Complete the business-export chain by splitting manifest-ready rows into role-specific lists

Constraint: Keep this checkpoint offline-only and avoid touching real business data, datasets, or model artifacts
Rejected: Leave role splitting as a manual next-session step | The export chain is more usable when reference/query/excluded lists are produced automatically
Confidence: high
Scope-risk: narrow
Directive: Treat the split outputs as staging lists and keep final project-manifest adaptation explicit in the downstream integration step
Tested: Normalized the sample CSV, ran split_business_manifest_ready.py, verified 1 reference + 1 query + 1 excluded row, and rechecked 73 relative links
Not-tested: Did not run against a live business export or feed the split outputs into the full training pipeline

@@ -59,6 +59,7 @@
    - Manifest 规范：`docs/business-manifest-and-type-role-spec.md`
    - Manifest 规范：`docs/business-manifest-and-type-role-spec.md`
    - 导出 cookbook：`docs/business-export-cookbook.md`
    - 导出 cookbook：`docs/business-export-cookbook.md`
    - 规范化脚本：`acr-engine/scripts/normalize_business_export.py`
    - 规范化脚本：`acr-engine/scripts/normalize_business_export.py`
+   - 角色拆分脚本：`acr-engine/scripts/split_business_manifest_ready.py`
 2. 补 cap64 multi-seed aggregate。
 2. 补 cap64 multi-seed aggregate。
 3. 更新：
 3. 更新：
    - `docs/open-dataset-workflow.md`
    - `docs/open-dataset-workflow.md`

+#!/usr/bin/env python3
+from __future__ import annotations
+
+import argparse
+import json
+from collections import Counter
+from pathlib import Path
+
+
+def load_jsonl(path: Path) -> list[dict]:
+    return [json.loads(line) for line in path.read_text().splitlines() if line.strip()]
+
+
+def write_json(path: Path, rows: list[dict]) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    path.write_text(json.dumps(rows, ensure_ascii=False, indent=2))
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description='Split manifest-ready business JSONL into reference/query/excluded JSON files')
+    parser.add_argument('--input', required=True)
+    parser.add_argument('--output-dir', required=True)
+    args = parser.parse_args()
+
+    input_path = Path(args.input).resolve()
+    output_dir = Path(args.output_dir).resolve()
+    rows = load_jsonl(input_path)
+
+    grouped = {'reference': [], 'query': [], 'excluded': []}
+    for row in rows:
+        role = row.get('role', 'excluded')
+        grouped.setdefault(role, []).append(row)
+
+    write_json(output_dir / 'reference.json', grouped.get('reference', []))
+    write_json(output_dir / 'query.json', grouped.get('query', []))
+    write_json(output_dir / 'excluded.json', grouped.get('excluded', []))
+
+    summary = {
+        'input_rows': len(rows),
+        'role_counts': dict(Counter(row.get('role', 'excluded') for row in rows)),
+        'outputs': {
+            'reference': str((output_dir / 'reference.json').resolve()),
+            'query': str((output_dir / 'query.json').resolve()),
+            'excluded': str((output_dir / 'excluded.json').resolve()),
+        },
+    }
+    print(json.dumps(summary, ensure_ascii=False, indent=2))
+
+
+if __name__ == '__main__':
+    main()

+## 2026-06-02 manifest-ready 角色拆分脚本交付 checkpoint
+
+完成项：
+- 新增 `acr-engine/scripts/split_business_manifest_ready.py`
+- 已把业务规范化输出继续推进为 `reference/query/excluded` 三类 JSON 清单。
+
+结论：
+- 下个 session 从业务导出到角色拆分已经形成连续脚本链路。
+- 后续只需要补最终项目 manifest 适配，而不必再手工分角色。
+
 ## 2026-06-02 业务导出规范化脚本交付 checkpoint
 ## 2026-06-02 业务导出规范化脚本交付 checkpoint
 
 
 完成项：
 完成项：

@@ -121,3 +121,25 @@ cd /workspace/acr-engine
 2. 应用 `business_type_role_mapping.json`
 2. 应用 `business_type_role_mapping.json`
 3. 自动补 `role / bucket / source_dataset / split` 默认值
 3. 自动补 `role / bucket / source_dataset / split` 默认值
 4. 输出 manifest-ready JSONL
 4. 输出 manifest-ready JSONL
+
+
+## 7. 拆分为角色清单
+
+如果你已经拿到了 manifest-ready JSONL，还可以继续用：
+- [../acr-engine/scripts/split_business_manifest_ready.py](../acr-engine/scripts/split_business_manifest_ready.py)
+
+示例：
+
+```bash
+cd /workspace/acr-engine
+/usr/local/miniconda3/bin/python scripts/split_business_manifest_ready.py \
+  --input /tmp/business_asset_manifest_ready.jsonl \
+  --output-dir /tmp/business_asset_manifest_split
+```
+
+它会输出：
+- `reference.json`
+- `query.json`
+- `excluded.json`
+
+这样下个 session 可以更快把业务素材继续整形成训练/评测所需清单。

@@ -79,6 +79,8 @@ flowchart LR
   - [../acr-engine/scripts/print_business_type_mapping.py](../acr-engine/scripts/print_business_type_mapping.py)
   - [../acr-engine/scripts/print_business_type_mapping.py](../acr-engine/scripts/print_business_type_mapping.py)
 - 规范化脚本：
 - 规范化脚本：
   - [../acr-engine/scripts/normalize_business_export.py](../acr-engine/scripts/normalize_business_export.py)
   - [../acr-engine/scripts/normalize_business_export.py](../acr-engine/scripts/normalize_business_export.py)
+- 角色拆分脚本：
+  - [../acr-engine/scripts/split_business_manifest_ready.py](../acr-engine/scripts/split_business_manifest_ready.py)
 
 
 示例命令：
 示例命令：
 
 

@@ -260,6 +260,7 @@
    - Manifest/角色映射看：[business-manifest-and-type-role-spec.md](./business-manifest-and-type-role-spec.md)
    - Manifest/角色映射看：[business-manifest-and-type-role-spec.md](./business-manifest-and-type-role-spec.md)
    - SQL/CSV/JSONL 导出参考：[business-export-cookbook.md](./business-export-cookbook.md)
    - SQL/CSV/JSONL 导出参考：[business-export-cookbook.md](./business-export-cookbook.md)
    - 规范化脚本：`acr-engine/scripts/normalize_business_export.py`
    - 规范化脚本：`acr-engine/scripts/normalize_business_export.py`
+   - 角色拆分脚本：`acr-engine/scripts/split_business_manifest_ready.py`
 2. 对比 cap48 与 cap64 的不一致现象，补充分规模结论。
 2. 对比 cap48 与 cap64 的不一致现象，补充分规模结论。
 3. 继续补 cap64 multi-seed，而不是只保留单 seed。
 3. 继续补 cap64 multi-seed，而不是只保留单 seed。
 4. 继续优化 `hybrid`，重点降低波动并提升 hard case 稳定性。
 4. 继续优化 `hybrid`，重点降低波动并提升 hard case 稳定性。