wkc/ccdi

Files

2026-03-20 14:52:08 +08:00

23 KiB

Raw Blame History

LSFX Mock Random Hit Rule Backend Implementation Plan

For agentic workers: REQUIRED: Use superpowers:subagent-driven-development (if subagents available) or superpowers:executing-plans to implement this plan. Steps use checkbox (- [ ]) syntax for tracking.

Goal: 改造 lsfx-mock-server 的行内流水生成链路，使大额交易与第一期新增规则都按 logId 稳定随机命中一部分，同时补齐 LARGE_PURCHASE_TRANSACTION 所需的最小数据库基线数据。

Architecture: 保持现有 FileService -> StatementService -> 缓存分页 主链路不变，不新增平行模块或兼容性双轨。FileService 只负责生成并保存规则命中计划，statement_rule_samples.py 只负责按规则代码装配命中样本，StatementService 负责把命中样本与噪声流水合并后缓存；LARGE_PURCHASE_TRANSACTION 不伪造成银行流水，单独通过 SQL 基线脚本补齐真实来源表数据。

Tech Stack: Python 3, FastAPI, pytest, PyMySQL, MySQL, Bash

File Structure

lsfx-mock-server/services/file_service.py: 在 FileRecord 中保存稳定随机的规则命中计划，并让上传链路与行内流水链路都具备一致的计划字段。
lsfx-mock-server/services/statement_rule_samples.py: 维护大额交易与第一期新增规则的样本 builder 映射，按命中计划生成最小命中样本集合。
lsfx-mock-server/services/statement_service.py: 从 FileRecord 读取规则命中计划，拼装命中样本与噪声流水，统一分配 ID 并缓存。
lsfx-mock-server/tests/test_file_service.py: 锁定命中计划的生成与持久化语义。
lsfx-mock-server/tests/test_statement_service.py: 锁定样本装配、缓存稳定性与按规则子集命中。
lsfx-mock-server/tests/integration/test_full_workflow.py: 验证 getJZFileOrZjrcuFile -> getBSByLogId 端到端链路。
sql/migration/2026-03-20-lsfx-mock-random-hit-rule-purchase-baseline.sql: 为 LARGE_PURCHASE_TRANSACTION 补最小采购业务数据。
docs/reports/implementation/2026-03-20-lsfx-mock-random-hit-rule-backend-record.md: 记录本次后端实施内容。
docs/tests/records/2026-03-20-lsfx-mock-random-hit-rule-backend-verification.md: 记录测试与数据库验证结果。

Task 1: 在 FileService 中持久化稳定随机的规则命中计划

Files:

Modify: lsfx-mock-server/services/file_service.py
Modify: lsfx-mock-server/tests/test_file_service.py
Reference: docs/design/2026-03-20-lsfx-mock-random-hit-rule-design.md
Step 1: Write the failing test

在 lsfx-mock-server/tests/test_file_service.py 中先补两条最小失败用例，锁定“同一 logId 命中计划稳定”和“fetch_inner_flow() 会把命中计划落到 FileRecord”：

def test_build_rule_hit_plan_should_be_deterministic_for_same_log_id():
    service = FileService(staff_identity_repository=FakeStaffIdentityRepository())

    plan1 = service._build_rule_hit_plan(10001)
    plan2 = service._build_rule_hit_plan(10001)

    assert plan1 == plan2
    assert 2 <= len(plan1["large_transaction_hit_rules"]) <= 4
    assert 2 <= len(plan1["phase1_hit_rules"]) <= 4


def test_fetch_inner_flow_should_persist_rule_hit_plan(monkeypatch):
    service = FileService(staff_identity_repository=FakeStaffIdentityRepository())
    monkeypatch.setattr(
        service,
        "_build_rule_hit_plan",
        lambda log_id: {
            "large_transaction_hit_rules": ["HOUSE_OR_CAR_EXPENSE", "TAX_EXPENSE"],
            "phase1_hit_rules": ["GAMBLING_SENSITIVE_KEYWORD", "FOREX_BUY_AMT"],
        },
    )

    response = service.fetch_inner_flow(
        {
            "groupId": 1001,
            "customerNo": "test_customer_001",
            "dataChannelCode": "test_code",
            "requestDateId": 20240101,
            "dataStartDateId": 20240101,
            "dataEndDateId": 20240131,
            "uploadUserId": 902001,
        }
    )
    log_id = response["data"][0]
    record = service.file_records[log_id]

    assert record.large_transaction_hit_rules == ["HOUSE_OR_CAR_EXPENSE", "TAX_EXPENSE"]
    assert record.phase1_hit_rules == ["GAMBLING_SENSITIVE_KEYWORD", "FOREX_BUY_AMT"]

