kangshifu-board/legacy/tools/seed_opportunities.py

"""seed_opportunities.py —— 造机会卡示例数据（KfOpportunity）+ 上传元数据 + 一条"卡住"的现场数据。

为什么单独一个类：机会卡有独立生命周期（评分、状态、决策记录、优先级），跟"成果"不是一回事。
理由见 README「为什么新建 KfOpportunity」。

可重复执行（幂等）：
  - 每张卡带 seedKey；已存在同 seedKey 的卡：只补缺失的上传元数据 / 关联成果 / 历史评分，不覆盖人工改过的 status/priority/decision；
  - 评分历史 [seed:xxx] 按 seedKey 去重，不会越跑越多；
  - 现场演示用的卡住数据同样按 seedKey 幂等。

  PARSE_MASTER_KEY=... python3 tools/seed_opportunities.py             # 写入
  PARSE_MASTER_KEY=... python3 tools/seed_opportunities.py --dry-run   # 只看要写什么
  PARSE_MASTER_KEY=... python3 tools/seed_opportunities.py --reset-scores  # 把示例评分恢复到初始值

凭据只从环境变量读，不落盘。
"""
import json
import os
import sys
import urllib.error
import urllib.request
from datetime import datetime, timedelta, timezone

BASE = "https://kangshifu.fmode.cn/parse"
APP_ID = "FKieRbIoGt_main"
MK = os.environ.get("PARSE_MASTER_KEY", "")
DRY = "--dry-run" in sys.argv
RESET_SCORES = "--reset-scores" in sys.argv

H = {"X-Parse-Application-Id": APP_ID, "Content-Type": "application/json"}
if MK:
    H["X-Parse-Master-Key"] = MK

NOW = datetime.now(timezone.utc)


def iso(dt):
    return dt.strftime("%Y-%m-%dT%H:%M:%S.") + "%03dZ" % (dt.microsecond // 1000)


def hours_ago(h):
    return iso(NOW - timedelta(hours=h))


def call(method, path, body=None):
    data = json.dumps(body, ensure_ascii=False).encode() if body is not None else None
    req = urllib.request.Request(BASE + path, data=data, headers=H, method=method)
    try:
        with urllib.request.urlopen(req, timeout=30) as r:
            return json.loads(r.read().decode() or "{}")
    except urllib.error.HTTPError as e:
        return {"_error": e.code, "_body": e.read().decode()[:400]}


def fetch(cls):
    return call("GET", f"/classes/{cls}?limit=1000&order=createdAt").get("results", [])


# ---------------------------------------------------------------- 上传元数据
# 所有"被上传的东西"都带这五个字段；老数据补齐 uploadedByKey 的值由这里统一推导。
UPLOAD_METHODS = {"session": "会话上传", "skill": "Skill 上传", "manual": "手工录入"}


def upload_meta(person, key, agent_name, agent_id, method, at, session=None):
    m = {
        "uploadedBy": person,
        "uploadedByKey": key,
        "agentName": agent_name,
        "agentId": agent_id,
        "uploadMethod": method,
        "uploadedAt": at,
    }
    if session:
        m["sourceSession"] = session
    return m


MAIN = ("康师傅产品创新助手", "agent-main-01")     # 主分身：学员在会话里把想法交给它
RESEARCH = ("品类调研 Agent", "agent-research-02")  # 调研 Agent：自动抓 VOC / 竞品
CHECK = ("机会评估 Agent", "agent-check-03")        # 评估 Agent：按评分模型打分
DESIGN = ("概念设计 Agent", "agent-design-04")      # 设计 Agent：把机会变成方案

