
+    candidate = payload.get("candidate") or {}
+    candidate_json = json.dumps({k: v for k, v in candidate.items() if k != "resumeText"}, ensure_ascii=False, indent=2)[:4000]
+    resume_text = (candidate.get("resumeText") or "")[:16000]
+    score = base.get("score")
+    risk_summary = "；".join(_list(base.get("risks"))[:3]) or "暂未发现明显风险"
+
+    user = (
+        "请基于下面的岗位要求与脱敏简历，产出一份对外初筛分析报告的细化字段，只返回 JSON，不要 Markdown。\n"
+        "字段与规则：\n"
+        "- conclusion：一句话结论（不超过 60 字），涵盖匹配要点与主要风险提示，与给定综合得分口径一致；\n"
+        "- grade：按综合得分定级：>=85 为 A，>=72 为 B，>=60 为 C，否则 D；\n"
+        "- highlights：核心亮点数组，每项 1 句话并带简历证据（如“主导 XX 项目，将 YY 提升 Z%”），最多 4 项，没有则给 1 条客观说明；\n"
+        "- jdCompare：逐条比对数组，每项含 requirement、evidence、conclusion 三个字段：requirement 取岗位硬性/加分/排除要求原文；"
+        "evidence 是候选人简历中的证据短句（无证据给空字符串）；conclusion 只能是“匹配 / 部分匹配 / 未体现 / 待确认”之一；\n"
+        "- dimensionNotes：对象，含 hard、experience、evidence、competency、risk 五个键，每个键 1-2 句话说明该维度得分理由或待确认点。\n"
+        "评估铁律：候选人在简历“自我评价 / 个人优势”中的主观表述（如“抗压能力强”）不作为亮点或比对证据，"
+        "只能作为面试待验证线索；证据只能来自工作/项目/实习的职责、产出与数据、教育背景、证书与明确技能。\n"
+        f"综合得分：{score}；基础结论：{base.get('decision')}；风险摘要：{risk_summary}\n"
+        f"岗位要求：{criteria_json}\n"
+        f"岗位信息：{job_json}\n"
+        f"脱敏候选人：{candidate_json}\n"
+        f"脱敏简历正文：\n{resume_text}"
+    )
+    parsed = await _chat_json(
+        settings,
+        "你是企业招聘系统中的资深 HR 报告撰写助手。只返回 JSON；结论必须与给定综合得分一致，不得推翻分数或虚构简历证据。",
+        user,
+        temperature=0,
+    )
+    if not isinstance(parsed, dict):
+        return None
+    parsed.setdefault("dimensionNotes", {})
+    parsed.setdefault("jdCompare", [])
+    parsed.setdefault("highlights", [])
+    parsed["version"] = 2
+    parsed["enrichProvider"] = f"llm:{settings.effective_llm_model}"
+    parsed["targetJob"] = str(safe_job.get("title") or "")
+    return parsed
+
+
+async def generate_v2_analysis(candidate_data: dict[str, Any], job_data: dict[str, Any]) -> dict[str, Any]:
+    """生成 v2 富报告对象 R（基础评估 + 富报告字段）。"""
+    payload = {"candidate": candidate_data, "job": job_data}
+    base = await analyze_resume(payload)
+    enriched = await llm_v2_fields(payload, base)
+    if enriched is None:
+        enriched = local_v2_fields(base, job_data, job_criteria(job_data))
+    result = {**base, **enriched}
+    result.setdefault("grade", grade_for_score(base.get("score")))
+    result.setdefault("conclusion", f"{base.get('decision')}（{base.get('score')} 分）。")
+    return result

backend/tests/test_reports.py

@@ -6,6 +6,7 @@
 - 关闭模式（auth_reports_client）：管理端点工作区隔离；公网只读匿名可访问。
 """
 
+import json
 from urllib.parse import parse_qs, urlparse
 
 import pytest
@@ -197,6 +198,100 @@ def test_public_read_wrong_token_or_unknown(share_open_client):
     assert body["ok"] is False and body["reason"] == "not_found"
 
 
+# ---- Phase B v2：富报告 SSE 生成 + analysis 覆盖冻结 ----
+
+def _v2_state():
+    job = {
+        "id": "job-v2-1",
+        "title": "数据分析师",
+        "department": "数据部",
+        "jd": "负责业务数据体系与经营分析，输出可执行结论。",
+        "mustHave": "本科及以上\n3年以上数据分析经验",
+        "niceToHave": "SQL\nPython",
+        "knockout": "无数据项目经历",
+        "competency": "业务理解\n沟通协同",
+        "matchKeywords": "SQL,Python,数据分析",
+    }
+    candidate = {
+        "id": "cand-v2-1",
+        "jobId": "job-v2-1",
+        "jobTitle": "数据分析师",
+        "name": "李四",
+        "source": "Boss",
+        "stage": "初筛通过",
+        "education": "本科",
+        "school": "某大学",
+        "major": "统计学",
+        "years": "3年",
+        "skills": ["SQL", "Python"],
+        "resumeText": "三年数据分析经验，熟悉 SQL、Python，主导过订单分析项目，将报表效率提升 40%，服务过电商客户。",
+    }
+    return {"jobs": [job], "candidates": [candidate], "offers": [], "tasks": [], "eventLog": []}
+
+
+def _sse_events(text):
+    events = []
+    for line in str(text).splitlines():
+        line = line.strip()
+        if line.startswith("data:"):
+            try:
+                events.append(json.loads(line[5:].strip()))
+            except json.JSONDecodeError:
+                continue
+    return events
+
+
+def test_v2_stream_local_generation(share_open_client):
+    assert share_open_client.put("/api/state", json=_v2_state()).status_code == 200
+    response = share_open_client.post("/api/reports/generate-v2/stream", json={"candidateId": "cand-v2-1"})
+    assert response.status_code == 200
+    assert response.headers["content-type"].startswith("text/event-stream")
+    events = _sse_events(response.text)
+    done = [event for event in events if event.get("type") == "done"]
+    assert len(done) == 1
+    analysis = done[0]["analysis"]
+    assert analysis.get("version") == 2
+    assert analysis.get("grade") in ("A", "B", "C", "D")
+    assert analysis.get("score") is not None
+    assert isinstance(analysis.get("jdCompare"), list) and analysis["jdCompare"]
+    assert analysis.get("conclusion")
+
+
+def test_v2_stream_missing_job_error(share_open_client):
+    state = _v2_state()
+    state["candidates"][0]["jobId"] = ""
+    state["jobs"] = []
+    assert share_open_client.put("/api/state", json=state).status_code == 200
+    response = share_open_client.post("/api/reports/generate-v2/stream", json={"candidateId": "cand-v2-1"})
+    events = _sse_events(response.text)
+    errors = [event for event in events if event.get("type") == "error"]
+    assert errors and "关联岗位" in errors[0]["message"]
+
+
+def test_create_report_with_analysis_override_v2(share_open_client):
+    assert share_open_client.put("/api/state", json=_v2_state()).status_code == 200
+    analysis = {
+        "version": 2,
+        "score": 82,
+        "decision": "建议复核后进入初筛",
+        "conclusion": "总体匹配，需复核数据分析深度。",
+        "grade": "B",
+        "highlights": ["主导订单分析项目，报表效率提升 40%"],
+        "jdCompare": [{"requirement": "本科及以上", "evidence": "统招本科", "conclusion": "匹配"}],
+        "dimensionNotes": {"hard": "学历达标", "experience": "3 年经验", "evidence": "有项目证据", "competency": "待面试观察", "risk": "无明显风险"},
+    }
+    created = share_open_client.post(
+        "/api/reports", json={"candidateId": "cand-v2-1", "analysis": analysis}
+    )
+    assert created.status_code == 200
+    report = created.json()["report"]
+    assert report["snapshot"]["analysis"]["version"] == 2
+    assert report["snapshot"]["analysis"]["grade"] == "B"
+    # 「仅快照」不改写候选行 analysis
+    state = share_open_client.get("/api/state").json()
+    assert "analysis" not in state["candidates"][0]
+
+
 # ---- SPA 直链回退（公网分享页深链需返回 index.html）----
 