Step 2: Run test to verify it fails

Run:

cd lsfx-mock-server
pytest tests/test_file_service.py -k "rule_hit_plan or persist_rule_hit_plan" -v

Expected:

FAIL
原因是 FileRecord 尚未保存规则命中计划，FileService 也没有 _build_rule_hit_plan() 能力
Step 3: Write minimal implementation

在 lsfx-mock-server/services/file_service.py 中只做最小改动：

为 FileRecord 新增两个字段，采用现有 Python 代码风格的 snake_case：

large_transaction_hit_rules: List[str] = field(default_factory=list)
phase1_hit_rules: List[str] = field(default_factory=list)

新增固定规则池与稳定随机 helper：

def _build_rule_hit_plan(self, log_id: int) -> dict:
    rng = random.Random(f"rule-plan:{log_id}")
    return {
        "large_transaction_hit_rules": self._pick_rule_subset(rng, LARGE_TRANSACTION_RULE_CODES, 2, 4),
        "phase1_hit_rules": self._pick_rule_subset(rng, PHASE1_RULE_CODES, 2, 4),
    }

在 _create_file_record()、upload_file()、fetch_inner_flow() 中都把规则命中计划写入 FileRecord，避免 getBSByLogId 读取上传链路记录时出现字段缺失。

Step 4: Run test to verify it passes

Run:

cd lsfx-mock-server
pytest tests/test_file_service.py -k "rule_hit_plan or persist_rule_hit_plan" -v

Expected:

PASS
FileRecord 已具备稳定随机的规则命中计划
Step 5: Commit

git add lsfx-mock-server/services/file_service.py lsfx-mock-server/tests/test_file_service.py
git commit -m "持久化Mock随机命中规则计划"

Task 2: 将 statement_rule_samples.py 拆成按规则代码装配的样本生成器

Files:

Modify: lsfx-mock-server/services/statement_rule_samples.py
Modify: lsfx-mock-server/tests/test_statement_service.py
Step 1: Write the failing test

在 lsfx-mock-server/tests/test_statement_service.py 中补两条失败用例，锁定“只生成被选中的规则样本，而不是默认全量命中”：

def test_build_seed_statements_for_rule_plan_should_only_include_requested_phase1_rules():
    plan = {
        "large_transaction_hit_rules": [],
        "phase1_hit_rules": ["GAMBLING_SENSITIVE_KEYWORD", "FOREX_BUY_AMT"],
    }

    statements = build_seed_statements_for_rule_plan(
        group_id=1000,
        log_id=20001,
        rule_plan=plan,
    )

    assert any("游戏" in item["userMemo"] for item in statements)
    assert any("购汇" in item["userMemo"] for item in statements)
    assert not any("证券" in item["userMemo"] for item in statements)


def test_build_seed_statements_for_rule_plan_should_generate_withdraw_cnt_samples():
    plan = {
        "large_transaction_hit_rules": [],
        "phase1_hit_rules": ["WITHDRAW_CNT"],
    }

    statements = build_seed_statements_for_rule_plan(
        group_id=1000,
        log_id=20001,
        rule_plan=plan,
    )

    assert len([
        item for item in statements
        if "微信提现" in item["userMemo"] or "支付宝提现" in item["userMemo"]
    ]) >= 4

Step 2: Run test to verify it fails

Run:

cd lsfx-mock-server
pytest tests/test_statement_service.py -k "rule_plan_should_only_include or withdraw_cnt_samples" -v

Expected:

FAIL
原因是当前样本模块仍然是固定大额交易全量样本，没有“按规则计划装配”入口
Step 3: Write minimal implementation

在 lsfx-mock-server/services/statement_rule_samples.py 中做结构性改造，但只落最短路径：

保留现有基础 identity 与 _build_statement() 工具函数
将大额交易样本拆成按规则代码可调用的 builder
新增第一期规则 builder 映射
提供统一入口 build_seed_statements_for_rule_plan(...)

