#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
services/code_review_pipeline_service.py
Post-Deploy AI Agent Code Review Pipeline

觸發時機：CD 健康檢查通過後，由 Gitea Action webhook 呼叫
Pipeline：
  Step 1  system          讀取變更檔案內容
  Step 2  Hermes          程式碼掃描（bugs / security / performance）
  Step 3  OpenClaw        架構品質評估（Gemini 2.5 Flash）
  Step 4  ElephantAlpha   決策協調（severity 判定 + auto-fix 裁量）
  Step 5  NemoTron        行動派遣（action_plans 寫入 + AiderHeal 觸發）

結果輸出：
  - ai_insights（type='code_review_result'）
  - action_plans（type='code_review_fix'）
  - Telegram 告警（啟動 / 完成 / 錯誤）
  - 前端 /code-review/ 即時 polling 狀態
"""

import json
import logging
import os
import re
import threading
from datetime import datetime
from typing import Any, Dict, List, Optional

from database.manager import get_session
from sqlalchemy import text

logger = logging.getLogger(__name__)

# ── Pipeline 全域狀態（供前端 polling）─────────────────────────────────────
_current_pipeline: Dict[str, Any] = {}
_pipeline_lock = threading.Lock()

GEMINI_API_KEY = os.getenv("GEMINI_API_KEY", "")
REVIEW_MODEL   = os.getenv("OPENCLAW_MODEL", "gemini-2.5-flash-preview-05-20")
INTERNAL_TOKEN = os.getenv("INTERNAL_WEBHOOK_TOKEN", "")


# ═══════════════════════════════════════════════════════════════════════════════
# Pipeline Class
# ═══════════════════════════════════════════════════════════════════════════════

class CodeReviewPipeline:
    """
    5-Step post-deploy code review pipeline.
    Call pipeline.run() inside a daemon thread.
    """

    def __init__(self, commit_sha: str, changed_files: List[str],
                 branch: str = "main", deploy_type: str = "sync"):
        self.commit_sha   = commit_sha
        self.branch       = branch
        self.deploy_type  = deploy_type
        self.started_at   = datetime.now()
        self.pipeline_id  = f"cr_{commit_sha[:8]}_{self.started_at.strftime('%Y%m%d_%H%M%S')}"

        # 只 review Python + YAML 檔（跳過靜態資源）
        self.changed_files = [
            f for f in changed_files
            if f.endswith(('.py', '.yaml', '.yml', '.json'))
            and not f.startswith(('node_modules/', '.git/'))
        ]

        self.state: Dict[str, Any] = {
            "pipeline_id":       self.pipeline_id,
            "commit_sha":        commit_sha,
            "branch":            branch,
            "changed_files":     self.changed_files,
            "status":            "running",
            "current_step":      0,
            "total_steps":       5,
            "steps":             [],
            "findings":          [],
            "severity_summary":  {"critical": 0, "high": 0, "medium": 0, "low": 0},
            "openclaw_report":   "",
            "ea_decision":       {},
            "auto_fix_triggered": False,
            "started_at":        self.started_at.isoformat(),
            "completed_at":      None,
            "message":           "",
        }
        self._sync_global()

    # ── State helpers ─────────────────────────────────────────────────────────

    def _sync_global(self):
        global _current_pipeline
        with _pipeline_lock:
            _current_pipeline = dict(self.state)

    def _step_start(self, num: int, name: str, agent: str):
        self.state["current_step"] = num
        self.state["steps"].append({
            "step":         num,
            "name":         name,
            "agent":        agent,
            "status":       "running",
            "started_at":   datetime.now().isoformat(),
            "completed_at": None,
            "summary":      "",
        })
        self._sync_global()
        logger.info("[CodeReview] ▶ Step %d/5 %s (%s)", num, name, agent)

    def _step_done(self, num: int, summary: str, ok: bool = True):
        for s in self.state["steps"]:
            if s["step"] == num:
                s["status"]       = "ok" if ok else "error"
                s["completed_at"] = datetime.now().isoformat()
                s["summary"]      = summary[:300]
        self._sync_global()
        logger.info("[CodeReview] %s Step %d — %s", "✓" if ok else "✗", num, summary[:100])

    def _finish(self, status: str, message: str):
        self.state["status"]       = status
        self.state["completed_at"] = datetime.now().isoformat()
        self.state["message"]      = message
        self._sync_global()

