@hongmaple0820/med-scale-research-os 0.43.0

package/target-research-platform/src/dashboard/deliverables.py

@@ -0,0 +1,710 @@
+"""研究者面向的交付物目录与靶点对比数据提取。"""
+
+from __future__ import annotations
+
+import json
+import re
+from pathlib import Path
+from typing import Any
+
+
+# 研究者真正关心的交付物（按优先级）
+_DELIVERABLE_RULES: list[dict[str, Any]] = [
+    {
+        "pattern": re.compile(r".*_manuscript_zhongliu\.md$", re.I),
+        "category": "manuscript",
+        "title_zh": "中华肿瘤杂志格式稿",
+        "description_zh": "按《中华肿瘤杂志》排版的中文投稿草稿，含图表嵌入与逐篇文献评述。",
+        "priority": 1,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_manuscript\.md$", re.I),
+        "category": "manuscript",
+        "title_zh": "期刊论文草稿（Markdown）",
+        "description_zh": "中文 IMRaD 研究论文，含真实图表与文献深度解读。",
+        "priority": 2,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_manuscript\.docx$", re.I),
+        "category": "manuscript",
+        "title_zh": "期刊论文草稿（Word）",
+        "description_zh": "可下载的 Word 版稿件骨架，便于按目标期刊格式排版。",
+        "priority": 3,
+        "previewable": False,
+    },
+    {
+        "pattern": re.compile(r".*_manuscript_quality\.json$", re.I),
+        "category": "manuscript",
+        "title_zh": "论文质量门禁报告",
+        "description_zh": "自动检查稿件结构、边界声明、文献评述来源、图表与元数据完整性。",
+        "priority": 4,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_manuscript_fact_check\.json$", re.I),
+        "category": "manuscript",
+        "title_zh": "医学事实核查报告",
+        "description_zh": "核查 PMID、研究类型、关键数字、图表引用与医学边界表述。",
+        "priority": 5,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_literature_evidence_pack\.json$", re.I),
+        "category": "manuscript",
+        "title_zh": "文献证据富化包",
+        "description_zh": "可复用的 PubMed/OpenAlex/AMiner 文献证据缓存，支持离线复核与再次生成。",
+        "priority": 6,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_delivery_manifest\.json$", re.I),
+        "category": "manuscript",
+        "title_zh": "交付清单与溯源索引",
+        "description_zh": "列出交付包文件、hash、数据源、质量门禁、事实核查和人工审阅待办。",
+        "priority": 7,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_memory_summary\.md$", re.I),
+        "category": "manuscript",
+        "title_zh": "\u8bb0\u5fc6\u64ad\u79cd\u6458\u8981",
+        "description_zh": "\u9762\u5411 gbrain / scale-local \u7684\u533b\u5b66\u4ea4\u4ed8\u6458\u8981\uff0c\u6c47\u603b\u8d28\u91cf\u95e8\u7981\u3001\u4e8b\u5b9e\u6838\u67e5\u3001\u6587\u732e\u8bc1\u636e\u4e0e\u4eba\u5de5\u5ba1\u9605\u8981\u70b9\u3002",
+        "priority": 7.1,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_memory_seed\.json$", re.I),
+        "category": "manuscript",
+        "title_zh": "\u8bb0\u5fc6\u64ad\u79cd\u8bb0\u5f55",
+        "description_zh": "\u8bb0\u5f55\u8bb0\u5fc6\u4f9b\u5e94\u5546\u3001\u5df2\u5bfc\u5165\u6587\u4ef6\u3001\u7f3a\u5931\u89d2\u8272\u548c\u8b66\u544a\uff0c\u7528\u4e8e\u9a8c\u8bc1\u533b\u5b66\u4ea4\u4ed8\u5df2\u6210\u529f\u5199\u5165\u957f\u671f\u8bb0\u5fc6\u3002",
+        "priority": 7.2,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_memory_learning_candidate\.json$", re.I),
+        "category": "manuscript",
+        "title_zh": "\u533b\u5b66\u5b66\u4e60\u5019\u9009\u8bb0\u5f55",
+        "description_zh": "\u8bb0\u5f55\u8fd9\u6b21\u533b\u5b66\u4ea4\u4ed8\u662f\u5426\u53ef\u8fdb\u5165\u957f\u671f\u8bb0\u5fc6/\u77e5\u8bc6\u6c89\u6dc0\u5ba1\u9605\uff0c\u542b recommended action\u3001promotable \u548c Memory Brain \u5bfc\u5165\u72b6\u6001\u3002",
