#!/usr/bin/env python3
"""从实测产物构建前端展示种子：NeuroPulse (project 5201)。"""

from __future__ import annotations

import csv
import json
import os
import re
import subprocess
from datetime import datetime
from pathlib import Path

ROOT = Path(__file__).resolve().parents[1]
OUT = ROOT / "web" / "src" / "data" / "seeds" / "neuropulseSeed.json"

PROJECT_ID = 5201
PROJECT_NAME = "神经外科双频超声吸引系统 (CUSA-NeuroPulse)"
DEVICE_TYPE = "超声软组织手术吸引设备"

DOC_MAP = {
    "intake": {
        "work_dir": "data/dev/workspaces/bizdev/intake_1788920542147409000",
        "title": "用户初始需求收集表",
        "stage": "V0.1",
        "state": "approved",
        "summary": "CUSA-NeuroPulse 立项需求摸底与临床预期用途基线（真实 MiniMax 产出）。",
    },
    "research": {
        "work_dir": "data/dev/workspaces/bizdev/research_1788921024562992000",
        "title": "调研报告（竞品、市场与法规初筛）",
        "stage": "V0.5",
        "state": "approved",
        "summary": "竞品/市场/法规初筛调研报告（真实 MiniMax 产出）。",
    },
    "feasibility_dev": {
        "work_dir": "data/dev/workspaces/bizdev/feasibility_dev_1788922272891355000",
        "title": "设计开发可行性评估报告",
        "stage": "V0.5",
        "state": "approved",
        "summary": "设计开发可行性评估（真实 MiniMax 产出）。",
    },
    "feasibility_reg": {
        "work_dir": "data/dev/workspaces/bizdev/feasibility_reg_1788922655324377000",
        "title": "注册路径可行性评估报告",
        "stage": "V0.5",
        "state": "approved",
        "summary": "NMPA 三类创新通道注册路径可行性（已终审批准）。",
    },
    "user_requirements": {
        "work_dir": "data/dev/workspaces/bizdev/user_requirements_1788923266008791000",
        "title": "用户需求表",
        "stage": "V0.5",
        "state": "reviewing",
        "summary": "用户需求表 UR（真实 MiniMax 产出，审核中）。",
    },
    "prelim_design": {
        "work_dir": "data/dev/workspaces/bizdev/prelim_design_1788923504095801000",
        "title": "系统初步设计方案",
        "stage": "V0.5",
        "state": "approving",
        "summary": "系统初步设计方案（真实 MiniMax 产出，待批准）。",
    },
    "cdo_quote": {
        "work_dir": "data/dev/workspaces/bizdev/cdo_quote_1788924027615053000",
        "title": "CDO商务报价单",
        "stage": "V1.0",
        "state": "drafting",
        "summary": "CDO 商务报价与工时拆解（真实 MiniMax 产出）。",
    },
}


def sid(uid: int) -> str:
    return f"sess-neuropulse-u{uid}"


def read_md(rel: str) -> str:
    p = ROOT / rel / "output.md"
    if not p.exists():
        return f"# 缺失产物\n\n路径不存在: `{rel}/output.md`"
    return p.read_text(encoding="utf-8")


def load_db_messages():
    cmd = [
        "docker",
        "exec",
        "shuimu-context-pg-dev",
        "psql",
        "-U",
        "app",
        "-d",
        "agent_system_dev",
        "-c",
        r"\copy (SELECT session_id, role, content, COALESCE(intent,''), created_at FROM chat_messages WHERE session_id LIKE 'sess_full_p5201%' ORDER BY id) TO STDOUT WITH CSV",
    ]
    raw = subprocess.check_output(cmd, text=True)
    rows = list(csv.reader(raw.splitlines()))
    # map backend session -> UI session
    mapped = []
    for sess, role, content, intent, created in rows:
        uid = 1
        if sess.endswith("_u2"):
            uid = 2
        elif sess.endswith("_u3"):
            uid = 3
        mapped.append(
            {
                "session_id": sid(uid),
                "role": role,
                "content": content,
                "intent": intent,
                "created_at": created,
            }
        )
    return mapped