 def test_spa_deep_link_fallback_serves_index(client):

docs/adr/0002-candidate-report-share.md

@@ -26,3 +26,19 @@
 - 正向：分享产物与实时评估解耦，符合「导出时刻所见即所得」；不破坏整包 state 机制；匿名面小（只读 + 令牌 + 白名单 + 默认关）。
 - 代价/注意：报告独立存储 → **整包「导出备份」不含报告**，需要时后续补报告专用导出/迁移；快照内容在生成后不随候选人资料更新（这是刻意语义，分享页应显示「报告生成时间」）；公开接口需在 `security.py` 放行名单中精确按前缀处理，避免误放其它 `/api/share` 业务。
 - 后续可做：Phase B 更丰富的 AI 报告模板（一句话结论 + 维度/亮点/风险 + JD 逐条比对表，同 AI秒聘）；PDF（水印）导出；自动过期；人才池报告；报告备份/迁移。
+
+
+## Phase B（v2）追加决策（2026-09-03，已实现 P0+P1）
+
+- 状态：已接受
+- **v2 生成方式**：先生成 v1 兼容基础评估（复用 `analyze_resume` 的 LLM → local-structured 回退链），
+  再做一次「富化」LLM 调用产出 conclusion/grade/highlights/jdCompare/dimensionNotes；无 Key 时用本地规则组装，
+  绝不 500。输出**单一富对象 R** = 基础评估 + v2 字段 + `version:2`（保留 v1 兼容键，抽屉「评估」页可继续渲染）。
+- **SSE 流式**：新增 `POST /api/reports/generate-v2/stream`（stage/done/error），沿用 JD 生成的 SSE 模式，不引入任务队列。
+- **维度评分**：沿用现有 scoreBreakdown 五维（硬性/经验/项目证据/胜任力/风险），不引入岗位自定义权重维度。
+- **预览与写回（决策 5）**：v2 生成先给预览；「仅生成分享报告」= `POST /api/reports` 携带 analysis 覆盖冻结快照
+  （不改候选行）；「同时刷新 AI 评估」= 先写回 `candidate.analysis`（补 jobSnapshot/inputKey/generatedAt）→ persist →
+  再按 v1 语义从候选行冻结。两种都产出带 report_id 的分享链接，旧 v1 快照/旧分析兼容渲染。
+- **PDF（P1，轻量）**：不做服务端渲染器；公网页「打印 / 存 PDF」+ 每页固定轻量水印（报告编号/内部资料/生成人），
+  浏览器另存 PDF 即得水印版。
+- 边界更新：v2 仍限个人工作区简历库抽屉；人才池报告、自动过期（表字段已预留）、报告备份/迁移仍为后续项。

vue-app/src/api/recruitment.js

@@ -179,12 +179,64 @@ export async function listCandidateReports(candidateId) {
   return data
 }
 
-export async function createCandidateReport(candidateId) {
-  const { data } = await http.post('/api/reports', { candidateId })
+export async function createCandidateReport(candidateId, analysis = null) {
+  const body = { candidateId }
+  if (analysis && typeof analysis === 'object') body.analysis = analysis
+  const { data } = await http.post('/api/reports', body)
   if (data.error) throw new Error(data.error)
   return data
 }
 
+// v2 富报告 SSE 流式生成（Phase B）：stage / done / error 事件。
+export async function generateV2ReportStream(candidateId, handlers = {}) {
+  const { onStage, onDone, onError } = handlers
+  let lastError = ''
+
+  let base = ''
+  if (typeof window !== 'undefined' && window.RECRUITMENT_API_BASE_URL) {
+    base = String(window.RECRUITMENT_API_BASE_URL).replace(/\/+$/, '')
+  }
+  const token = readToken()
+  const response = await fetch(`${base}/api/reports/generate-v2/stream`, {
+    method: 'POST',
+    headers: {
+      'Content-Type': 'application/json',
+      ...(token ? { Authorization: `Bearer ${token}` } : {}),
+    },
+    body: JSON.stringify({ candidateId }),
+  })
+  if (!response.ok || !response.body) {
+    throw new Error(`报告生成请求失败（HTTP ${response.status}）`)
+  }
+  const reader = response.body.getReader()
+  const decoder = new TextDecoder('utf-8')
+  let buffer = ''
+  for (;;) {
+    const { done, value } = await reader.read()
+    if (done) break
+    buffer += decoder.decode(value, { stream: true })
+    const lines = buffer.split('\n')
+    buffer = lines.pop() || ''
+    for (const line of lines) {
+      const trimmed = line.trim()
+      if (!trimmed.startsWith('data:')) continue
+      let event
+      try {
+        event = JSON.parse(trimmed.slice(5).trim())
+      } catch {
+        continue
+      }
+      if (event.type === 'stage' && onStage) onStage(event.stage, event.message || '')
+      else if (event.type === 'done' && onDone) onDone(event.analysis)
+      else if (event.type === 'error') {
+        lastError = event.message || '报告生成失败'
+        if (onError) onError(lastError)
+      }
+    }
+  }
+  if (lastError) throw new Error(lastError)
+}
+
 export async function setCandidateReportStatus(reportId, status) {
   const { data } = await http.patch(`/api/reports/${reportId}`, { status })
   if (data.error) throw new Error(data.error)