+        "priority": 7.25,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_memory_learning_candidate\.md$", re.I),
+        "category": "manuscript",
+        "title_zh": "\u533b\u5b66\u5b66\u4e60\u5019\u9009\u6458\u8981",
+        "description_zh": "\u9762\u5411\u4eba\u5de5\u590d\u5ba1\u7684 Markdown \u5b66\u4e60\u6458\u8981\uff0c\u6c47\u603b\u8d28\u91cf\u95e8\u7981\u3001\u4e8b\u5b9e\u6838\u67e5\u3001\u6587\u732e\u8bc1\u636e\u548c\u8b66\u544a\u3002",
+        "priority": 7.26,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_delivery\.zip$", re.I),
+        "category": "manuscript",
+        "title_zh": "医学研究交付 ZIP 包",
+        "description_zh": "面向研究者审阅的完整交付包，含论文、图表、元数据、质量报告和事实核查报告。",
+        "priority": 8,
+        "previewable": False,
+    },
+    {
+        "pattern": re.compile(r"figures/.*\.png$", re.I),
+        "category": "manuscript",
+        "title_zh": "论文图表（PNG）",
+        "description_zh": "表达热图、试验流程图、靶点排名图等发表级图表。",
+        "priority": 9,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r"report_.*\.html$", re.I),
+        "category": "evidence",
+        "title_zh": "可视化研究报告",
+        "description_zh": "整合四阶段结论的 HTML 报告，适合汇报与分享。",
+        "priority": 3,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r"report_.*\.docx$", re.I),
+        "category": "evidence",
+        "title_zh": "Word 研究报告",
+        "description_zh": "可下载的 Word 版完整报告，便于批注与流转。",
+        "priority": 4,
+        "previewable": False,
+    },
+    {
+        "pattern": re.compile(r".*_business_plan\.md$", re.I),
+        "category": "business",
+        "title_zh": "商业计划书",
+        "description_zh": "研发里程碑、成本估算与市场机会分析。",
+        "priority": 5,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_stage1_report\.md$", re.I),
+        "category": "expression",
+        "title_zh": "表达分析报告",
+        "description_zh": "GTEx / TCGA / HPA 多源表达与预后挖掘结论。",
+        "priority": 6,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_stage2_report\.md$", re.I),
+        "category": "screening",
+        "title_zh": "靶点筛选报告",
+        "description_zh": "候选靶点排名、成药性评分与差异化依据。",
+        "priority": 7,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_stage3_report\.md$", re.I),
+        "category": "research",
+        "title_zh": "深度调研报告",
+        "description_zh": "文献综述、临床试验格局与竞争强度分析。",
+        "priority": 8,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_milestones\.json$", re.I),
+        "category": "business",
+        "title_zh": "里程碑数据",
+        "description_zh": "各研发阶段周期、成功率与累计成本。",
+        "priority": 9,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r"pipeline_output\.json$", re.I),
+        "category": "summary",
+        "title_zh": "管线汇总数据",
+        "description_zh": "四阶段结构化汇总，供二次分析或 Agent 读取。",
+        "priority": 10,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_review_tasks\.json$", re.I),
+        "category": "manuscript",
+        "title_zh": "Agent \u6587\u732e\u8bc4\u8ff0\u5f85\u529e",
+        "description_zh": "\u5217\u51fa PMID\u3001\u7814\u7a76\u7c7b\u578b\u4e0e\u8bc4\u8ff0\u63d0\u793a\uff0c\u7528\u4e8e\u8865\u5145 Agent/LLM \u6df1\u5ea6\u8bc4\u8ff0\u3002",
+        "priority": 6.5,
+        "previewable": True,
+    },
+    {
+        "pattern": re.compile(r".*_agent_reviews\.json$", re.I),
+        "category": "manuscript",
+        "title_zh": "Agent \u6df1\u5ea6\u8bc4\u8ff0\u7f13\u5b58",
+        "description_zh": "Agent \u5df2\u5b8c\u6210\u7684\u9010\u7bc7\u4e2d\u6587\u8bc4\u8ff0\u7f13\u5b58\uff0c\u53ef\u76f4\u63a5\u56de\u5199\u8bba\u6587\u6b63\u6587\u3002",