def load_early_dialogues_from_results():
    path = Path("/tmp/full_lifecycle_p5201_results.json")
    if not path.exists():
        return []
    data = json.loads(path.read_text(encoding="utf-8"))
    out = []
    for d in data.get("dialogues") or []:
        uid = d.get("user_id") or 1
        name = d.get("user") or "张工"
        # skip empty timeout
        user_msg = d.get("input") or ""
        reply = d.get("reply") or ""
        intent = d.get("intent") or ""
        if not user_msg:
            continue
        out.append({"uid": uid, "name": name, "role": "user", "content": user_msg, "intent": ""})
        if reply:
            out.append(
                {
                    "uid": uid,
                    "name": "水木东方协同参谋",
                    "role": "assistant",
                    "content": reply,
                    "intent": intent,
                }
            )
        elif "生成《用户初始需求收集表》" in user_msg:
            out.append(
                {
                    "uid": uid,
                    "name": "水木东方协同参谋",
                    "role": "assistant",
                    "content": (
                        "已为您启动《用户初始需求收集表》编制任务。智能体小队已完成立项画像整理，"
                        "正式稿已写入交付物大盘（doc-neuropulse-intake）。您可在右侧工作流文档中打开查阅，"
                        "并按编制人流程认领后送审。"
                    ),
                    "intent": "TRIGGER_TASK",
                }
            )
    # reminders (full from DB preferred; keep order markers)
    for key in (
        "author_after_reject",
        "reviewer_pending",
        "approver_pending",
        "reviewer_after_resubmit",
        "approver_after_pass",
    ):
        item = (data.get("reminders") or {}).get(key)
        if not isinstance(item, dict) or not item.get("input"):
            continue
        # DB has fuller replies for these; skip adding truncated duplicates here
    return out


def enrich_assistant_from_reply_files(messages):
    """用 dialogue/reply.md 全文替换截断的同类答复。"""
    replies = []
    for p in (ROOT / "data/dev/workspaces/dialogue").glob("*/reply.md"):
        text = p.read_text(encoding="utf-8").strip()
        if text:
            replies.append(text)

    def better(old: str) -> str:
        if not old:
            return old
        head = old[:80].replace("\n", " ")
        for r in replies:
            if r.startswith(old[:40]) or old[:40] in r[:120] or head[:30] in r[:200]:
                if len(r) > len(old):
                    return r
        return old

    for m in messages:
        if m.get("role") == "assistant":
            m["content"] = better(m["content"])
    return messages


def build_messages():
    early = load_early_dialogues_from_results()
    db = load_db_messages()

    # 先装 early（张工主线），再装 DB（避免重复时以 DB 全文覆盖同类）
    by_session: dict[str, list] = {sid(1): [], sid(2): [], sid(3): []}
    idx = 0

    def push(uid, role, content, intent="", author=None, ts=None):
        nonlocal idx
        idx += 1
        session = sid(uid)
        hhmm = ts or f"10:{10 + idx:02d}"
        if "T" in str(hhmm):
            # 2026-09-09 11:28:52.xxx+08
            try:
                hhmm = str(hhmm)[11:16]
            except Exception:
                hhmm = f"11:{idx:02d}"
        msg = {
            "id": f"m-np-{idx}",
            "sessionId": session,
            "role": role,
            "content": content,
            "timestamp": hhmm,
        }
        if role == "user":
            msg["authorName"] = {1: "张工", 2: "李主管", 3: "王总监"}.get(uid, "用户")
        else:
            msg["authorName"] = author or "水木东方协同参谋"
            if intent:
                msg["intentTag"] = intent
        by_session[session].append(msg)

    # system banner for each
    for uid, uname in [(1, "张工"), (2, "李主管"), (3, "王总监")]:
        push(
            uid,
            "system",
            f"水木东方协同参谋小队已就位。当前项目【{PROJECT_NAME}】进入【V0.5 方案成型与法规对齐】。"
            f"您正以「{uname}」身份与智能体私聊，其他成员对话不可见。",
            ts="09:50",
        )

    for e in early:
        push(e["uid"], e["role"], e["content"], e.get("intent") or "", e.get("name"), ts=None)