vue-app/src/components/reports/ReportContentView.vue

+<script setup>
+// 候选人分析报告内容渲染（ADR 0002 Phase A/B 通用）：
+// - v1 快照：得分/结论/摘要/风险/评估明细（与抽屉「评估」同源）；
+// - v2 富对象：额外渲染等级、一句话结论、维度评分条、核心亮点、JD 逐条比对。
+// 数据全部来自后端冻结的 snapshot，页面/抽屉预览共用，避免两套渲染分叉。
+import { computed } from 'vue'
+
+import { providerLabel, scoreBreakdownItems } from '@/utils/scoring'
+import { assessmentDecision, assessmentReportRows, assessmentScoreText, riskProfileRows } from '@/utils/report'
+
+const props = defineProps({
+  report: { type: Object, required: true },
+})
+
+const snapshot = computed(() => props.report?.snapshot || {})
+const analysis = computed(() => snapshot.value.analysis || {})
+const isV2 = computed(() => analysis.value.version === 2)
+
+const candidate = computed(() => ({
+  id: snapshot.value.candidateId,
+  name: snapshot.value.candidateName,
+  jobTitle: snapshot.value.jobTitle,
+  analysis: analysis.value,
+  matchReport: snapshot.value.matchReport,
+  quickMatchScore: snapshot.value.quickMatchScore,
+}))
+
+const scoreText = computed(() => assessmentScoreText(candidate.value, 'ai'))
+const decisionInfo = computed(() => assessmentDecision(candidate.value, 'ai'))
+const reportRows = computed(() => assessmentReportRows(candidate.value, 'ai'))
+const riskProfile = computed(() => riskProfileRows(analysis.value))
+const dimensionItems = computed(() => scoreBreakdownItems(analysis.value))
+const author = computed(() => snapshot.value.author || {})
+
+const gradeTone = computed(() => {
+  const grade = String(analysis.value.grade || '').toUpperCase()
+  if (grade === 'A') return 'success'
+  if (grade === 'B') return 'primary'
+  if (grade === 'C') return 'warning'
+  return 'danger'
+})
+
+const jdConclusionMeta = (conclusion = '') => {
+  const map = {
+    匹配: { type: 'success', text: '匹配' },
+    部分匹配: { type: 'primary', text: '部分匹配' },
+    未体现: { type: 'danger', text: '未体现' },
+    待确认: { type: 'warning', text: '待确认' },
+  }
+  return map[conclusion] || { type: 'info', text: conclusion || '待确认' }
+}
+
+const toneTag = (level) =>
+  ({
+    ok: { type: 'success', text: '正常' },
+    hint: { type: 'warning', text: '提示' },
+    warn: { type: 'danger', text: '需确认' },
+  })[level] || { type: 'info', text: level }
+
+const DIM_NOTES_KEY = ['hard', 'experience', 'evidence', 'competency', 'risk']
+
+const dimensionNote = (index) => {
+  const notes = analysis.value.dimensionNotes || {}
+  const key = DIM_NOTES_KEY[index]
+  return key ? notes[key] : ''
+}
+
+const toText = (item) => {
+  if (item && typeof item === 'object') {
+    if (item.title && item.detail) return `${item.title}：${item.detail}`
+    return item.title || item.text || item.detail || JSON.stringify(item)
+  }
+  return String(item ?? '')
+}
+
+const formatTime = (iso = '') => {
+  if (!iso) return ''
+  const date = new Date(iso)
+  if (Number.isNaN(date.getTime())) return iso
+  const pad = (n) => String(n).padStart(2, '0')
+  return `${date.getFullYear()}-${pad(date.getMonth() + 1)}-${pad(date.getDate())} ${pad(date.getHours())}:${pad(date.getMinutes())}`
+}
+</script>
+
+<template>
+  <div class="report-content">
+    <section class="candidate-head">
+      <div class="name-row">
+        <h1>{{ snapshot.candidateName || '未命名候选人' }}</h1>
+        <el-tag size="small" effect="plain">{{ snapshot.jobTitle || '待匹配岗位' }}</el-tag>
+        <el-tag v-if="snapshot.source" size="small" type="info" effect="plain">{{ snapshot.source }}</el-tag>
+        <el-tag v-if="isV2" size="small" :type="gradeTone" effect="dark" class="grade-tag">
+          等级 {{ analysis.grade }}
+        </el-tag>
+        <el-tag v-if="isV2" size="small" type="primary" effect="plain">v2 富评估</el-tag>
+      </div>
+      <div class="meta-row">
+        <span>报告编号：{{ report.reportId || '—' }}</span>
+        <span>生成时间：{{ formatTime(report.generatedAt) }}</span>
+        <span>评估方式：{{ providerLabel(analysis.provider || '') }}</span>
+        <span>生成人：{{ author.name || author.username || '—' }}</span>
+      </div>
+    </section>
+
+    <section class="summary-strip" :class="`tone-${decisionInfo.tone}`">
+      <div class="score-block">
+        <strong class="score-value">{{ scoreText }}</strong>
+        <span>综合得分</span>
+      </div>
+      <div class="decision-block">
+        <span class="eyebrow">初筛建议</span>
+        <strong>{{ decisionInfo.text }}</strong>
+        <p>{{ decisionInfo.note }}</p>
+      </div>
+    </section>
+
+    <section v-if="isV2 && analysis.conclusion" class="block">
+      <h4>一句话结论</h4>
+      <p class="conclusion-text">{{ analysis.conclusion }}</p>
+    </section>
+
+    <section v-if="analysis.summary" class="block">
+      <h4>评估摘要</h4>
+      <p class="summary-text">{{ analysis.summary }}</p>
+    </section>
+
+    <section v-if="isV2 && dimensionItems.length" class="block">
+      <h4>维度评分</h4>
+      <div class="dimension-list">
+        <div
+          v-for="(item, index) in dimensionItems"
+          :key="item.label"
+          class="dimension-item"
+          :class="{ total: item.total }"
+        >
+          <div class="dimension-head">
+            <b>{{ item.label }}</b>
+            <strong>{{ item.value < 0 ? item.value : `${item.value}/${item.max}` }}</strong>
+          </div>
+          <div class="dimension-track">
+            <i
+              :style="{
+                width: item.total
+                  ? `${Math.min(100, Math.max(0, item.value))}%`
+                  : `${Math.min(100, (Math.abs(item.value) / item.max) * 100)}%`,
+              }"
+            />
+          </div>
+          <p class="dimension-hint">{{ item.hint }}</p>
+          <p v-if="dimensionNote(index)" class="dimension-note">{{ dimensionNote(index) }}</p>
+        </div>
+      </div>
+    </section>
+
+    <section v-if="isV2 && analysis.highlights && analysis.highlights.length" class="block">
+      <h4>核心亮点</h4>
+      <ul class="highlight-list">
+        <li v-for="(item, index) in analysis.highlights" :key="index">{{ toText(item) }}</li>
+      </ul>
+    </section>
+
+    <section v-if="isV2 && analysis.jdCompare && analysis.jdCompare.length" class="block">
+      <h4>JD 逐条比对</h4>
+      <div class="jd-compare">
+        <div v-for="(row, index) in analysis.jdCompare" :key="index" class="jd-row">
+          <div class="jd-requirement">
+            <span>岗位要求</span>
+            <p>{{ toText(row.requirement) }}</p>
+          </div>
+          <div class="jd-evidence">
+            <span>候选人证据</span>
+            <p>{{ toText(row.evidence) || '未在简历中找到明确证据' }}</p>
+          </div>
+          <el-tag :type="jdConclusionMeta(row.conclusion).type" size="small" effect="light">
+            {{ jdConclusionMeta(row.conclusion).text }}
+          </el-tag>
+        </div>
+      </div>
+    </section>
+
+    <section v-if="riskProfile.length" class="block">
+      <h4>风险点评估</h4>
+      <div class="risk-list">
+        <div v-for="item in riskProfile" :key="item.dimension" class="risk-item" :class="`level-${item.level}`">
+          <el-tag :type="toneTag(item.level).type" size="small" effect="light">
+            {{ toneTag(item.level).text }}
+          </el-tag>
+          <b>{{ item.label }}</b>
+          <p>{{ item.summary }}</p>
+          <span v-for="(ev, index) in item.evidence || []" :key="`${item.dimension}-${index}`"> 依据：{{ ev }} </span>
+        </div>
+      </div>
+    </section>
+
+    <section v-if="reportRows.length" class="block">
+      <h4>评估明细</h4>
+      <div v-for="row in reportRows" :key="row.label" class="detail-card" :class="`tone-${row.tone}`">
+        <div class="detail-head">
+          <b>{{ row.label }}</b>
+          <el-tag :type="row.tone" size="small" effect="light">{{ row.status }}</el-tag>
+        </div>
+        <ul v-if="row.items.length">
+          <li v-for="(item, index) in row.items" :key="index">{{ item }}</li>