# ---------------------------------------------------------------- 机会卡
# 每张卡都挂到库里真实存在的成果（KfArtifact / KfProject），保证"点开看报告/证据"有东西。
# scores：[评分人, 分数, 评语, 几小时前, seedKey]
CARDS = [
    {
        "seedKey": "seed-opp-1",
        "title": "夜宵场景低卡高蛋白拌面",
        "oneLiner": "加班到十点想吃饱又怕胖：低卡高蛋白 + 1 分钟即食，5 个夜宵场景里 3 个需求未被满足",
        "stage": "S3 机会识别",
        "category": "拌面",
        "priority": "S",
        "status": "candidate",
        "evidenceLevel": "high",
        "quoteCount": 12,
        "citeCount": 8,
        "upload": upload_meta("张伟", "sess-01", *MAIN, "session", hours_ago(6), session="sess-01"),
        "linkedProject": "QFvrQsUPyD",
        "linkedArtifacts": ["KHEScCO5dT", "5icUD1vUVN", "BjBza4kr2U"],
        "decision": {
            "type": "concept", "label": "进入概念验证",
            "reason": "S3 阶段 5 位成员独立提到夜宵场景的「饱腹 + 低卡」矛盾；已有 12 条带原文的 VOC 支撑。建议 S5 做一次概念测试。",
            "by": "张伟", "at": hours_ago(4),
        },
        "scores": [
            ("李娜", 86, "场景真实，VOC 密度高，可以做", 9),
            ("吴敏", 90, "和现有夜宵 SKU 不冲突，值得立项", 7),
            ("周涛", 84, "高蛋白卖点需要工艺支持，先验证", 5),
            ("评价 Agent", 88, "人群规模 × 需求强度 × 可行性 综合 88", 3),
        ],
    },
    {
        "seedKey": "seed-opp-2",
        "title": "办公室 3 分钟速食拌面",
        "oneLiner": "午休只有 10 分钟、不想吃汤汤水水：免煮、不冒汤、可带去办公室的拌面",
        "stage": "S4 验证评估",
        "category": "拌面",
        "priority": "A",
        "status": "approved",
        "evidenceLevel": "high",
        "quoteCount": 9,
        "citeCount": 5,
        "upload": upload_meta("李娜", "sess-02", *MAIN, "session", hours_ago(30), session="sess-02"),
        "linkedProject": "PzNTNt7ttt",
        "linkedArtifacts": ["BjBza4kr2U", "yGwGJ9aPu6"],
        "decision": {
            "type": "approve", "label": "立项",
            "reason": "S4 验证完成：三家竞品均无免汤方案，目标人群 2.7 亿中约 18% 处于该场景；评分模型 82 分，过线。",
            "by": "刘洋", "at": hours_ago(20),
        },
        "scores": [
            ("刘洋", 84, "验证数据齐，建议立项", 26),
            ("王芳", 80, "价格带要卡住 8 元", 22),
            ("评价 Agent", 82, "需求 85 / 竞争 78 / 可行性 83", 20),
        ],
    },
    {
        "seedKey": "seed-opp-3",
        "title": "地域风味麻酱拌面",
        "oneLiner": "现有拌面酱包集中在麻酱 / 香辣两类；西北油泼、西南麻辣是搜索热但 SKU 空档",
        "stage": "S4 验证评估",
        "category": "拌面",
        "priority": "A",
        "status": "candidate",
        "evidenceLevel": "medium",
        "quoteCount": 6,
        "citeCount": 3,
        "upload": upload_meta("刘洋", "sess-04", *RESEARCH, "skill", hours_ago(52)),
        "linkedProject": "",
        "linkedArtifacts": ["QVrfdasAkg", "C6AFWrqyTQ"],
        "decision": {
            "type": "hold", "label": "暂缓（补原文）",
            "reason": "方向认可，但支撑它的 6 条记录里只有 3 条带原文，销量数据待调研 Agent 回传。补齐前不排产。",
            "by": "赵磊", "at": hours_ago(40),
        },
        "scores": [
            ("赵磊", 72, "方向对，证据不够", 45),
            ("陈静", 78, "和夜宵低卡可以合并做双口味", 38),
            ("评价 Agent", 75, "证据强度拖后腿", 36),
        ],
    },
    {
        "seedKey": "seed-opp-4",
        "title": "露营便携免煮拌面",
        "oneLiner": "露营带的都是泡面，想吃拌面又得煮：冷水泡 8 分钟、自带拌酱的免煮方案",
        "stage": "S5 场景与卖点设计",
        "category": "杯面",
        "priority": "S",
        "status": "candidate",
        "evidenceLevel": "medium",
        "quoteCount": 7,
        "citeCount": 4,
        "upload": upload_meta("陈静", "sess-05", *DESIGN, "skill", hours_ago(2)),
        "linkedProject": "",
        "linkedArtifacts": ["mdJAkc5u6X", "w5DDt2Qek4"],
        "decision": {
            "type": "concept", "label": "进入概念验证",
            "reason": "露营搜索量 2026 春夏环比 +41%；已有方案具备卖点抓手（「不烧水也能拌」）。等 S4 评分补齐。",
            "by": "陈静", "at": hours_ago(1.5),
        },
        "scores": [
            ("孙悦", 79, "场景有意思，量可能不大", 8),
            ("郑凯", 81, "卖点很清晰，可以做包装测试", 2),
        ],
    },
    {
        "seedKey": "seed-opp-5",
        "title": "健身人群高蛋白面",