    # ── Main pipeline ─────────────────────────────────────────────────────────

    def run(self):
        """Execute full pipeline. Designed to run in daemon thread."""
        try:
            self._notify_start()

            # Step 1 ─ read files
            self._step_start(1, "讀取變更檔案", "system")
            file_contents = self._read_changed_files()
            if not file_contents:
                self._step_done(1, "無有效 Python/YAML 變更，跳過 Review")
                self._finish("skipped", "無有效變更檔案")
                return
            self._step_done(1, f"讀取 {len(file_contents)} 個檔案")

            # Step 2 ─ Hermes scan
            self._step_start(2, "Hermes 程式碼掃描", "Hermes")
            findings = self._hermes_scan(file_contents)
            cnt = self.state["severity_summary"]
            self._step_done(2, f"CRITICAL={cnt['critical']} HIGH={cnt['high']} MEDIUM={cnt['medium']} LOW={cnt['low']}")

            # Step 3 ─ OpenClaw assessment
            self._step_start(3, "OpenClaw 架構品質評估", "OpenClaw")
            openclaw_report = self._openclaw_assess(file_contents, findings)
            self.state["openclaw_report"] = openclaw_report
            self._step_done(3, openclaw_report[:120] if openclaw_report else "（Gemini 未回應）")

            # Step 4 ─ ElephantAlpha orchestration
            self._step_start(4, "ElephantAlpha 決策協調", "ElephantAlpha")
            ea = self._ea_orchestrate(findings, openclaw_report)
            self.state["ea_decision"] = ea
            self._step_done(4, f"優先度={ea.get('priority','?')} auto_fix={ea.get('auto_fix',False)}")

            # Step 5 ─ NemoTron dispatch
            self._step_start(5, "NemoTron 行動派遣", "NemoTron")
            dispatch = self._nemotron_dispatch(ea, findings)
            self._step_done(5, f"寫入 {dispatch['actions']} 筆 action_plans，自動修復={'是' if dispatch['auto_fix'] else '否'}")

            # Persist + notify
            self._save_to_db(findings, openclaw_report, ea)
            self._notify_complete(findings, openclaw_report, ea)
            self._finish("completed", "Pipeline 執行完成")

        except Exception as e:
            logger.error("[CodeReview] Pipeline 例外: %s", e, exc_info=True)
            self._finish("error", str(e)[:200])
            self._notify_error(str(e))

    # ── Step 1：讀取檔案 ───────────────────────────────────────────────────────

    def _read_changed_files(self) -> Dict[str, str]:
        project_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
        contents: Dict[str, str] = {}
        for rel_path in self.changed_files[:8]:          # 最多 8 個檔案
            abs_path = os.path.join(project_root, rel_path)
            try:
                with open(abs_path, encoding="utf-8", errors="ignore") as fh:
                    raw = fh.read()
                    contents[rel_path] = raw[:6000] + ("\n... (截斷)" if len(raw) > 6000 else "")
            except OSError:
                logger.debug("[CodeReview] 無法讀取 %s（部署路徑不同？）", rel_path)
        return contents

    # ── Step 2：Hermes 掃描 ───────────────────────────────────────────────────

    def _hermes_scan(self, files: Dict[str, str]) -> List[Dict]:
        try:
            from services.ai_provider import ai_provider_service

            # 最多送 4 個檔案給 Hermes，避免 context overflow
            files_text = "\n\n".join(
                f"### {name}\n```python\n{content}\n```"
                for name, content in list(files.items())[:4]
            )

            prompt = f"""你是資深程式碼審查工程師，請掃描以下程式碼並列出所有問題。

{files_text}

輸出格式：純 JSON 陣列，每項包含以下欄位：
- severity: "CRITICAL" | "HIGH" | "MEDIUM" | "LOW"
- type: "bug" | "security" | "performance" | "maintainability"
- file: 檔案名稱
- line_hint: 約略行號或函式名稱
- description: 問題說明（繁體中文，精簡一句）
- suggestion: 修復建議（繁體中文，精簡一句）