+        "priority": 6.6,
+        "previewable": True,
+    },
+]
+
+_CATEGORY_LABELS = {
+    "manuscript": "期刊论文",
+    "evidence": "证据报告",
+    "business": "商业规划",
+    "expression": "表达分析",
+    "screening": "靶点筛选",
+    "research": "深度调研",
+    "summary": "汇总数据",
+}
+
+_ADVANCED_RULES: list[dict[str, Any]] = [
+    {
+        "pattern": re.compile(r".*_stage[1-4]_results\.json$", re.I),
+        "title_zh": "原始分析数据",
+        "description_zh": "完整 JSON 原始结果，供技术人员复核。",
+    },
+    {
+        "pattern": re.compile(r"provenance/.*\.json$", re.I),
+        "title_zh": "数据溯源记录",
+        "description_zh": "API 调用来源与请求参数，用于审计追溯。",
+    },
+]
+
+
+def _match_rule(filename: str, rel_path: str, rules: list[dict[str, Any]]) -> dict[str, Any] | None:
+    text = f"{rel_path}/{filename}".replace("\\", "/")
+    for rule in rules:
+        if rule["pattern"].search(filename) or rule["pattern"].search(text):
+            return rule
+    return None
+
+
+def _artifact_entry(file_path: Path, output_dir: Path, rule: dict[str, Any]) -> dict[str, Any]:
+    rel = str(file_path.relative_to(output_dir)).replace("\\", "/")
+    suffix = file_path.suffix.lower()
+    return {
+        "filename": file_path.name,
+        "path": str(file_path.resolve()),
+        "relative_path": rel,
+        "suffix": suffix,
+        "size": file_path.stat().st_size,
+        "category": rule.get("category", "advanced"),
+        "category_label": _CATEGORY_LABELS.get(rule.get("category", ""), "高级数据"),
+        "title_zh": rule["title_zh"],
+        "description_zh": rule["description_zh"],
+        "priority": rule.get("priority", 99),
+        "previewable": rule.get("previewable", suffix in {".md", ".html", ".json"}),
+        "downloadable": True,
+    }
+
+
+def _matches_current_run(file_path: Path, output_dir: Path, target: str | None, disease: str | None) -> bool:
+    if not target:
+        return True
+    rel = str(file_path.relative_to(output_dir)).replace("\\", "/").lower()
+    target_l = target.lower()
+    disease_l = (disease or "").lower().replace(" ", " ")
+    if target_l in rel:
+        return True
+    if disease_l and disease_l in rel:
+        return True
+    if rel.startswith("evidence_") and target_l in rel.replace("_", " "):
+        return True
+    return False
+
+
+def _dedupe_by_title(items: list[dict[str, Any]]) -> list[dict[str, Any]]:
+    best: dict[str, dict[str, Any]] = {}
+    for item in items:
+        key = item["title_zh"]
+        current = best.get(key)
+        if not current or item["size"] >= current["size"]:
+            best[key] = item
+    return sorted(best.values(), key=lambda item: (item["priority"], item["title_zh"]))
+
+
+def list_researcher_deliverables(
+    output_dir: Path,
+    *,
+    current_target: str | None = None,
+    current_disease: str | None = None,
+) -> dict[str, list[dict[str, Any]]]:
+    deliverables: list[dict[str, Any]] = []
+    advanced: list[dict[str, Any]] = []
+    if not output_dir.exists():
+        return {"deliverables": [], "advanced": [], "scope": "empty"}
+
+    for file_path in sorted(output_dir.rglob("*")):
+        if not file_path.is_file():
+            continue
+        if file_path.suffix.lower() not in {".json", ".md", ".html", ".pdf", ".docx", ".csv", ".png", ".svg", ".zip"}:
+            continue
+        rel = str(file_path.relative_to(output_dir)).replace("\\", "/")
+        if rel.startswith("dashboard_runtime/") or rel.startswith("cache/") or rel.startswith("demo/"):