最小结构如下：

LARGE_TRANSACTION_BUILDERS = {
    "HOUSE_OR_CAR_EXPENSE": build_house_or_car_samples,
    "TAX_EXPENSE": build_tax_samples,
    "SINGLE_LARGE_INCOME": build_single_large_income_samples,
    "CUMULATIVE_INCOME": build_cumulative_income_samples,
    "ANNUAL_TURNOVER": build_annual_turnover_supporting_samples,
    "LARGE_CASH_DEPOSIT": build_large_cash_deposit_samples,
    "FREQUENT_CASH_DEPOSIT": build_frequent_cash_deposit_samples,
    "LARGE_TRANSFER": build_large_transfer_samples,
}

PHASE1_RULE_BUILDERS = {
    "GAMBLING_SENSITIVE_KEYWORD": build_gambling_sensitive_keyword_samples,
    "SPECIAL_AMOUNT_TRANSACTION": build_special_amount_transaction_samples,
    "SUSPICIOUS_INCOME_KEYWORD": build_suspicious_income_keyword_samples,
    "FOREX_BUY_AMT": build_forex_buy_samples,
    "FOREX_SELL_AMT": build_forex_sell_samples,
    "STOCK_TFR_LARGE": build_stock_transfer_large_samples,
    "LARGE_STOCK_TRADING": build_large_stock_trading_samples,
    "WITHDRAW_CNT": build_withdraw_cnt_samples,
}


def build_seed_statements_for_rule_plan(group_id, log_id, rule_plan, **kwargs):
    statements = []
    for rule_code in rule_plan["large_transaction_hit_rules"]:
        statements.extend(LARGE_TRANSACTION_BUILDERS[rule_code](group_id, log_id, **kwargs))
    for rule_code in rule_plan["phase1_hit_rules"]:
        statements.extend(PHASE1_RULE_BUILDERS[rule_code](group_id, log_id, **kwargs))
    return statements

要求：

不再默认返回全量命中样本
每个 builder 只构造命中该规则所需的最小流水集合
WITHDRAW_CNT 虽然是对象型规则，但依然通过银行流水样本构造触发条件
Step 4: Run test to verify it passes

Run:

cd lsfx-mock-server
pytest tests/test_statement_service.py -k "rule_plan_should_only_include or withdraw_cnt_samples" -v

Expected:

PASS
样本模块已按规则子集装配，不再默认全量命中
Step 5: Commit

git add lsfx-mock-server/services/statement_rule_samples.py lsfx-mock-server/tests/test_statement_service.py
git commit -m "拆分Mock规则样本构造器"

Task 3: 让 StatementService 按命中计划生成样本并保持缓存稳定

Files:

Modify: lsfx-mock-server/services/statement_service.py
Modify: lsfx-mock-server/tests/test_statement_service.py
Modify: lsfx-mock-server/tests/integration/test_full_workflow.py
Step 1: Write the failing test

先在 lsfx-mock-server/tests/test_statement_service.py 补一条服务层测试，明确 StatementService 必须读取 FileRecord 中的命中计划，而不是继续默认全量样本：

def test_generate_statements_should_follow_rule_hit_plan_from_file_record():
    file_service = FileService(staff_identity_repository=FakeStaffIdentityRepository())
    statement_service = StatementService(file_service=file_service)

    response = file_service.fetch_inner_flow(
        {
            "groupId": 1001,
            "customerNo": "customer_rule_plan",
            "dataChannelCode": "test_code",
            "requestDateId": 20240101,
            "dataStartDateId": 20240101,
            "dataEndDateId": 20240131,
            "uploadUserId": 902001,
        }
    )
    log_id = response["data"][0]
    record = file_service.file_records[log_id]
    record.large_transaction_hit_rules = ["HOUSE_OR_CAR_EXPENSE"]
    record.phase1_hit_rules = ["GAMBLING_SENSITIVE_KEYWORD"]

    statements = statement_service._generate_statements(group_id=1001, log_id=log_id, count=200)

    assert any("房产首付款" in item["userMemo"] for item in statements)
    assert any("游戏" in item["userMemo"] for item in statements)
    assert not any("购汇" in item["userMemo"] for item in statements)