只輸出 JSON 陣列，不含其他文字。無問題時輸出 []"""

            resp = ai_provider_service.generate(
                prompt=prompt,
                provider="ollama",
                model="hermes3:latest",
                temperature=0.1,
                timeout=120,
            )
            if not (resp and resp.success):
                logger.warning("[CodeReview] Hermes 未回應: %s", getattr(resp, "error", ""))
                return []

            match = re.search(r"\[.*\]", resp.content, re.DOTALL)
            if not match:
                return []
            findings = json.loads(match.group())

            # 更新 severity 計數
            for f in findings:
                sev = f.get("severity", "LOW").lower()
                if sev in self.state["severity_summary"]:
                    self.state["severity_summary"][sev] += 1
            self.state["findings"] = findings
            self._sync_global()
            return findings

        except Exception as e:
            logger.warning("[CodeReview] Hermes 掃描失敗: %s", e)
            return []

    # ── Step 3：OpenClaw 評估 ──────────────────────────────────────────────────

    def _openclaw_assess(self, files: Dict[str, str], findings: List[Dict]) -> str:
        if not GEMINI_API_KEY:
            logger.warning("[CodeReview] GEMINI_API_KEY 未設定，跳過 OpenClaw")
            return ""
        try:
            import google.generativeai as genai
            genai.configure(api_key=GEMINI_API_KEY)
            model = genai.GenerativeModel(
                model_name=REVIEW_MODEL,
                generation_config=genai.types.GenerationConfig(
                    temperature=0.3, max_output_tokens=1500,
                ),
                system_instruction=(
                    "你是 OpenClaw 程式碼品質戰略分析師，以技術主管視角評估部署後程式碼。"
                    "語言：繁體中文。風格：精準、數據導向、可執行建議。"
                ),
            )
            sev = self.state["severity_summary"]
            findings_json = json.dumps(findings[:8], ensure_ascii=False, indent=2)
            files_list = "\n".join(f"- {k} ({len(v)} 字元)" for k, v in list(files.items())[:5])

            prompt = f"""【部署資訊】Commit {self.commit_sha[:8]} @ {self.branch}
【變更檔案】
{files_list}

【Hermes 掃描摘要】CRITICAL={sev['critical']} HIGH={sev['high']} MEDIUM={sev['medium']} LOW={sev['low']}
【Hermes 詳細問題】
{findings_json}

請產出程式碼品質評估（使用 HTML <b> 標題，150字以內）：

<b>🔍 整體風險等級</b>（CRITICAL / HIGH / MEDIUM / LOW，一句理由）
<b>⚠️ 最需關注問題</b>（TOP 2，具體說明）
<b>💡 架構優化方向</b>（1 條長期建議）
<b>✅ 本次部署亮點</b>（值得肯定的地方）"""

            resp = model.generate_content(prompt, request_options={"timeout": 90})
            return resp.text or ""
        except Exception as e:
            logger.warning("[CodeReview] OpenClaw 評估失敗: %s", e)
            return ""

    # ── Step 4：ElephantAlpha 決策 ─────────────────────────────────────────────

    def _ea_orchestrate(self, findings: List[Dict], openclaw_report: str) -> Dict:
        sev = self.state["severity_summary"]
        critical_n = sev["critical"]
        high_n     = sev["high"]

        # 嘗試呼叫 ElephantAlpha 做精細判斷
        try:
            from services.elephant_service import elephant_service

            top3 = json.dumps(
                [f for f in findings if f.get("severity") in ("CRITICAL", "HIGH")][:3],
                ensure_ascii=False,
            )
            prompt = f"""你是 Elephant Alpha，負責協調 Code Review 後的修復決策。

【部署】commit={self.commit_sha[:8]} branch={self.branch}
【問題統計】CRITICAL={critical_n} HIGH={high_n} MEDIUM={sev['medium']} LOW={sev['low']}
【Top 問題】{top3}
【OpenClaw評估摘要】{openclaw_report[:300]}