+            continue
+
+        rule = _match_rule(file_path.name, rel, _DELIVERABLE_RULES)
+        if rule:
+            if _matches_current_run(file_path, output_dir, current_target, current_disease):
+                deliverables.append(_artifact_entry(file_path, output_dir, rule))
+            continue
+
+        adv_rule = _match_rule(file_path.name, rel, _ADVANCED_RULES)
+        if adv_rule and _matches_current_run(file_path, output_dir, current_target, current_disease):
+            advanced.append(_artifact_entry(file_path, output_dir, adv_rule))
+
+    deliverables = _dedupe_by_title(deliverables)
+    advanced = _dedupe_by_title(advanced)
+
+    scope = "current_run" if current_target else "all_runs"
+    return {"deliverables": deliverables, "advanced": advanced, "scope": scope}
+
+
+def _read_json(path: Path) -> dict[str, Any] | None:
+    try:
+        return json.loads(path.read_text(encoding="utf-8"))
+    except Exception:
+        return None
+
+
+def _find_stage_file(output_dir: Path, stage: str, target: str, disease: str) -> Path | None:
+    stage_dir = output_dir / stage
+    if not stage_dir.exists():
+        return None
+    candidates = [
+        stage_dir / f"{target}_{disease}_stage4_results.json" if stage == "stage4" else None,
+        stage_dir / f"{target}_stage1_results.json" if stage == "stage1" else None,
+        stage_dir / f"{target}_{disease}_stage3_results.json" if stage == "stage3" else None,
+        stage_dir / f"{disease}_stage2_results.json" if stage == "stage2" else None,
+    ]
+    for candidate in candidates:
+        if candidate and candidate.exists():
+            return candidate
+    return None
+
+
+def _extract_stage1_metrics(data: dict[str, Any]) -> dict[str, Any]:
+    summary = data.get("summary") or {}
+    validation = data.get("validation") or {}
+    return {
+        "expression_confidence": summary.get("confidence_level") or validation.get("confidence_level"),
+        "top_tissue": (summary.get("key_findings") or [None])[0],
+        "data_sources": ", ".join(summary.get("data_sources") or []),
+    }
+
+
+def _extract_stage2_metrics(data: dict[str, Any], target: str) -> dict[str, Any]:
+    summary = data.get("summary") or {}
+    top_candidates = summary.get("top_candidates") or []
+    target_row = next((row for row in top_candidates if row.get("gene") == target), None)
+    return {
+        "top_candidate": (top_candidates[0] or {}).get("gene") if top_candidates else None,
+        "target_rank": next((i + 1 for i, row in enumerate(top_candidates) if row.get("gene") == target), None),
+        "target_druggability": target_row.get("druggability_score") if target_row else None,
+        "target_score": target_row.get("score") if target_row else None,
+        "candidate_count": len(top_candidates),
+    }
+
+
+def _extract_stage3_metrics(data: dict[str, Any]) -> dict[str, Any]:
+    clinical = (((data.get("stages") or {}).get("clinical_pipeline") or {}).get("data") or {})
+    sections = clinical.get("sections") or {}
+    pipeline = sections.get("clinical_pipeline") or clinical
+    summary = pipeline.get("summary") or {}
+    phase_dist = summary.get("phase_distribution") or pipeline.get("phase_distribution") or {}
+    return {
+        "trial_count": summary.get("total_trials") or pipeline.get("total_trials"),
+        "active_trials": summary.get("active_trials") or pipeline.get("active_trials"),