再在 lsfx-mock-server/tests/integration/test_full_workflow.py 补一条链路测试，锁定同一 logId 首次生成后分页稳定：

def test_inner_flow_bank_statement_should_keep_same_rule_subset(client):
    fetch_response = client.post(
        "/watson/api/project/getJZFileOrZjrcuFile",
        data={
            "groupId": 1001,
            "customerNo": "customer_subset",
            "dataChannelCode": "test_code",
            "requestDateId": 20240101,
            "dataStartDateId": 20240101,
            "dataEndDateId": 20240131,
            "uploadUserId": 902001,
        },
    )
    log_id = fetch_response.json()["data"][0]

    page1 = client.post(
        "/watson/api/project/getBSByLogId",
        data={"groupId": 1001, "logId": log_id, "pageNow": 1, "pageSize": 10},
    ).json()
    page2 = client.post(
        "/watson/api/project/getBSByLogId",
        data={"groupId": 1001, "logId": log_id, "pageNow": 1, "pageSize": 10},
    ).json()

    assert page1["data"]["bankStatementList"] == page2["data"]["bankStatementList"]

Step 2: Run test to verify it fails

Run:

cd lsfx-mock-server
pytest tests/test_statement_service.py -k "follow_rule_hit_plan" -v
pytest tests/integration/test_full_workflow.py -k "same_rule_subset" -v

Expected:

FAIL
原因是 StatementService 还没有从 FileRecord 读取规则命中计划，也没有按子集装配样本
Step 3: Write minimal implementation

在 lsfx-mock-server/services/statement_service.py 中接通规则命中计划：

from services.statement_rule_samples import build_seed_statements_for_rule_plan


def _generate_statements(self, group_id: int, log_id: int, count: int) -> List[Dict]:
    record = self.file_service.get_file_record(log_id) if self.file_service is not None else None
    rule_plan = {
        "large_transaction_hit_rules": record.large_transaction_hit_rules if record else [],
        "phase1_hit_rules": record.phase1_hit_rules if record else [],
    }

    seeded_statements = build_seed_statements_for_rule_plan(
        group_id=group_id,
        log_id=log_id,
        rule_plan=rule_plan,
        primary_enterprise_name=primary_enterprise_name,
        primary_account_no=primary_account_no,
        staff_id_card=record.staff_id_card if record else None,
        family_id_cards=record.family_id_cards if record else None,
    )
    total_count = max(count, len(seeded_statements))
    statements = list(seeded_statements)
    for _ in range(total_count - len(seeded_statements)):
        statements.append(
            self._generate_random_statement(
                group_id,
                log_id,
                primary_enterprise_name,
                primary_account_no,
                allowed_identity_cards,
                rng,
            )
        )
    return self._assign_statement_ids(statements, group_id, log_id)

要求：

命中样本生成完后仍统一走 _assign_statement_ids()
继续保留 200 条固定总数和缓存语义
未命中的规则绝不回退为默认全量样本
Step 4: Run test to verify it passes

Run:

cd lsfx-mock-server
pytest tests/test_statement_service.py -k "follow_rule_hit_plan or fixed_total_count_200 or cached_result" -v
pytest tests/integration/test_full_workflow.py -k "same_rule_subset or share_same_primary_binding" -v

Expected:

PASS
同一 logId 的规则命中子集稳定，且主绑定信息没有回归
Step 5: Commit

git add lsfx-mock-server/services/statement_service.py lsfx-mock-server/tests/test_statement_service.py lsfx-mock-server/tests/integration/test_full_workflow.py
git commit -m "接通Mock随机命中流水生成链路"

Task 4: 为 LARGE_PURCHASE_TRANSACTION 补最小数据库基线脚本

Files:

Create: sql/migration/2026-03-20-lsfx-mock-random-hit-rule-purchase-baseline.sql
Reference: sql/ccdi_purchase_transaction.sql
Reference: ccdi-project/src/main/resources/mapper/ccdi/project/CcdiBankTagAnalysisMapper.xml
Step 1: Write the failing verification query

先准备最小验证脚本思路，确认当前库里还没有本次联调专用的采购事项基线数据：

import pymysql

cursor.execute(
    "SELECT COUNT(1) AS cnt FROM ccdi_purchase_transaction WHERE purchase_id = 'LSFXMOCKPUR001'"
)
assert cursor.fetchone()["cnt"] == 0