+        </ul>
+        <p v-else class="detail-empty">暂无</p>
+      </div>
+    </section>
+
+    <footer class="report-footer">
+      <p>本报告由 AI 辅助生成，基于生成时刻的简历与岗位信息冻结，仅供内部初筛参考，请以人工复核与面试为准。</p>
+      <p>不含候选人联系方式与原简历文件。</p>
+    </footer>
+  </div>
+</template>
+
+<style lang="scss" scoped>
+.report-content {
+  width: 100%;
+}
+
+.candidate-head {
+  padding: 4px 0 14px;
+
+  .name-row {
+    display: flex;
+    align-items: center;
+    flex-wrap: wrap;
+    gap: 8px;
+
+    h1 {
+      margin: 0;
+      font-size: 24px;
+    }
+
+    .grade-tag {
+      font-weight: 700;
+    }
+  }
+
+  .meta-row {
+    display: flex;
+    flex-wrap: wrap;
+    gap: 6px 18px;
+    margin-top: 10px;
+    font-size: 12px;
+    color: var(--muted);
+  }
+}
+
+.summary-strip {
+  display: flex;
+  align-items: stretch;
+  border: 1px solid var(--line);
+  border-radius: 12px;
+  overflow: hidden;
+  margin-bottom: 16px;
+
+  .score-block {
+    display: flex;
+    flex-direction: column;
+    align-items: center;
+    justify-content: center;
+    gap: 4px;
+    min-width: 168px;
+    padding: 16px 18px;
+    background: rgba(255, 255, 255, 0.8);
+
+    .score-value {
+      font-size: 30px;
+      line-height: 1;
+    }
+
+    span {
+      font-size: 12px;
+      color: var(--muted);
+    }
+  }
+
+  .decision-block {
+    flex: 1;
+    min-width: 0;
+    padding: 14px 18px;
+    display: flex;
+    flex-direction: column;
+    justify-content: center;
+    border-left: 1px solid var(--line);
+
+    .eyebrow {
+      font-size: 11px;
+      letter-spacing: 1px;
+      color: var(--muted);
+    }
+
+    strong {
+      margin: 4px 0 2px;
+      font-size: 17px;
+    }
+
+    p {
+      margin: 0;
+      font-size: 12px;
+      color: var(--muted);
+      line-height: 1.5;
+    }
+  }
+
+  &.tone-success {
+    border-color: color-mix(in srgb, var(--green) 55%, var(--line));
+
+    .score-value,
+    .decision-block strong {
+      color: var(--green);
+    }
+  }
+
+  &.tone-warning {
+    border-color: color-mix(in srgb, var(--yellow) 55%, var(--line));
+
+    .score-value,
+    .decision-block strong {
+      color: var(--yellow);
+    }
+  }
+
+  &.tone-danger {
+    border-color: color-mix(in srgb, var(--rose) 55%, var(--line));
+
+    .score-value,
+    .decision-block strong {
+      color: var(--rose);
+    }
+  }
+
+  &.tone-info {
+    .score-value,
+    .decision-block strong {
+      color: var(--blue);
+    }
+  }
+}
+
+.block {
+  margin-bottom: 16px;
+
+  h4 {
+    margin: 0 0 8px;
+    font-size: 14px;
+  }
+
+  .conclusion-text {
+    margin: 0;
+    padding: 12px 14px;
+    border: 1px solid var(--line);
+    border-left: 3px solid var(--blue);
+    border-radius: 10px;
+    background: color-mix(in srgb, var(--blue) 6%, #fff);
+    font-size: 14px;
+    line-height: 1.7;
+    color: var(--ink);
+    font-weight: 500;
+  }
+
+  .summary-text {
+    margin: 0;
+    padding: 12px 14px;
+    border: 1px solid var(--line);
+    border-radius: 10px;
+    background: rgba(255, 255, 255, 0.7);
+    font-size: 13px;
+    line-height: 1.7;
+    color: var(--ink);
+  }
+}
+
+.dimension-list {
+  display: flex;
+  flex-direction: column;
+  gap: 8px;
+
+  .dimension-item {
+    padding: 9px 12px;
+    border: 1px solid var(--line);
+    border-radius: 10px;
+    background: rgba(255, 255, 255, 0.7);
+
+    &.total {
+      border-color: color-mix(in srgb, var(--blue) 45%, var(--line));
+      background: color-mix(in srgb, var(--blue) 5%, #fff);
+    }
+
+    .dimension-head {
+      display: flex;
+      align-items: center;
+      justify-content: space-between;
+      gap: 10px;
+
+      b {
+        font-size: 13px;
+      }
+
+      strong {
+        font-size: 13px;
+      }
+    }
+
+    .dimension-track {
+      margin: 6px 0 4px;
+      height: 6px;
+      border-radius: 999px;
+      background: var(--line);
+      overflow: hidden;
+
+      i {
+        display: block;
+        height: 100%;
+        border-radius: 999px;
+        background: linear-gradient(90deg, var(--blue), var(--cyan));
+      }
+    }
+
+    .dimension-hint,
+    .dimension-note {
+      margin: 0;
+      font-size: 12px;
+      color: var(--muted);
+      line-height: 1.5;
+    }
+
+    .dimension-note {
+      margin-top: 2px;
+      color: var(--ink);
+    }
+  }
+}
+
+.highlight-list {
+  margin: 0;
+  padding: 0;
+  list-style: none;
+  display: flex;
+  flex-direction: column;
+  gap: 6px;
+
+  li {
+    padding: 8px 12px;
+    border: 1px solid var(--line);
+    border-left: 3px solid var(--green);
+    border-radius: 8px;
+    background: color-mix(in srgb, var(--green) 6%, #fff);
+    font-size: 13px;
+    line-height: 1.6;
+  }
+}
+
+.jd-compare {
+  display: flex;
+  flex-direction: column;
+  gap: 8px;
+
+  .jd-row {
+    display: grid;
+    grid-template-columns: 30% 1fr auto;
+    gap: 12px;
+    align-items: start;
+    padding: 10px 12px;
+    border: 1px solid var(--line);
+    border-radius: 10px;
+    background: rgba(255, 255, 255, 0.7);
+
+    .jd-requirement,
+    .jd-evidence {
+      min-width: 0;
+
+      span {
+        font-size: 11px;
+        color: var(--muted);
+      }
+
+      p {
+        margin: 3px 0 0;
+        font-size: 13px;
+        line-height: 1.55;
+        color: var(--ink);
+        word-break: break-word;
+      }
+    }
+  }
+}
+
+.risk-list {
+  display: flex;
+  flex-direction: column;
+  gap: 8px;
+
+  .risk-item {
+    display: grid;
+    grid-template-columns: auto auto 1fr;
+    align-items: center;
+    gap: 4px 10px;
+    padding: 10px 12px;
+    border-radius: 10px;
+    border: 1px solid var(--line);
+    background: rgba(255, 255, 255, 0.6);
+
+    b {
+      font-size: 13px;
+    }
+
+    p {
+      grid-column: 1 / -1;
+      margin: 2px 0 0;
+      font-size: 13px;
+      line-height: 1.6;
+    }
+
+    span {
+      grid-column: 1 / -1;
+      font-size: 12px;
+      color: var(--muted);
+      line-height: 1.5;
+    }
+
+    &.level-warn {
+      border-color: color-mix(in srgb, var(--rose) 45%, var(--line));
+      background: color-mix(in srgb, var(--rose) 6%, #fff);
+    }
+
+    &.level-hint {
+      border-color: color-mix(in srgb, var(--yellow) 45%, var(--line));
+      background: color-mix(in srgb, var(--yellow) 7%, #fff);
+    }
+  }
+}
+
+.detail-card {
+  margin-bottom: 10px;
+  padding: 12px 14px;
+  border: 1px solid var(--line);
+  border-left-width: 3px;
+  border-radius: 10px;
+  background: rgba(255, 255, 255, 0.7);
+
+  &.tone-success {
+    border-left-color: var(--green);
+  }
+
+  &.tone-warning {
+    border-left-color: var(--yellow);
+  }
+
+  &.tone-danger {
+    border-left-color: var(--rose);
+  }
+
+  &.tone-info {
+    border-left-color: var(--blue);
+  }
+
+  .detail-head {
+    display: flex;
+    align-items: center;
+    justify-content: space-between;
+    gap: 10px;
+    margin-bottom: 6px;
+
+    b {
+      font-size: 14px;
+    }
+  }
+
+  ul {
+    margin: 0;
+    padding-left: 18px;
+    display: flex;
+    flex-direction: column;
+    gap: 5px;
+
+    li {
+      font-size: 13px;
+      line-height: 1.6;
+      color: var(--ink);
+    }
+  }
+
+  .detail-empty {
+    margin: 0;
+    font-size: 13px;
+    color: var(--muted);
+  }
+}
+
+.report-footer {
+  margin-top: 18px;
+  padding-top: 12px;
+  border-top: 1px dashed var(--line);
+  font-size: 12px;
+  color: var(--muted);
+  line-height: 1.6;
+
+  p {
+    margin: 2px 0;
+  }
+}
+
+@media (max-width: 620px) {
+  .summary-strip {
+    flex-direction: column;
+
+    .score-block {
+      min-width: 0;
+      padding: 14px;
+    }
+
+    .decision-block {
+      border-left: 0;
+      border-top: 1px solid var(--line);
+    }
+  }
+
+  .jd-compare .jd-row {
+    grid-template-columns: 1fr auto;
+  }
+}
+</style>