+        "competition_intensity": summary.get("competition_intensity") or pipeline.get("competition_intensity"),
+        "max_clinical_phase": summary.get("max_phase") or pipeline.get("max_phase"),
+        "approved_drugs": summary.get("approved_drugs_count") or len(summary.get("approved_drugs") or []),
+    }
+
+
+def _extract_stage4_metrics(data: dict[str, Any]) -> dict[str, Any]:
+    ms = (((data.get("stages") or {}).get("milestones") or {}).get("data") or {})
+    adj = (ms.get("stage_adjustments") or {}).get("adjustment_details") or {}
+    competitive = adj.get("competitive_landscape") or {}
+    milestones = ms.get("milestones") or []
+    last = milestones[-1] if milestones else {}
+    return {
+        "overall_success_rate": ms.get("overall_success_rate"),
+        "total_duration_months": ms.get("total_duration_months"),
+        "total_cost_estimate": ms.get("total_cost_estimate"),
+        "estimated_completion": ms.get("estimated_completion"),
+        "druggability_score": adj.get("druggability_score"),
+        "expression_confidence_adj": adj.get("expression_confidence"),
+        "has_approved_drugs": competitive.get("has_approved_drugs"),
+        "competition_intensity": competitive.get("competition_intensity"),
+        "max_clinical_phase": competitive.get("max_clinical_phase"),
+        "final_phase_name": last.get("phase_name") or last.get("phase"),
+    }
+
+
+def discover_target_profiles(output_dir: Path) -> list[dict[str, Any]]:
+    profiles: list[dict[str, Any]] = []
+    stage4_dir = output_dir / "stage4"
+    if not stage4_dir.exists():
+        return profiles
+
+    for path in sorted(stage4_dir.glob("*_stage4_results.json")):
+        data = _read_json(path)
+        if not data:
+            continue
+        target = str(data.get("target") or path.name.split("_")[0])
+        disease = str(data.get("disease") or "unknown")
+        stage1 = _read_json(_find_stage_file(output_dir, "stage1", target, disease) or Path())
+        stage2 = _read_json(_find_stage_file(output_dir, "stage2", target, disease) or Path())
+        stage3 = _read_json(_find_stage_file(output_dir, "stage3", target, disease) or Path())
+
+        metrics = {
+            "stage1": _extract_stage1_metrics(stage1 or {}),
+            "stage2": _extract_stage2_metrics(stage2 or {}, target),
+            "stage3": _extract_stage3_metrics(stage3 or {}),
+            "stage4": _extract_stage4_metrics(data),
+        }
+        profiles.append({
+            "target": target,
+            "disease": disease,
+            "timestamp": data.get("timestamp"),
+            "path": str(path.resolve()),
+            "metrics": metrics,
+        })
+
+    return profiles
+
+
+def build_comparison_report(output_dir: Path) -> dict[str, Any]:
+    profiles = discover_target_profiles(output_dir)
+    if not profiles:
+        return {
+            "available": False,
+            "profiles": [],
+            "rows": [],
+            "recommendation": "暂无可对比靶点，请先完成至少一次完整分析。",
+        }
+
+    rows = [
+        ("表达置信度", "stage1", "expression_confidence", "text"),
+        ("主要表达发现", "stage1", "top_tissue", "text"),
+        ("靶点综合评分", "stage2", "target_score", "score"),
+        ("成药性评分", "stage2", "target_druggability", "score"),
+        ("候选排名", "stage2", "target_rank", "rank"),
+        ("临床试验总数", "stage3", "trial_count", "count"),
+        ("活跃试验数", "stage3", "active_trials", "count"),
+        ("竞争强度", "stage3", "competition_intensity", "text"),