Step 2: Run verification to confirm baseline data is absent

Run:

python3 - <<'PY'
import pymysql
from pathlib import Path
import re

text = Path('ruoyi-admin/src/main/resources/application-dev.yml').read_text(encoding='utf-8')
match = re.search(r"url:\s*jdbc:mysql://(?P<host>[^:/?#]+):(?P<port>\d+)/(?P<db>[^?\n]+).*?\n\s*username:\s*(?P<user>[^\n]+)\n\s*password:\s*(?P<pwd>[^\n]+)", text, re.S)
conn = pymysql.connect(
    host=match.group('host'),
    port=int(match.group('port')),
    user=match.group('user').strip(),
    password=match.group('pwd').strip(),
    database=match.group('db').strip(),
    charset='utf8mb4',
    cursorclass=pymysql.cursors.DictCursor,
)
with conn, conn.cursor() as cursor:
    cursor.execute("SELECT COUNT(1) AS cnt FROM ccdi_purchase_transaction WHERE purchase_id = 'LSFXMOCKPUR001'")
    print(cursor.fetchone()['cnt'])
PY

Expected:

输出 0
说明 LARGE_PURCHASE_TRANSACTION 所需联调基线数据尚未落库
Step 3: Write minimal implementation

创建 sql/migration/2026-03-20-lsfx-mock-random-hit-rule-purchase-baseline.sql，只插入最小必要采购记录，不扩展多余字段。脚本建议采用“先删后插”的幂等方式：

DELETE FROM ccdi_purchase_transaction WHERE purchase_id = 'LSFXMOCKPUR001';

INSERT INTO ccdi_purchase_transaction (
  purchase_id, purchase_category, project_name, subject_name, subject_desc,
  purchase_qty, budget_amount, bid_amount, actual_amount, contract_amount, settlement_amount,
  purchase_method, supplier_name, contact_person, contact_phone, supplier_uscc, supplier_bank_account,
  apply_date, plan_approve_date, announce_date, bid_open_date, contract_sign_date,
  expected_delivery_date, actual_delivery_date, acceptance_date, settlement_date,
  applicant_id, applicant_name, apply_department,
  purchase_leader_id, purchase_leader_name, purchase_department,
  created_by, updated_by
)
SELECT
  'LSFXMOCKPUR001', '设备采购', 'LSFX Mock 联调',
  '反洗钱终端设备采购', '用于命中 LARGE_PURCHASE_TRANSACTION 真实规则',
  1, 188000.00, 186000.00, 186000.00, 186000.00, 186000.00,
  '竞争性谈判', '兰溪市联调供应链有限公司', '联调联系人', '13800000000', '91330781MOCKPUR001', '6222000000001234',
  CURRENT_DATE, CURRENT_DATE, CURRENT_DATE, CURRENT_DATE, CURRENT_DATE,
  CURRENT_DATE, CURRENT_DATE, CURRENT_DATE, CURRENT_DATE,
  CAST(s.staff_id AS CHAR), s.name, '纪检初核部',
  NULL, NULL, NULL,
  'admin', 'admin'
FROM ccdi_base_staff s
WHERE COALESCE(TRIM(CAST(s.staff_id AS CHAR)), '') <> ''
  AND COALESCE(TRIM(s.name), '') <> ''
LIMIT 1;

要求：

只补一条采购记录
actual_amount 必须大于 100000
不手写 mysql -e 直接执行中文 SQL，后续执行必须使用 bin/mysql_utf8_exec.sh
Step 4: Run script and verify it works

Run:

bin/mysql_utf8_exec.sh sql/migration/2026-03-20-lsfx-mock-random-hit-rule-purchase-baseline.sql
python3 - <<'PY'
import pymysql
from pathlib import Path
import re

text = Path('ruoyi-admin/src/main/resources/application-dev.yml').read_text(encoding='utf-8')
match = re.search(r"url:\s*jdbc:mysql://(?P<host>[^:/?#]+):(?P<port>\d+)/(?P<db>[^?\n]+).*?\n\s*username:\s*(?P<user>[^\n]+)\n\s*password:\s*(?P<pwd>[^\n]+)", text, re.S)
conn = pymysql.connect(
    host=match.group('host'),
    port=int(match.group('port')),
    user=match.group('user').strip(),
    password=match.group('pwd').strip(),
    database=match.group('db').strip(),
    charset='utf8mb4',
    cursorclass=pymysql.cursors.DictCursor,
)
with conn, conn.cursor() as cursor:
    cursor.execute("""
        SELECT purchase_id, actual_amount, supplier_name
        FROM ccdi_purchase_transaction
        WHERE purchase_id = 'LSFXMOCKPUR001'
    """)
    print(cursor.fetchone())
PY