vue-app/src/components/resumes/ResumeDetail.vue

@@ -6,17 +6,19 @@ import { ElMessageBox } from 'element-plus'
 
 import http from '@/api'
 import * as reportApi from '@/api/recruitment'
+import ReportContentView from '@/components/reports/ReportContentView.vue'
 import { useRecruitmentStore } from '@/stores/recruitment'
 import { CANDIDATE_TAG_OPTIONS, candidateTagTone } from '@/utils/constants'
 import { candidateWorkflow, workflowStageClass } from '@/utils/task'
 import { candidateScreeningConclusion, quickScore } from '@/utils/resume'
 import { isAiEvaluationReport } from '@/utils/scoring'
-import { analysisNeedsRefresh } from '@/utils/analysis'
+import { analysisInputKey, analysisNeedsRefresh } from '@/utils/analysis'
 import { resumeDisplayName, candidateAge, candidateYears } from '@/utils/candidate'
 import { uid } from '@/utils/normalize'
 import { assessmentDecision, assessmentReportRows, assessmentScoreText, riskProfileRows } from '@/utils/report'
 import { getFilteredCandidates } from '@/utils/matching'
-import { jobCandidates } from '@/utils/links'
+import { buildJobSnapshot } from '@/utils/job'
+import { jobCandidates, matchedJobForCandidate } from '@/utils/links'
 import { showToast } from '@/utils/toast'
 
 const store = useRecruitmentStore()
@@ -321,6 +323,10 @@ const reportListLoading = ref(false)
 const reportList = ref([])
 const reportGenerating = ref(false)
 const reportShareEnabled = ref(true)
+const reportGeneratingStage = ref('')
+const reportPreviewOpen = ref(false)
+const reportPreview = ref(null)
+const reportCommitting = ref(false)
 
 const formatReportTime = (iso = '') => {
   if (!iso) return ''
@@ -378,30 +384,76 @@ const generateReport = async () => {
     return
   }
   reportGenerating.value = true
+  reportGeneratingStage.value = '准备中…'
   try {
     if (!store.localMatch(candidate.id)) throw new Error('请先为候选人选择并确认关联岗位')
-    const previous = candidate.analysis
-    const reusable = isAiEvaluationReport(previous) && !analysisNeedsRefresh(candidate)
-    if (!reusable) {
-      showToast('正在生成 AI 评估报告...', 'info')
-      candidate.analysis = await store.analyzeResume(candidate)
+    await store.saveStateToServer()
+    let analysis = null
+    await reportApi.generateV2ReportStream(candidate.id, {
+      onStage: (_stage, message) => {
+        reportGeneratingStage.value = message || 'AI 正在生成报告…'
+      },
+      onDone: (value) => {
+        analysis = value
+      },
+    })
+    if (!analysis) throw new Error('报告生成失败，请稍后重试')
+    reportPreview.value = {
+      reportId: '预览',
+      generatedAt: new Date().toISOString(),
+      snapshot: {
+        candidateId: candidate.id,
+        candidateName: candidate.name,
+        jobTitle: candidate.jobTitle,
+        source: candidate.source,
+        analysis,
+      },
+    }
+    reportPreviewOpen.value = true
+  } catch (error) {
+    showToast(error.message || '分析报告生成失败', 'error')
+  } finally {
+    reportGenerating.value = false
+    reportGeneratingStage.value = ''
+  }
+}
+
+// 预览确认（决策 5）：仅冻结快照（不写回），或先刷新 candidate.analysis 再冻结。
+const commitPreview = async (writeBack = false) => {
+  const candidate = selected.value
+  const analysis = reportPreview.value?.snapshot?.analysis
+  if (!candidate || !analysis || reportCommitting.value) return
+  reportCommitting.value = true
+  try {
+    if (writeBack) {
+      const job = matchedJobForCandidate(candidate, store.jobs) || {}
+      const jobSnapshot = buildJobSnapshot(job)
+      analysis.jobSnapshot = jobSnapshot
+      analysis.inputKey = analysisInputKey(candidate, jobSnapshot)
+      analysis.generatedAt = new Date().toISOString()
+      candidate.analysis = analysis
       store.resumeDetailMode = 'ai'
+      await store.saveStateToServer()
     }
-    await store.saveStateToServer()
-    const payload = await reportApi.createCandidateReport(candidate.id)
+    const payload = writeBack
+      ? await reportApi.createCandidateReport(candidate.id)
+      : await reportApi.createCandidateReport(candidate.id, analysis)
     const report = payload.report
     store.logEvent('report_generated', candidate, {
       reportId: report.reportId,
       score: reportSnapshotScore(report.snapshot),
+      reportVersion: report.snapshot?.analysis?.version === 2 ? 'v2' : 'v1',
     })
     store.persist('分析报告已生成')
+    reportPreviewOpen.value = false
+    reportPreview.value = null
     await copyText(report.shareUrl)
     showToast('分析报告已生成，链接已复制到剪贴板', 'success')
     await loadReports()
   } catch (error) {
-    showToast(error.message || '分析报告生成失败', 'error')
+    showToast(error.message || '报告创建失败', 'error')
   } finally {
-    reportGenerating.value = false
+    reportCommitting.value = false
   }
 }
 