+        ("最高临床阶段", "stage3", "max_clinical_phase", "phase"),
+        ("整体成功率", "stage4", "overall_success_rate", "percent"),
+        ("研发总周期(月)", "stage4", "total_duration_months", "months"),
+        ("总成本估算(百万美元)", "stage4", "total_cost_estimate", "money"),
+        ("预计完成时间", "stage4", "estimated_completion", "text"),
+        ("已有上市药物", "stage4", "has_approved_drugs", "bool"),
+    ]
+
+    recommendation = _build_recommendation(profiles)
+    return {
+        "available": len(profiles) >= 1,
+        "profiles": profiles,
+        "rows": [{"label_zh": label, "stage": stage, "key": key, "format": fmt} for label, stage, key, fmt in rows],
+        "recommendation": recommendation,
+    }
+
+
+def _build_recommendation(profiles: list[dict[str, Any]]) -> str:
+    if len(profiles) == 1:
+        p = profiles[0]
+        m4 = p["metrics"]["stage4"]
+        m2 = p["metrics"]["stage2"]
+        return (
+            f"当前仅分析 {p['target']}（{p['disease']}）。"
+            f"成药性评分 {m2.get('target_druggability', '—')}，"
+            f"整体成功率 { _fmt_percent(m4.get('overall_success_rate')) }，"
+            f"建议再分析 1–2 个对照靶点后使用对比功能。"
+        )
+
+    def score(profile: dict[str, Any]) -> float:
+        m2 = profile["metrics"]["stage2"]
+        m4 = profile["metrics"]["stage4"]
+        drug = float(m2.get("target_druggability") or 0)
+        success = float(m4.get("overall_success_rate") or 0)
+        return drug * 0.6 + success * 0.4
+
+    ranked = sorted(profiles, key=score, reverse=True)
+    best = ranked[0]
+    second = ranked[1]
+    return (
+        f"综合成药性与发展成功率，{best['target']}（{best['disease']}）当前优先级更高；"
+        f"对照靶点 {second['target']} 可作为差异化或风险对照继续深挖。"
+    )
+
+
+def _fmt_percent(value: Any) -> str:
+    if value is None:
+        return "—"
+    try:
+        return f"{float(value) * 100:.2f}%"
+    except (TypeError, ValueError):
+        return str(value)
+
+
+def discover_latest_run(output_dir: Path) -> tuple[str | None, str | None]:
+    stage4 = output_dir / "stage4"
+    if not stage4.exists():
+        return None, None
+    candidates = list(stage4.glob("*_business_plan.md"))
+    if not candidates:
+        candidates = list(stage4.glob("*_stage4_results.json"))
+    if not candidates:
+        return None, None
+    latest = max(candidates, key=lambda p: p.stat().st_mtime)
+    stem = latest.stem
+    if stem.endswith("_business_plan"):
+        base = stem[: -len("_business_plan")]
+    elif stem.endswith("_stage4_results"):
+        base = stem[: -len("_stage4_results")]
+    else:
+        return None, None
+    if "_" in base:
+        target, disease = base.split("_", 1)
+        return target, disease.replace("_", " ")
+    return base, None
+
+
+def _find_report_file(output_dir: Path, patterns: list[str]) -> str | None:
+    for pattern in patterns:
+        matches = sorted(output_dir.glob(pattern), key=lambda p: p.stat().st_mtime, reverse=True)
+        if matches:
+            return str(matches[0].resolve())
+    return None
+
+
+def build_report_bundle(
+    output_dir: Path,
+    *,
+    current_target: str | None = None,
+    current_disease: str | None = None,
+) -> dict[str, Any]:
+    """构建在线阅读用的分析报告包（非学术论文）。"""
+    target = current_target or None
+    disease = current_disease or None
+    if not target:
+        target, disease = discover_latest_run(output_dir)
+
+    if not target:
+        return {
+            "available": False,
+            "target": None,
+            "disease": None,