        "oneLiner": "健身人群把速食面当「欺骗餐」，要的是每份 20g 以上蛋白、低脂、能算进宏量",
        "stage": "S2 用户需求与痛点",
        "category": "汤面",
        "priority": "B",
        "status": "rejected",
        "evidenceLevel": "low",
        "quoteCount": 2,
        "citeCount": 1,
        "upload": upload_meta("孙悦", "sess-07", *MAIN, "session", hours_ago(76), session="sess-07"),
        "linkedProject": "",
        "linkedArtifacts": ["0TufH1peEG", "rSqN03GPux", "WrWlJ1lSHC"],
        "decision": {
            "type": "reject", "label": "否决",
            "reason": "三条支撑记录里只有 2 条原文，且蛋白成本会把单价推到 15 元以上，和现有价格带冲突。留档，不做。",
            "by": "吴敏", "at": hours_ago(60),
        },
        "scores": [
            ("吴敏", 58, "成本算不过来", 70),
            ("周涛", 62, "需求真实但规模小", 66),
            ("评价 Agent", 55, "人群规模 ★★，可行性 ★★", 62),
        ],
    },
    {
        "seedKey": "seed-opp-6",
        "title": "银发低钠汤面",
        "oneLiner": "50 岁以上人群控盐诉求明确，现有方便面钠含量普遍 2000mg 以上，无低钠 SKU",
        "stage": "S1 品类市场分析",
        "category": "汤面",
        "priority": "B",
        "status": "candidate",
        "evidenceLevel": "medium",
        "quoteCount": 4,
        "citeCount": 2,
        "upload": upload_meta("赵磊", "sess-06", *MAIN, "manual", hours_ago(50)),
        "linkedProject": "",
        "linkedArtifacts": ["Lor4psRme0", "rvyuQ3985Q"],
        "decision": None,
        "scores": [
            ("赵磊", 70, "合规是卖点，工艺要改", 48),
        ],
    },
    {
        "seedKey": "seed-opp-7",
        "title": "夏日清爽凉拌面",
        "oneLiner": "夏季 6–8 月拌面搜索 +35%，但现有产品多为「热拌」，冷拌 / 冰镇吃法没有对应 SKU",
        "stage": "S3 机会识别",
        "category": "拌面",
        "priority": "A",
        "status": "candidate",
        "evidenceLevel": "medium",
        "quoteCount": 5,
        "citeCount": 3,
        "upload": upload_meta("周涛", "sess-08", *RESEARCH, "skill", hours_ago(0.4)),
        "linkedProject": "",
        "linkedArtifacts": ["9Mj4LV3a3P", "czLP44qeOl", "W42NYTiJFm"],
        "decision": {
            "type": "hold", "label": "指派评估",
            "reason": "季节性明确，需要 S4 用评分模型验证「冷藏口感」这一工艺风险。",
            "by": "吴敏", "at": hours_ago(0.2),
        },
        "scores": [
            ("周涛", 76, "季节性太强，看全年怎么讲", 0.3),
        ],
    },
    {
        "seedKey": "seed-opp-8",
        "title": "学生党实惠大份拌面",
        "oneLiner": "宿舍场景吃一碗不饱：同价位加量 30%，以「三包面饼」作为卖点",
        "stage": "S3 机会识别",
        "category": "炒面",
        "priority": "S",
        "status": "candidate",
        "evidenceLevel": "medium",
        "quoteCount": 5,
        "citeCount": 2,
        "upload": upload_meta("王芳", "sess-03", *MAIN, "session", hours_ago(28), session="sess-03"),
        "linkedProject": "gC9Ra0bnXf",
        "linkedArtifacts": ["OG7w0dxbja", "HmZG9wFEq7"],
        "decision": {
            "type": "concept", "label": "进入概念验证",
            "reason": "「吃不饱」在差评里频次高于「难吃」；加量 + 不涨价可行性待工艺评估。",
            "by": "王芳", "at": hours_ago(24),
        },
        "scores": [
            ("王芳", 91, "最接近真实痛点的一张卡", 26),
            ("陈静", 88, "可以和学生套餐组合", 24),
            ("评价 Agent", 89, "需求 92 / 竞争 80 / 可行性 88", 22),
        ],
    },
    # 早前用 skill 上传、后被判重复的一张卡：保留它，池子里才有"否决"的真实样本
    {
        "seedKey": "seed-opp-9",
        "title": "夜宵低卡拌面（与 seed-opp-1 重复）",
        "oneLiner": "陈静在 S3 独立上传的机会卡，内容与张伟的夜宵低卡高蛋白重合",
        "stage": "S3 机会识别",
        "category": "拌面",
        "priority": "B",
        "status": "rejected",
        "evidenceLevel": "low",
        "quoteCount": 0,
        "citeCount": 0,
        "upload": upload_meta("陈静", "sess-05", *MAIN, "skill", hours_ago(68)),
        "linkedProject": "",
        "linkedArtifacts": ["C6AFWrqyTQ"],
        "decision": {
            "type": "merge", "label": "否决（并入 seed-opp-1）",
            "reason": "与张伟的夜宵低卡高蛋白拌面重复，评分并入原卡，本卡留档。",
            "by": "吴敏", "at": hours_ago(64),
        },
        "scores": [],
    },
]