@@ -1085,7 +1137,7 @@ const saveInterview = (candidate) => {
       </footer>
     </div>
 
-    <el-dialog v-model="reportDialogOpen" title="分析报告" width="620px" append-to-body class="report-panel-dialog">
+    <el-dialog v-model="reportDialogOpen" :close-on-click-modal="false" title="分析报告" width="620px" append-to-body class="report-panel-dialog">
       <div class="report-panel">
         <div class="report-panel-head">
           <div class="report-panel-identity">
@@ -1093,7 +1145,7 @@ const saveInterview = (candidate) => {
             <span>{{ selected?.jobTitle || '待匹配岗位' }}</span>
           </div>
           <el-button type="primary" :loading="reportGenerating" :disabled="!reportShareEnabled" @click="generateReport">
-            生成新报告
+            生成 v2 报告
           </el-button>
         </div>
         <el-alert
@@ -1106,8 +1158,12 @@ const saveInterview = (candidate) => {
           description="当前部署未开启 REPORT_SHARE_ENABLED，生成的链接外部无法访问；请联系管理员在配置中开启。"
         />
         <p class="report-hint">
-          每次「生成新报告」会基于当前 AI
-          评估冻结一份快照并创建公网链接；分享页仅含姓名、岗位与评估结论，不含联系方式与原简历。
+          每次「生成 v2 报告」会先生成富报告并弹窗预览，可在「仅生成分享报告」与「同时刷新 AI
+          评估」之间选择；确认后才会冻结快照并创建公网链接。分享页仅含姓名、岗位与评估结论，不含联系方式与原简历。
+        </p>
+        <p v-if="reportGenerating" class="report-stage">
+          <el-icon class="is-loading"><Loading /></el-icon>
+          {{ reportGeneratingStage }}
         </p>
         <div v-if="reportListLoading" class="report-list-loading">加载中…</div>
         <div v-else-if="!reportList.length" class="report-empty">还没有分析报告，点击右上角「生成新报告」创建。</div>
@@ -1157,6 +1213,33 @@ const saveInterview = (candidate) => {
         </div>
       </div>
     </el-dialog>
+
+    <el-dialog
+      v-model="reportPreviewOpen"
+      title="报告预览（v2 富评估）"
+      width="820px"
+      top="6vh"
+      append-to-body
+      class="report-preview-dialog"
+    >
+      <div class="report-preview-body">
+        <ReportContentView v-if="reportPreview" :report="reportPreview" />
+      </div>
+      <template #footer>
+        <div class="preview-footer">
+          <span class="preview-tip">
+            预览尚未保存。「同时刷新 AI 评估并分享」会把该富报告写回候选人评估（抽屉「评估」页同步变丰富）后再生成链接。
+          </span>
+          <div class="preview-actions">
+            <el-button :disabled="reportCommitting" @click="reportPreviewOpen = false">取消</el-button>
+            <el-button :loading="reportCommitting" @click="commitPreview(false)">仅生成分享报告</el-button>
+            <el-button type="primary" :loading="reportCommitting" @click="commitPreview(true)">
+              同时刷新 AI 评估并分享
+            </el-button>
+          </div>
+        </div>
+      </template>
+    </el-dialog>
   </el-drawer>
 </template>
 <style lang="scss" scoped>
@@ -2173,6 +2256,47 @@ const saveInterview = (candidate) => {
     }
   }
 }
+// 报告生成进度与 v2 预览（Phase B）
+.report-stage {
+  display: flex;
+  align-items: center;
+  gap: 6px;
+  margin: 0 0 10px;
+  padding: 8px 10px;
+  border-radius: 8px;
+  background: color-mix(in srgb, var(--blue) 6%, #fff);
+  color: var(--ink);
+  font-size: 12px;
+}
+
+.report-preview-dialog {
+  .report-preview-body {
+    max-height: 62vh;
+    overflow: auto;
+    padding: 2px 4px;
+  }
+
+  .preview-footer {
+    display: flex;
+    flex-direction: column;
+    align-items: stretch;
+    gap: 10px;
+
+    .preview-tip {
+      font-size: 12px;
+      line-height: 1.6;
+      color: var(--muted);
+    }
+
+    .preview-actions {
+      display: flex;
+      justify-content: flex-end;
+      flex-wrap: wrap;
+      gap: 8px;
+    }
+  }
+}
+
 // 分析报告面板（ADR 0002）
 .report-panel {
   .report-panel-head {

vue-app/src/utils/report.js

+// 把评估字段里的条目统一转成可读文本：LLM 可能返回对象（如风险
+// {type,finding,basis}）或 JSON 字符串，直接插值会显示成原始 JSON。
+function plainText(item) {
+  if (item && typeof item === 'object') {
+    const type = item.type || item.label || item.dimension || ''
+    const finding = item.finding || item.summary || item.text || item.detail || ''
+    const basis = item.basis || item.evidence || ''
+    let text = String(finding || '').trim()
+    const basisText = String(basis || '').trim()
+    if (basisText) text = text ? `${text}（依据：${basisText}）` : `依据：${basisText}`
+    if (type) text = text ? `【${type}】${text}` : `【${type}】`
+    return text || JSON.stringify(item)
+  }
+  const value = String(item ?? '').trim()
+  if (value.startsWith('{')) {
+    try {
+      const parsed = JSON.parse(value)
+      if (parsed && typeof parsed === 'object') return plainText(parsed)
+    } catch {
+      // 保持原文
+    }
+  }
+  return value
+}
+
 function listItems(value) {
-  return Array.isArray(value) ? value.filter((item) => String(item || '').trim()) : []
+  return Array.isArray(value) ? value.map(plainText).filter((item) => String(item || '').trim()) : []
 }
 