+            "note": "暂无可读报告。请先运行一次完整分析。",
+            "sections": [],
+            "summary_cards": [],
+        }
+
+    disease_token = (disease or "").replace(" ", " ")
+    sections = [
+        {
+            "id": "stage1",
+            "title_zh": "表达分析",
+            "subtitle_zh": "多源表达与预后挖掘（GTEx / TCGA / HPA）",
+            "kind": "markdown",
+            "path": _find_report_file(output_dir, [
+                f"stage1/{target}_stage1_report.md",
+                f"stage1/*{target}*stage1_report.md",
+            ]),
+        },
+        {
+            "id": "stage2",
+            "title_zh": "靶点筛选",
+            "subtitle_zh": "候选靶点排名与成药性评分",
+            "kind": "markdown",
+            "path": _find_report_file(output_dir, [
+                f"stage2/{disease_token}_stage2_report.md",
+                f"stage2/*{disease_token}*stage2_report.md",
+                "stage2/*_stage2_report.md",
+            ]),
+        },
+        {
+            "id": "stage3",
+            "title_zh": "深度调研",
+            "subtitle_zh": "文献、临床试验与竞争格局（不是学术论文）",
+            "kind": "markdown",
+            "path": _find_report_file(output_dir, [
+                f"stage3/{target}_{disease_token}_stage3_report.md",
+                f"stage3/*{target}*stage3_report.md",
+            ]),
+        },
+        {
+            "id": "stage4",
+            "title_zh": "商业规划",
+            "subtitle_zh": "里程碑、成功率与成本估算",
+            "kind": "markdown",
+            "path": _find_report_file(output_dir, [
+                f"stage4/{target}_{disease_token}_business_plan.md",
+                f"stage4/*{target}*business_plan.md",
+            ]),
+        },
+        {
+            "id": "evidence",
+            "title_zh": "完整可视化报告",
+            "subtitle_zh": "四阶段整合 HTML 证据包",
+            "kind": "html",
+            "path": _find_report_file(output_dir, [
+                f"evidence_{target}_{disease_token.replace(' ', '_')}/report_{target}.html",
+                f"evidence_*/report_{target}.html",
+                "evidence_*/report_*.html",
+            ]),
+        },
+        {
+            "id": "manuscript",
+            "title_zh": "期刊论文草稿",
+            "subtitle_zh": "中文 IMRaD 结构稿件（含结构化摘要、图表与逐篇文献评述，投稿前需人工审校）",
+            "kind": "markdown",
+            "path": _find_report_file(output_dir, [
+                f"manuscripts/{target}_{disease_token}_manuscript.md",
+                f"manuscripts/{target}_{disease_token.replace(' ', '_')}_manuscript.md",
+                "manuscripts/*_manuscript.md",
+            ]),
+        },
+    ]
+
+    stage4_path = _find_report_file(output_dir, [
+        f"stage4/{target}_{disease_token}_stage4_results.json",
+        f"stage4/*{target}*stage4_results.json",
+    ])
+    stage4_json = _read_json(Path(stage4_path)) if stage4_path else None
+
+    summary_cards: list[dict[str, str]] = []
+    if stage4_json:
+        ms = (((stage4_json.get("stages") or {}).get("milestones") or {}).get("data") or {})
+        adj = (ms.get("stage_adjustments") or {}).get("adjustment_details") or {}
+        competitive = adj.get("competitive_landscape") or {}
+        summary_cards = [
+            {"label": "靶点", "value": target},
+            {"label": "疾病", "value": disease or "—"},
+            {"label": "成药性评分", "value": str(adj.get("druggability_score", "—"))},
+            {"label": "整体成功率", "value": _fmt_percent(ms.get("overall_success_rate"))},
+            {"label": "研发周期", "value": f"{ms.get('total_duration_months', '—')} 个月"},