# ---------------------------------------------------------------- 现场演示：一个卡住的人
# 现场看板要能列出"谁卡住了"：一条有时长但在 S4 停了 30 小时、没有任何成果的会话。
STUCK_SESSION = {
    "seedKey": "seed-stuck-session",
    "className": "KfSession",
    "fields": {
        "studentName": "何雨",
        "groupId": "G3",
        "stage": "S4验证评估",
        "step": "1/3",
        "msgCount": 13,
        "lastActiveAt": {"__type": "Date", "iso": hours_ago(30)},
    },
}

CARD_FIELDS_SCHEMA = {
    "title": {"type": "String"},
    "oneLiner": {"type": "String"},
    "stage": {"type": "String"},
    "category": {"type": "String"},
    "priority": {"type": "String"},
    "status": {"type": "String"},
    "evidenceLevel": {"type": "String"},
    "quoteCount": {"type": "Number"},
    "citeCount": {"type": "Number"},
    "uploadedBy": {"type": "String"},
    "uploadedByKey": {"type": "String"},
    "agentName": {"type": "String"},
    "agentId": {"type": "String"},
    "uploadMethod": {"type": "String"},
    "uploadedAt": {"type": "Date"},
    "sourceSession": {"type": "String"},
    "linkedProject": {"type": "String"},
    "linkedArtifacts": {"type": "Array"},
    "decision": {"type": "Object"},
    "scores": {"type": "Array"},
    "decisionLog": {"type": "Array"},
    "seedKey": {"type": "String"},
    "searchText": {"type": "String"},
}

UPLOAD_FIELDS_SCHEMA = {
    "uploadedByKey": {"type": "String"},
    "agentName": {"type": "String"},
    "agentId": {"type": "String"},
    "uploadMethod": {"type": "String"},
    "uploadedAt": {"type": "Date"},
}


def seed_scores(card):
    out = []
    for person, score, comment, hours in card["scores"]:
        out.append({
            "user": person,
            "score": score,
            "comment": comment,
            "at": {"__type": "Date", "iso": hours_ago(hours)},
            "seed": card["seedKey"],
        })
    return out


def search_text(d):
    return " ".join(str(d.get(k) or "") for k in ("title", "oneLiner", "category", "stage", "priority", "uploadedBy", "agentName")).lower()


def main():
    if not MK:
        sys.exit("PARSE_MASTER_KEY 未设置")
    if "results" not in call("GET", "/classes/KfProject?limit=1"):
        sys.exit("masterKey 验证失败")

    # 1) schema
    #    KfOpportunity 首次运行时按 CARD_FIELDS_SCHEMA 一次建全（Parse 拒绝重复改已有字段，
    #    所以已存在就跳过）；KfArtifact / KfProject 只补上传元数据字段。
    has_opp = bool(call("GET", "/schemas/KfOpportunity").get("fields"))
    plan = [] if has_opp else [("KfOpportunity", CARD_FIELDS_SCHEMA)]
    plan += [("KfArtifact", UPLOAD_FIELDS_SCHEMA), ("KfProject", UPLOAD_FIELDS_SCHEMA)]
    for cls, fields in plan:
        if DRY:
            print("[dry] schema", cls, list(fields))
            continue
        r = call("PUT", f"/schemas/{cls}", {"className": cls, "fields": fields})
        print("schema", cls, "ok" if "fields" in r else r)