 // 需求 17：风险固定维度（跳槽频率 / 竞业协议 / 职业空窗 / 高层职业道德）

vue-app/src/views/ShareReportView.vue

 <script setup>
-// 分析报告公网分享页（ADR 0002）：免登录只读展示候选人 AI 初筛分析快照。
-// 数据来自后端匿名只读接口 /api/share/report/{reportId}?token=...；
-// 渲染逻辑复用 utils/report.js（与抽屉「评估」同源），快照不含电话/邮箱/原简历。
+// 分析报告公网分享页（ADR 0002）：免登录只读展示候选人分析报告快照。
+// 内容渲染交给 ReportContentView（v1/v2 通用）；本页负责外壳、加载/失效态与打印水印（P1 轻量 PDF）。
 import { computed, onMounted, ref } from 'vue'
 import { useRoute } from 'vue-router'
 
 import { fetchSharedReport } from '@/api/recruitment'
-import { providerLabel } from '@/utils/scoring'
-import { assessmentDecision, assessmentReportRows, assessmentScoreText, riskProfileRows } from '@/utils/report'
+import ReportContentView from '@/components/reports/ReportContentView.vue'
 
 const route = useRoute()
 
@@ -15,37 +13,15 @@ const loading = ref(true)
 const errorReason = ref('') // '' | 'disabled' | 'not_found'
 const report = ref(null)
 
-const snapshot = computed(() => report.value?.snapshot || {})
-const candidate = computed(() => ({
-  id: snapshot.value.candidateId,
-  name: snapshot.value.candidateName,
-  jobTitle: snapshot.value.jobTitle,
-  analysis: snapshot.value.analysis,
-  matchReport: snapshot.value.matchReport,
-  quickMatchScore: snapshot.value.quickMatchScore,
-}))
-
-const scoreText = computed(() => (report.value ? assessmentScoreText(candidate.value, 'ai') : ''))
-const decisionInfo = computed(() => (report.value ? assessmentDecision(candidate.value, 'ai') : {}))
-const reportRows = computed(() => (report.value ? assessmentReportRows(candidate.value, 'ai') : []))
-const riskProfile = computed(() => riskProfileRows(snapshot.value.analysis || {}))
-const author = computed(() => snapshot.value.author || {})
-const provider = computed(() => providerLabel(snapshot.value.analysis?.provider || ''))
-
-const toneTag = (level) =>
-  ({
-    ok: { type: 'success', text: '正常' },
-    hint: { type: 'warning', text: '提示' },
-    warn: { type: 'danger', text: '需确认' },
-  })[level] || { type: 'info', text: level }
-
-const formatTime = (iso = '') => {
+const author = computed(() => report.value?.snapshot?.author || {})
+const generatedDate = computed(() => {
+  const iso = report.value?.generatedAt || ''
   if (!iso) return ''
   const date = new Date(iso)
   if (Number.isNaN(date.getTime())) return iso
   const pad = (n) => String(n).padStart(2, '0')
   return `${date.getFullYear()}-${pad(date.getMonth() + 1)}-${pad(date.getDate())} ${pad(date.getHours())}:${pad(date.getMinutes())}`
-}
+})
 
 const printPage = () => window.print()
 
@@ -78,7 +54,7 @@ onMounted(async () => {
             <span>候选人 AI 初筛分析报告</span>
           </div>
         </div>
-        <el-button size="small" class="print-btn no-print" @click="printPage">
+        <el-button v-if="report" size="small" class="print-btn no-print" @click="printPage">
           <el-icon><Printer /></el-icon>
           打印 / 存 PDF
         </el-button>
@@ -97,69 +73,12 @@ onMounted(async () => {
       </div>
 
       <template v-else-if="report">
-        <section class="candidate-head">
-          <div class="name-row">
-            <h1>{{ snapshot.candidateName || '未命名候选人' }}</h1>
-            <el-tag size="small" effect="plain">{{ snapshot.jobTitle || '待匹配岗位' }}</el-tag>
-            <el-tag v-if="snapshot.source" size="small" type="info" effect="plain">{{ snapshot.source }}</el-tag>
-          </div>
-          <div class="meta-row">
-            <span>报告编号：{{ report.reportId }}</span>
-            <span>生成时间：{{ formatTime(report.generatedAt) }}</span>
-            <span>评估方式：{{ provider }}</span>
-            <span>生成人：{{ author.name || author.username || '—' }}</span>
-          </div>
-        </section>
-
-        <section class="summary-strip" :class="`tone-${decisionInfo.tone}`">
-          <div class="score-block">
-            <strong class="score-value">{{ scoreText }}</strong>
-            <span>综合得分</span>
-          </div>
-          <div class="decision-block">
-            <span class="eyebrow">初筛建议</span>
-            <strong>{{ decisionInfo.text }}</strong>
-            <p>{{ decisionInfo.note }}</p>
-          </div>
-        </section>
-
-        <section v-if="snapshot.analysis?.summary" class="block">
-          <h4>评估摘要</h4>
-          <p class="summary-text">{{ snapshot.analysis.summary }}</p>
-        </section>
-
-        <section v-if="riskProfile.length" class="block">
-          <h4>风险点评估</h4>
-          <div class="risk-list">
-            <div v-for="item in riskProfile" :key="item.dimension" class="risk-item" :class="`level-${item.level}`">
-              <el-tag :type="toneTag(item.level).type" size="small" effect="light">
-                {{ toneTag(item.level).text }}
-              </el-tag>
-              <b>{{ item.label }}</b>
-              <p>{{ item.summary }}</p>
-              <span v-for="(ev, index) in item.evidence || []" :key="`${item.dimension}-${index}`">依据：{{ ev }}</span>
-            </div>
-          </div>
-        </section>
-
-        <section v-if="reportRows.length" class="block">
-          <h4>评估明细</h4>
-          <div v-for="row in reportRows" :key="row.label" class="detail-card" :class="`tone-${row.tone}`">
-            <div class="detail-head">
-              <b>{{ row.label }}</b>
-              <el-tag :type="row.tone" size="small" effect="light">{{ row.status }}</el-tag>
-            </div>
-            <ul v-if="row.items.length">
-              <li v-for="(item, index) in row.items" :key="index">{{ item }}</li>
-            </ul>
-            <p v-else class="detail-empty">暂无</p>
-          </div>
-        </section>
-
-        <footer class="report-footer">
-          <p>本报告由 AI 辅助生成，基于生成时刻的简历与岗位信息冻结，仅供内部初筛参考，请以人工复核与面试为准。</p>
-          <p>不含候选人联系方式与原简历文件。</p>
-        </footer>
+        <ReportContentView :report="report" />
+        <div class="print-watermark no-screen" aria-hidden="true">
+          <div>内部资料 · 仅供内部初筛参考</div>
+          <div>{{ report.reportId }}</div>
+          <div v-if="author.name || author.username">{{ author.name || author.username }} · {{ generatedDate }}</div>
+        </div>
       </template>
     </main>
   </div>
@@ -224,6 +143,10 @@ onMounted(async () => {
   }
 }
 