請以 JSON 回答（不含其他文字）：
{{
  "priority": "critical|high|medium|low",
  "auto_fix": true|false,
  "reasoning": "決策理由（繁體中文，一句話，需含具體數字）",
  "fix_files": ["需自動修復的檔案（最多3個，只填 CRITICAL/HIGH 問題的檔案）"],
  "human_review_needed": true|false
}}

規則：
- CRITICAL ≥ 1 → priority=critical, auto_fix=true
- HIGH ≥ 3 → priority=high, auto_fix=true
- HIGH 1-2 → priority=high, auto_fix=false, human_review_needed=true
- 其餘 → priority=medium|low, auto_fix=false"""

            resp = elephant_service.generate(
                prompt=prompt,
                json_mode=True,
                temperature=0.1,
                timeout=60,
            )
            if resp.success:
                return json.loads(resp.content)
        except Exception as e:
            logger.warning("[CodeReview] ElephantAlpha 決策失敗，回退規則: %s", e)

        # 規則 fallback
        auto_fix = critical_n > 0 or high_n >= 3
        priority = (
            "critical" if critical_n > 0 else
            "high"     if high_n     > 0 else
            "medium"   if sev["medium"] > 0 else "low"
        )
        fix_files = list({
            f.get("file", "") for f in findings
            if f.get("severity") in ("CRITICAL", "HIGH") and f.get("file")
        })[:3]

        return {
            "priority":            priority,
            "auto_fix":            auto_fix,
            "reasoning":           f"規則判斷：CRITICAL={critical_n} HIGH={high_n}，{'觸發自動修復' if auto_fix else '需人工審查'}",
            "fix_files":           fix_files,
            "human_review_needed": not auto_fix and (critical_n + high_n) > 0,
        }

    # ── Step 5：NemoTron 派遣 ──────────────────────────────────────────────────

    def _nemotron_dispatch(self, ea: Dict, findings: List[Dict]) -> Dict:
        auto_fix  = ea.get("auto_fix", False)
        fix_files = ea.get("fix_files", [])
        priority_map = {"critical": 1, "high": 2, "medium": 3, "low": 4}
        priority_num  = priority_map.get(ea.get("priority", "low"), 4)

        actions_created = 0
        session = get_session()
        try:
            # 每個需修復的檔案建立一筆 action_plan
            for fpath in fix_files[:3]:
                related = [f for f in findings if f.get("file") == fpath][:3]
                desc = f"Code Review 修復：{fpath}｜{', '.join(f.get('description','')[:40] for f in related)}"
                session.execute(text("""
                    INSERT INTO action_plans
                        (action_type, description, status, priority, metadata_json, created_at)
                    VALUES
                        ('code_review_fix', :desc, :status, :priority, :meta, NOW())
                """), {
                    "desc":     desc[:500],
                    "status":   "auto_pending" if auto_fix else "pending_review",
                    "priority": priority_num,
                    "meta":     json.dumps({
                        "pipeline_id":  self.pipeline_id,
                        "commit_sha":   self.commit_sha,
                        "file":         fpath,
                        "auto_fix":     auto_fix,
                        "ea_priority":  ea.get("priority"),
                        "findings":     related,
                    }, ensure_ascii=False),
                })
                actions_created += 1
            session.commit()
        except Exception as e:
            logger.warning("[CodeReview] action_plans 寫入失敗: %s", e)
            session.rollback()
        finally:
            session.close()

        # 觸發 AiderHeal（非阻塞）
        if auto_fix and fix_files:
            self.state["auto_fix_triggered"] = True
            self._sync_global()
            self._trigger_aider_heal(findings, fix_files)

        return {"actions": actions_created, "auto_fix": auto_fix}

    def _trigger_aider_heal(self, findings: List[Dict], fix_files: List[str]):
        """非阻塞觸發 AiderHeal 自動修復"""
        def _heal_worker():
            try:
                from services.aider_heal_executor import execute_code_fix
                for fpath in fix_files[:2]:             # 最多同時修 2 個檔案
                    related = [f for f in findings if f.get("file") == fpath]
                    if not related:
                        continue
                    worst = sorted(related, key=lambda x: {"CRITICAL":0,"HIGH":1,"MEDIUM":2,"LOW":3}.get(x.get("severity","LOW"),3))[0]
                    result = execute_code_fix(
                        error_type=f"code_review_{worst.get('type','bug')}",
                        error_message=worst.get("description", "Code Review 發現問題"),
                        target_file=fpath,
                        context={"suggestion": worst.get("suggestion", ""), "pipeline_id": self.pipeline_id},
                    )
                    logger.info("[CodeReview] AiderHeal %s → %s", fpath, result.get("message", ""))
            except Exception as e:
                logger.error("[CodeReview] AiderHeal 觸發失敗: %s", e)