    # DB messages (reminders / later turns) — append if not duplicate of last user text
    existing_user_texts = set()
    for sess_msgs in by_session.values():
        for m in sess_msgs:
            if m["role"] == "user":
                existing_user_texts.add(m["content"][:40])

    for row in db:
        uid = 1
        if row["session_id"].endswith("u2"):
            uid = 2
        elif row["session_id"].endswith("u3"):
            uid = 3
        if row["role"] == "user" and row["content"][:40] in existing_user_texts:
            continue
        push(uid, row["role"], row["content"], row.get("intent") or "", ts=row.get("created_at"))

    # flatten + enrich
    all_msgs = []
    for s in (sid(1), sid(2), sid(3)):
        all_msgs.extend(by_session[s])
    all_msgs = enrich_assistant_from_reply_files(all_msgs)
    return all_msgs, {s: len(by_session[s]) for s in by_session}


def main():
    messages, counts = build_messages()
    documents = {}
    deliverables = []
    for key, meta in DOC_MAP.items():
        doc_id = f"doc-neuropulse-{key}"
        md = read_md(meta["work_dir"])
        documents[doc_id] = {
            "docId": doc_id,
            "projectId": PROJECT_ID,
            "taskKey": key,
            "title": meta["title"],
            "stage": meta["stage"],
            "category": "workflow",
            "creatorId": 1,
            "creatorName": "张工",
            "version": "V1.0" if meta["state"] == "approved" else "V0.9",
            "lifecycleState": meta["state"],
            "isTemplateOnly": False,
            "isClaimed": True,
            "author": "张工",
            "reviewer": "李主管",
            "approver": "王总监",
            "updatedAt": "2026-09-09 11:30",
            "summary": meta["summary"],
            "markdown": md,
        }
        deliverables.append(
            {
                "docId": doc_id,
                "projectId": PROJECT_ID,
                "taskKey": key,
                "title": meta["title"],
                "stage": meta["stage"],
                "category": "workflow",
                "creatorId": 1,
                "creatorName": "张工",
                "isFixedWorkflow": True,
                "isTemplateOnly": False,
                "lifecycleState": meta["state"],
                "isClaimed": True,
                "author": "张工",
                "reviewer": "李主管",
                "approver": "王总监",
                "updatedAt": "2026-09-09 11:30",
                "summary": meta["summary"],
            }
        )

    sessions = []
    for uid, uname in [(1, "张工"), (2, "李主管"), (3, "王总监")]:
        sessions.append(
            {
                "sessionId": sid(uid),
                "userId": uid,
                "projectId": PROJECT_ID,
                "title": PROJECT_NAME,
                "deviceType": DEVICE_TYPE,
                "stage": "V0.5",
                "stageLabel": "V0.5 方案成型与法规对齐",
                "messageCount": counts.get(sid(uid), 0),
                "updatedAt": "刚刚",
            }
        )

    seed = {
        "generatedAt": datetime.now().isoformat(timespec="seconds"),
        "projectId": PROJECT_ID,
        "projectName": PROJECT_NAME,
        "activeSessionId": sid(1),
        "sessions": sessions,
        "messages": messages,
        "deliverables": deliverables,
        "documents": documents,
        "permissions": {
            "1": {d["docId"]: "author" for d in deliverables},
            "2": {d["docId"]: "reviewer" for d in deliverables},
            "3": {d["docId"]: "approver" for d in deliverables},
        },
    }

    OUT.parent.mkdir(parents=True, exist_ok=True)
    OUT.write_text(json.dumps(seed, ensure_ascii=False), encoding="utf-8")
    print(f"Wrote {OUT}")
    print(f"sessions={len(sessions)} messages={len(messages)} docs={len(documents)}")
    for k, v in documents.items():
        print(f"  {k}: markdown={len(v['markdown'])} state={v['lifecycleState']}")


if __name__ == "__main__":
    main()