+.no-screen {
+  display: none;
+}
+
 .state-box {
   padding: 72px 24px;
   text-align: center;
@@ -255,259 +178,6 @@ onMounted(async () => {
   }
 }
 
-.candidate-head {
-  padding: 18px 0 14px;
-
-  .name-row {
-    display: flex;
-    align-items: center;
-    flex-wrap: wrap;
-    gap: 8px;
-
-    h1 {
-      margin: 0;
-      font-size: 24px;
-    }
-  }
-
-  .meta-row {
-    display: flex;
-    flex-wrap: wrap;
-    gap: 6px 18px;
-    margin-top: 10px;
-    font-size: 12px;
-    color: var(--muted);
-  }
-}
-
-.summary-strip {
-  display: flex;
-  align-items: stretch;
-  gap: 0;
-  border: 1px solid var(--line);
-  border-radius: 12px;
-  overflow: hidden;
-  margin-bottom: 16px;
-
-  .score-block {
-    display: flex;
-    flex-direction: column;
-    align-items: center;
-    justify-content: center;
-    gap: 4px;
-    min-width: 168px;
-    padding: 16px 18px;
-    background: rgba(255, 255, 255, 0.8);
-
-    .score-value {
-      font-size: 30px;
-      line-height: 1;
-    }
-
-    span {
-      font-size: 12px;
-      color: var(--muted);
-    }
-  }
-
-  .decision-block {
-    flex: 1;
-    min-width: 0;
-    padding: 14px 18px;
-    display: flex;
-    flex-direction: column;
-    justify-content: center;
-    border-left: 1px solid var(--line);
-
-    .eyebrow {
-      font-size: 11px;
-      letter-spacing: 1px;
-      color: var(--muted);
-    }
-
-    strong {
-      margin: 4px 0 2px;
-      font-size: 17px;
-    }
-
-    p {
-      margin: 0;
-      font-size: 12px;
-      color: var(--muted);
-      line-height: 1.5;
-    }
-  }
-
-  &.tone-success {
-    border-color: color-mix(in srgb, var(--green) 55%, var(--line));
-
-    .score-value,
-    .decision-block strong {
-      color: var(--green);
-    }
-  }
-
-  &.tone-warning {
-    border-color: color-mix(in srgb, var(--yellow) 55%, var(--line));
-
-    .score-value,
-    .decision-block strong {
-      color: var(--yellow);
-    }
-  }
-
-  &.tone-danger {
-    border-color: color-mix(in srgb, var(--rose) 55%, var(--line));
-
-    .score-value,
-    .decision-block strong {
-      color: var(--rose);
-    }
-  }
-
-  &.tone-info {
-    .score-value,
-    .decision-block strong {
-      color: var(--blue);
-    }
-  }
-}
-
-.block {
-  margin-bottom: 16px;
-
-  h4 {
-    margin: 0 0 8px;
-    font-size: 14px;
-  }
-
-  .summary-text {
-    margin: 0;
-    padding: 12px 14px;
-    border: 1px solid var(--line);
-    border-radius: 10px;
-    background: rgba(255, 255, 255, 0.7);
-    font-size: 13px;
-    line-height: 1.7;
-    color: var(--ink);
-  }
-}
-
-.risk-list {
-  display: flex;
-  flex-direction: column;
-  gap: 8px;
-
-  .risk-item {
-    display: grid;
-    grid-template-columns: auto auto 1fr;
-    align-items: center;
-    gap: 4px 10px;
-    padding: 10px 12px;
-    border-radius: 10px;
-    border: 1px solid var(--line);
-    background: rgba(255, 255, 255, 0.6);
-
-    b {
-      font-size: 13px;
-    }
-
-    p {
-      grid-column: 1 / -1;
-      margin: 2px 0 0;
-      font-size: 13px;
-      line-height: 1.6;
-    }
-
-    span {
-      grid-column: 1 / -1;
-      font-size: 12px;
-      color: var(--muted);
-      line-height: 1.5;
-    }
-
-    &.level-warn {
-      border-color: color-mix(in srgb, var(--rose) 45%, var(--line));
-      background: color-mix(in srgb, var(--rose) 6%, #fff);
-    }
-
-    &.level-hint {
-      border-color: color-mix(in srgb, var(--yellow) 45%, var(--line));
-      background: color-mix(in srgb, var(--yellow) 7%, #fff);
-    }
-  }
-}
-
-.detail-card {
-  margin-bottom: 10px;
-  padding: 12px 14px;
-  border: 1px solid var(--line);
-  border-left-width: 3px;
-  border-radius: 10px;
-  background: rgba(255, 255, 255, 0.7);
-
-  &.tone-success {
-    border-left-color: var(--green);
-  }
-
-  &.tone-warning {
-    border-left-color: var(--yellow);
-  }
-
-  &.tone-danger {
-    border-left-color: var(--rose);
-  }
-
-  &.tone-info {
-    border-left-color: var(--blue);
-  }
-
-  .detail-head {
-    display: flex;
-    align-items: center;
-    justify-content: space-between;
-    gap: 10px;
-    margin-bottom: 6px;
-
-    b {
-      font-size: 14px;
-    }
-  }
-
-  ul {
-    margin: 0;
-    padding-left: 18px;
-    display: flex;
-    flex-direction: column;
-    gap: 5px;
-
-    li {
-      font-size: 13px;
-      line-height: 1.6;
-      color: var(--ink);
-    }
-  }
-
-  .detail-empty {
-    margin: 0;
-    font-size: 13px;
-    color: var(--muted);
-  }
-}
-
-.report-footer {
-  margin-top: 18px;
-  padding-top: 12px;
-  border-top: 1px dashed var(--line);
-  font-size: 12px;
-  color: var(--muted);
-  line-height: 1.6;
-
-  p {
-    margin: 2px 0;
-  }
-}
-
 @keyframes pulse {
   0%,
   100% {
@@ -522,20 +192,6 @@ onMounted(async () => {
 }
 
 @media (max-width: 620px) {
-  .summary-strip {
-    flex-direction: column;
-
-    .score-block {
-      min-width: 0;
-      padding: 14px;
-    }
-
-    .decision-block {
-      border-left: 0;
-      border-top: 1px solid var(--line);
-    }
-  }
-
   .report-head {
     align-items: flex-start;
   }
@@ -555,6 +211,27 @@ onMounted(async () => {
     box-shadow: none;
     border: 0;
     width: 100%;
+    max-width: none;
+  }
+
+  // 轻量水印（P1）：每页固定覆盖，避免截图外传
+  .print-watermark {
+    position: fixed;
+    inset: 0;
+    display: flex;
+    flex-direction: column;
+    align-items: center;
+    justify-content: center;
+    gap: 10px;
+    transform: rotate(-28deg);
+    font-size: 24px;
+    font-weight: 700;
+    letter-spacing: 3px;
+    color: #000;
+    opacity: 0.07;
+    pointer-events: none;
+    z-index: 9999;
+    white-space: nowrap;
   }
 }
 </style>