        t = threading.Thread(target=_heal_worker, daemon=True, name=f"aider-heal-{self.commit_sha[:8]}")
        t.start()

    # ── DB 持久化 ──────────────────────────────────────────────────────────────

    def _save_to_db(self, findings: List[Dict], openclaw_report: str, ea: Dict):
        session = get_session()
        try:
            session.execute(text("""
                INSERT INTO ai_insights
                    (insight_type, content, confidence, created_by,
                     status, metadata_json, period, created_at)
                VALUES
                    ('code_review_result', :content, :conf, 'code_review_pipeline',
                     'active', :meta, :period, NOW())
            """), {
                "content": json.dumps({
                    "findings":         findings,
                    "openclaw_report":  openclaw_report,
                    "ea_decision":      ea,
                    "severity_summary": self.state["severity_summary"],
                }, ensure_ascii=False)[:8000],
                "conf":   0.90,
                "meta":   json.dumps({
                    "pipeline_id":       self.pipeline_id,
                    "commit_sha":        self.commit_sha,
                    "branch":            self.branch,
                    "changed_files":     self.changed_files,
                    "auto_fix_triggered": self.state["auto_fix_triggered"],
                }, ensure_ascii=False),
                "period": datetime.now().strftime("%Y-%m-%d"),
            })
            session.commit()
            logger.info("[CodeReview] ai_insights 寫入成功 pipeline=%s", self.pipeline_id)
        except Exception as e:
            logger.error("[CodeReview] DB 寫入失敗: %s", e)
            session.rollback()
        finally:
            session.close()

    # ── Telegram 通知 ─────────────────────────────────────────────────────────

    def _notify_start(self):
        try:
            from services.telegram_templates import _send_telegram_raw
            files_list = "\n".join(f"  • {f}" for f in self.changed_files[:5])
            if len(self.changed_files) > 5:
                files_list += f"\n  （+{len(self.changed_files)-5} 個）"
            _send_telegram_raw(
                f"🔍 <b>Code Review 啟動</b>\n"
                f"══════════════════════════\n"
                f"📦 Commit　<code>{self.commit_sha[:8]}</code>  🌿 {self.branch}\n"
                f"📝 變更檔案：\n{files_list}\n"
                f"══════════════════════════\n"
                f"🤖 <i>Hermes → OpenClaw → Elephant Alpha → NemoTron</i>\n"
                f"📊 即時進度：https://mo.wooo.work/code-review/"
            )
        except Exception as e:
            logger.warning("[CodeReview] 啟動通知失敗: %s", e)

    def _notify_complete(self, findings: List[Dict], openclaw_report: str, ea: Dict):
        try:
            from services.telegram_templates import _send_telegram_raw
            sev      = self.state["severity_summary"]
            priority = ea.get("priority", "medium")
            auto_fix = ea.get("auto_fix", False)

            icon = {"critical": "🔴", "high": "🟠", "medium": "🟡", "low": "🟢"}.get(priority, "🟡")
            top_issues = [f for f in findings if f.get("severity") in ("CRITICAL", "HIGH")][:3]
            issues_lines = "\n".join(
                f"  {'🔴' if f['severity']=='CRITICAL' else '🟠'} [{f['severity']}] {f.get('description','')} — <i>{f.get('file','')}</i>"
                for f in top_issues
            ) or "  ✅ 無高風險問題"