    # 2) 机会卡：按 seedKey 幂等
    existing = {c.get("seedKey"): c for c in fetch("KfOpportunity") if c.get("seedKey")}
    artifacts = {a["objectId"]: a for a in fetch("KfArtifact")}
    for card in CARDS:
        body = {
            "title": card["title"], "oneLiner": card["oneLiner"], "stage": card["stage"],
            "category": card["category"], "priority": card["priority"], "status": card["status"],
            "evidenceLevel": card["evidenceLevel"], "quoteCount": card["quoteCount"],
            "citeCount": card["citeCount"], "seedKey": card["seedKey"],
            "linkedProject": card.get("linkedProject") or "",
            "linkedArtifacts": [i for i in card["linkedArtifacts"] if i in artifacts],
        }
        body.update(card["upload"])
        body["uploadedAt"] = {"__type": "Date", "iso": card["upload"]["uploadedAt"]}
        body["searchText"] = search_text({**card, **card["upload"]})
        dec = dict(card["decision"], at={"__type": "Date", "iso": card["decision"]["at"]}) if card["decision"] else None
        if dec:
            body["decision"] = dec
        body["decisionLog"] = [dec] if dec else []

        cur = existing.get(card["seedKey"])
        if not cur:
            body["scores"] = seed_scores(card)
            if DRY:
                print("[dry] create", card["title"], [s[0] for s in card["scores"]])
                continue
            r = call("POST", "/classes/KfOpportunity", body)
            print("create", card["title"], "ok" if "objectId" in r else r)
            continue

        # 已存在：只补缺失的字段；不覆盖人工改过的 status / priority / decision
        patch = {}
        for k in ("uploadedBy", "uploadedByKey", "agentName", "agentId", "uploadMethod", "uploadedAt",
                  "linkedProject", "linkedArtifacts", "evidenceLevel", "quoteCount", "citeCount", "oneLiner"):
            if not cur.get(k) and body.get(k):
                patch[k] = body[k]
        if not cur.get("decision") and dec:
            patch["decision"] = dec
            patch["decisionLog"] = [dec]
        live_scores = cur.get("scores") or []
        have = {(s.get("user"), s.get("seed")) for s in live_scores}
        if RESET_SCORES:
            live_scores = [s for s in live_scores if not s.get("seed")]
        missing = [s for s in seed_scores(card) if (s["user"], s["seed"]) not in have]
        if len(live_scores) != len(cur.get("scores") or []) or missing:
            patch["scores"] = live_scores + missing
        if patch:
            if DRY:
                print("[dry] patch", card["title"], list(patch))
                continue
            r = call("PUT", f"/classes/KfOpportunity/{cur['objectId']}", patch)
            print("patch", card["title"], list(patch), "ok" if "updatedAt" in r else r)
        else:
            print("keep ", card["title"], "(已存在且完整)")

    # 3) 老成果 / 项目补齐上传元数据（只补缺的）
    owner_method = {}
    for card in CARDS:
        owner_method[card["upload"]["uploadedBy"]] = (card["upload"], card["upload"]["uploadedByKey"])
    for cls in ("KfArtifact", "KfProject"):
        rows = fetch(cls)
        for row in rows:
            patch = {}
            if not row.get("uploadedByKey"):
                patch["uploadedByKey"] = row.get("ownerSession") or row.get("ownerName") or ""
            if not row.get("agentName"):
                patch["agentName"] = MAIN[0]
                patch["agentId"] = MAIN[1]
            if not row.get("uploadMethod"):
                patch["uploadMethod"] = "skill"  # 现有 26 条成果全部是 jarvis skill 写入
            if not row.get("uploadedAt") and row.get("createdAt"):
                patch["uploadedAt"] = {"__type": "Date", "iso": row.get("createdAt")}
            if not patch:
                continue
            if DRY:
                continue
            r = call("PUT", f"/classes/{cls}/{row['objectId']}", patch)
            if "updatedAt" not in r:
                print("ERR upload-meta", cls, row["objectId"], r)
        print(f"upload-meta {cls}: {len(rows)} 行已检查")

    # 4) 现场演示：卡住的人
    s = STUCK_SESSION
    have = [x for x in fetch(s["className"]) if x.get("seedKey") == s["seedKey"]]
    if have:
        print("keep  stuck-session", have[0]["objectId"])
    elif DRY:
        print("[dry] create stuck-session", s["fields"]["studentName"])
    else:
        r = call("POST", f"/classes/{s['className']}", {**s["fields"], "seedKey": s["seedKey"]})
        ok = "objectId" in r
        print("create stuck-session", "ok" if ok else r)
        if not ok:
            # KfSession 没有 seedKey 字段时，退回用姓名查找
            found = [x for x in fetch("KfSession") if x.get("studentName") == s["fields"]["studentName"]]
            print("  fallback find:", len(found))


if __name__ == "__main__":
    main()