+            {"label": "成本估算", "value": f"${ms.get('total_cost_estimate', '—')}M"},
+            {"label": "竞争强度", "value": str(competitive.get("competition_intensity", "—"))},
+            {"label": "已有上市药", "value": "是" if competitive.get("has_approved_drugs") else "否"},
+        ]
+
+    available_sections = [s for s in sections if s.get("path")]
+    return {
+        "available": bool(available_sections),
+        "target": target,
+        "disease": disease,
+        "note": (
+            "包含四阶段靶点分析汇报与期刊论文草稿。"
+            "论文草稿为基于公开数据的 IMRaD 骨架，不可直接投稿，需补充实验与作者信息。"
+        ),
+        "sections": sections,
+        "summary_cards": summary_cards,
+    }
+
+
+def summarize_json_preview(filename: str, payload: dict[str, Any]) -> list[dict[str, str]]:
+    """将 JSON 原始数据提炼为研究者可读的摘要条目。"""
+    rows: list[dict[str, str]] = []
+    lower = filename.lower()
+
+    if "business_plan" in lower or "stage4" in lower:
+        ms = (((payload.get("stages") or {}).get("milestones") or {}).get("data") or payload)
+        rows.extend([
+            {"label": "靶点", "value": str(payload.get("target") or ms.get("target") or "—")},
+            {"label": "疾病", "value": str(payload.get("disease") or ms.get("disease") or "—")},
+            {"label": "整体成功率", "value": _fmt_percent(ms.get("overall_success_rate"))},
+            {"label": "研发周期", "value": f"{ms.get('total_duration_months', '—')} 个月"},
+            {"label": "成本估算", "value": f"${ms.get('total_cost_estimate', '—')}M"},
+        ])
+        return rows
+
+    if "stage2" in lower:
+        summary = payload.get("summary") or {}
+        top = (summary.get("top_candidates") or [{}])[0]
+        rows.extend([
+            {"label": "疾病", "value": str(summary.get("disease") or payload.get("disease") or "—")},
+            {"label": "Top 候选", "value": str(top.get("gene") or "—")},
+            {"label": "Top 评分", "value": str(top.get("score") or "—")},
+            {"label": "成药性", "value": str(top.get("druggability_score") or "—")},
+        ])
+        return rows
+
+    if "stage3" in lower:
+        clinical = (((payload.get("stages") or {}).get("clinical_pipeline") or {}).get("data") or {})
+        summary = (clinical.get("sections") or {}).get("clinical_pipeline", clinical).get("summary") or clinical.get("summary") or {}
+        rows.extend([
+            {"label": "靶点", "value": str(payload.get("target") or "—")},
+            {"label": "临床试验数", "value": str(summary.get("total_trials") or "—")},
+            {"label": "活跃试验", "value": str(summary.get("active_trials") or "—")},
+            {"label": "竞争强度", "value": str(summary.get("competition_intensity") or "—")},
+        ])
+        return rows
+
+    if "stage1" in lower or payload.get("gene"):
+        summary = payload.get("summary") or {}
+        findings = summary.get("key_findings") or []
+        rows.extend([
+            {"label": "基因", "value": str(payload.get("gene") or "—")},
+            {"label": "置信度", "value": str(summary.get("confidence_level") or "—")},
+            {"label": "关键发现", "value": str(findings[0] if findings else "—")},
+        ])
+        return rows
+
+    if "milestone" in lower:
+        rows.append({"label": "里程碑数", "value": str(len(payload.get("milestones") or []))})
+        return rows
+
+    for key in ("target", "disease", "gene", "success", "status"):
+        if key in payload:
+            rows.append({"label": key, "value": str(payload[key])})
+    return rows[:8]