            msg = (
                f"{icon} <b>Code Review 完成</b>  ·  <code>{self.commit_sha[:8]}</code>\n"
                f"══════════════════════════\n"
                f"🔴 CRITICAL　<b>{sev['critical']}</b>　"
                f"🟠 HIGH　<b>{sev['high']}</b>　"
                f"🟡 MEDIUM　<b>{sev['medium']}</b>　"
                f"🟢 LOW　<b>{sev['low']}</b>\n"
                f"══════════════════════════\n"
                f"<b>⚠️ 主要問題</b>\n{issues_lines}\n"
            )
            if openclaw_report:
                msg += f"\n{openclaw_report[:400]}\n"

            fix_status = "🔧 已觸發自動修復（AiderHeal）" if auto_fix else (
                "👁 需人工審查" if ea.get("human_review_needed") else "✅ 無需修復動作"
            )
            msg += (
                f"══════════════════════════\n"
                f"🤖 Elephant Alpha：<b>{priority.upper()}</b>　{fix_status}\n"
                f"📊 完整報告：https://mo.wooo.work/code-review/"
            )
            _send_telegram_raw(msg)
        except Exception as e:
            logger.warning("[CodeReview] 完成通知失敗: %s", e)

    def _notify_error(self, error: str):
        try:
            from services.telegram_templates import _send_telegram_raw
            _send_telegram_raw(
                f"🚨 <b>Code Review Pipeline 失敗</b>\n"
                f"Commit：<code>{self.commit_sha[:8]}</code> @ {self.branch}\n"
                f"錯誤：{error[:200]}\n"
                f"📊 查看：https://mo.wooo.work/code-review/"
            )
        except Exception:
            pass


# ═══════════════════════════════════════════════════════════════════════════════
# 公開 API
# ═══════════════════════════════════════════════════════════════════════════════

def trigger_post_deploy_review(
    commit_sha:    str,
    changed_files: List[str],
    branch:        str = "main",
    deploy_type:   str = "sync",
) -> str:
    """
    啟動 Pipeline（後台 daemon thread）。
    回傳 pipeline_id。由 routes/code_review_routes.py 的 webhook 端點呼叫。
    """
    pipeline = CodeReviewPipeline(commit_sha, changed_files, branch, deploy_type)
    t = threading.Thread(
        target=pipeline.run,
        daemon=True,
        name=f"code-review-{commit_sha[:8]}",
    )
    t.start()
    logger.info("[CodeReview] 已派發 pipeline=%s files=%d", pipeline.pipeline_id, len(pipeline.changed_files))
    return pipeline.pipeline_id


def get_current_state() -> Dict[str, Any]:
    """前端 polling 用：取得目前 pipeline 即時狀態"""
    with _pipeline_lock:
        return dict(_current_pipeline)


def get_history(limit: int = 20) -> List[Dict]:
    """取得 ai_insights 中歷史 code_review_result 記錄"""
    session = get_session()
    try:
        rows = session.execute(text("""
            SELECT id, content, confidence, metadata_json, created_at, status
            FROM ai_insights
            WHERE insight_type = 'code_review_result'
            ORDER BY created_at DESC
            LIMIT :lim
        """), {"lim": limit}).fetchall()

        results = []
        for r in rows:
            meta, content = {}, {}
            try:
                meta    = json.loads(r[3]) if r[3] else {}
                content = json.loads(r[1]) if r[1] else {}
            except Exception:
                pass
            sev = content.get("severity_summary", {})
            results.append({
                "id":               r[0],
                "pipeline_id":      meta.get("pipeline_id", ""),
                "commit_sha":       meta.get("commit_sha", "")[:8],
                "branch":           meta.get("branch", ""),
                "changed_files":    meta.get("changed_files", []),
                "severity_summary": sev,
                "total_issues":     sum(sev.values()),
                "auto_fix":         meta.get("auto_fix_triggered", False),
                "ea_decision":      content.get("ea_decision", {}),
                "created_at":       r[4].isoformat() if r[4] else "",
                "status":           r[5] or "active",
            })
        return results
    except Exception as e:
        logger.warning("[CodeReview] 歷史讀取失敗: %s", e)
        return []
    finally:
        session.close()


def verify_internal_token(request_token: str) -> bool:
    """驗證 CD webhook 來源 token。未設定 env 時直接放行（dev 環境）"""
    if not INTERNAL_TOKEN:
        return True
    return request_token == INTERNAL_TOKEN