Expected:

SQL 脚本执行成功
查询结果返回 LSFXMOCKPUR001
actual_amount 大于 100000
Step 5: Commit

git add sql/migration/2026-03-20-lsfx-mock-random-hit-rule-purchase-baseline.sql
git commit -m "补齐Mock采购规则数据库基线"

Task 5: 做后端回归并沉淀实施与验证文档

Files:

Create: docs/reports/implementation/2026-03-20-lsfx-mock-random-hit-rule-backend-record.md
Create: docs/tests/records/2026-03-20-lsfx-mock-random-hit-rule-backend-verification.md
Reference: docs/design/2026-03-20-lsfx-mock-random-hit-rule-design.md
Reference: docs/plans/backend/2026-03-20-lsfx-mock-random-hit-rule-backend-implementation.md
Step 1: Run the full targeted regression

Run:

cd lsfx-mock-server
pytest tests/test_file_service.py tests/test_statement_service.py tests/test_api.py tests/integration/test_full_workflow.py -v

Expected:

PASS
随机命中计划、样本装配、缓存稳定与端到端链路均通过
Step 2: Verify LARGE_PURCHASE_TRANSACTION baseline separately

Run:

python3 - <<'PY'
import pymysql
from pathlib import Path
import re

text = Path('ruoyi-admin/src/main/resources/application-dev.yml').read_text(encoding='utf-8')
match = re.search(r"url:\s*jdbc:mysql://(?P<host>[^:/?#]+):(?P<port>\d+)/(?P<db>[^?\n]+).*?\n\s*username:\s*(?P<user>[^\n]+)\n\s*password:\s*(?P<pwd>[^\n]+)", text, re.S)
conn = pymysql.connect(
    host=match.group('host'),
    port=int(match.group('port')),
    user=match.group('user').strip(),
    password=match.group('pwd').strip(),
    database=match.group('db').strip(),
    charset='utf8mb4',
    cursorclass=pymysql.cursors.DictCursor,
)
with conn, conn.cursor() as cursor:
    cursor.execute("""
        SELECT COUNT(1) AS cnt
        FROM ccdi_purchase_transaction
        WHERE purchase_id = 'LSFXMOCKPUR001'
          AND actual_amount > 100000
    """)
    print(cursor.fetchone()['cnt'])
PY

Expected:

输出 1
说明采购基线数据已满足真实规则命中门槛
Step 3: Write implementation and verification records

在 docs/reports/implementation/2026-03-20-lsfx-mock-random-hit-rule-backend-record.md 中记录：

规则命中计划如何生成
样本模块如何按规则子集装配
StatementService 如何读取计划并缓存
LARGE_PURCHASE_TRANSACTION 为什么单独补数据库基线

在 docs/tests/records/2026-03-20-lsfx-mock-random-hit-rule-backend-verification.md 中记录：

pytest 执行时间与结果
SQL 基线脚本执行结果
采购基线查询结果
是否发现回归
Step 4: Check Git change scope

Run:

git status --short

Expected:

只包含本次任务相关文件
.DS_Store 不纳入提交
Step 5: Commit

git add docs/reports/implementation/2026-03-20-lsfx-mock-random-hit-rule-backend-record.md docs/tests/records/2026-03-20-lsfx-mock-random-hit-rule-backend-verification.md
git commit -m "补充Mock随机命中后端实施记录"

23 KiB Raw Blame History Unescape Escape

LSFX Mock Random Hit Rule Backend Implementation Plan

File Structure

Task 1: 在 FileService 中持久化稳定随机的规则命中计划

Task 2: 将 statement_rule_samples.py 拆成按规则代码装配的样本生成器

Task 3: 让 StatementService 按命中计划生成样本并保持缓存稳定

Task 4: 为 LARGE_PURCHASE_TRANSACTION 补最小数据库基线脚本

Task 5: 做后端回归并沉淀实施与验证文档

23 KiB

Raw Blame History