maestro/src/daemon/orchestrator.ts

import { Store } from '../store/index.js';
import type { Project, Task, ReviewVerdict } from '../model/types.js';
import { rankByScore } from '../model/scoring.js';
import { createWorktree, worktreeDiff, type WorktreeDiff, type WorktreeInfo } from '../executor/worktree.js';
import { runTask, type RunnerFn } from '../executor/runner.js';
import { runVerify, type VerifyFn } from '../executor/verify.js';
import { reviewCode, reviewSecurity, type ReviewerFn } from '../executor/reviewer.js';
import { pickModel } from '../executor/models.js';

/** 失败后默认最多自动重试次数（项目级 maxRetries 未设置时回退此值） */
export const DEFAULT_MAX_RETRIES = 2;

/** 指数退避基础等待（第 n 次重试等待 BASE * 2^(n-1)，上限 MAX） */
const BACKOFF_BASE_MS = 30_000;   // 30s
const BACKOFF_MAX_MS = 10 * 60_000; // 10min

/** 第 attempt 次重试（1-indexed）的等待毫秒 */
export function computeBackoffMs(attempt: number): number {
  return Math.min(BACKOFF_BASE_MS * (2 ** (attempt - 1)), BACKOFF_MAX_MS);
}

export interface OrchestratorLogger {
  info(msg: string): void;
  error(msg: string): void;
}

/** 依赖注入点：测试传 mock，生产用真实现 */
export interface OrchestratorDeps {
  runner: RunnerFn;
  reviewCode: ReviewerFn;        // code review（kind=reviewer 的 run）
  reviewSecurity: ReviewerFn;    // 安全审计（kind=security 的 run）
  verify: VerifyFn;
  createWorktree: (repoPath: string, taskId: string, baseBranch: string) => Promise<WorktreeInfo>;
  worktreeDiff: (repoPath: string, dir: string, branch: string, baseBranch: string) => Promise<WorktreeDiff>;
  /** 当前时间戳（ms）；测试注入可控时钟（默认 Date.now） */
  nowMs: () => number;
}

export interface Orchestrator {
  /** 跑一轮领取（同步领取 + 异步执行，不阻塞）。错误只记日志。 */
  tick(): void;
  /** 等待所有在途执行收尾（测试用） */
  drain(): Promise<void>;
  /** 在途任务 id（防同任务重复领取） */
  readonly inflight: ReadonlyMap<string, string>;
}

/**
 * 编排器核心（与定时器解耦，便于测试）。
 * 每轮对 status=active 且 autonomy≠manual 的项目：在途数 < concurrency 时领任务——
 * ready 叶子（deps 全 done；auto-easy 只领 easy）或 queued（重试/重启遗留）。
 * 成功 → setResult + exec_review；失败 → failed → 重试 ≤MAX_RETRIES 次 → needs_attention。
 */
export function createOrchestrator(store: Store, log: OrchestratorLogger, deps: Partial<OrchestratorDeps> = {}): Orchestrator {
  const d: OrchestratorDeps = { runner: runTask, reviewCode, reviewSecurity, verify: runVerify, createWorktree, worktreeDiff, nowMs: Date.now, ...deps };
  const inflight = new Map<string, string>();       // taskId → projectId
  const pending = new Set<Promise<void>>();
  /** 退避表：taskId → 下次可领取时间戳（ms）；daemon 重启后清空，已有 queued 任务即刻可领 */
  const backoffUntil = new Map<string, number>();

  /**
   * 本项目可领取的任务：queued（重试/孤儿）+ ready 叶子且 deps 全 done；auto-easy 只挑 easy。
   * 按调度分降序返回（score = 自身分 + 已完成依赖分 + 等待解锁的 blocked 任务分，见 model/scoring.ts）。
   */
  function claimable(project: Project): Array<{ task: Task; score: number }> {
    const nowMs = d.nowMs();
    const tasks = store.listTasks(project.id);
    const byId = new Map(tasks.map((t) => [t.id, t]));
    const parents = new Set(tasks.filter((t) => t.parentId).map((t) => t.parentId as string));
    const easyOnly = project.autonomy === 'auto-easy';
    const candidates = tasks.filter((t) => {
      if (inflight.has(t.id)) return false;
      if (t.status !== 'ready' && t.status !== 'queued') return false;
      if (parents.has(t.id)) return false;                                  // 非叶子（容器）跳过
      if (easyOnly && t.complexity !== 'easy') return false;
      const until = backoffUntil.get(t.id);
      if (until !== undefined && nowMs < until) return false;               // 退避冷却中
      return t.deps.every((dep) => byId.get(dep)?.status === 'done');
    });
    return rankByScore(candidates, tasks);
  }

  /** 跑一个独立复审 run（kind=reviewer/security）。任何失败折叠为 summary=失败原因、verdict=null，不抛错、不挡任务。 */
  async function runOneReview(
    kind: 'reviewer' | 'security',
    fn: ReviewerFn,
    task: Task,
    project: Project,
    wt: WorktreeInfo,
    executorReport: string,
  ): Promise<{ summary: string | null; verdict: ReviewVerdict | null }> {
    const label = kind === 'reviewer' ? 'code review' : '安全审计';
    let reviewRunId: string | null = null;
    try {
      const review = store.startRun(task.id, kind, { worktree: wt.dir, branch: wt.branch });
      reviewRunId = review.id;
      log.info(`${label} 任务 ${task.id} run=${review.id} model=${pickModel(task, project, 'reviewer')}`);
      const rv = await fn(task, project, wt, review.id, executorReport);
      store.finishRun(review.id, 'succeeded', {
        transcriptRef: rv.transcriptRef ?? undefined,
        claudeSessionId: rv.sessionId ?? undefined,
      });
      log.info(`任务 ${task.id} ${label} 完成 verdict=${rv.verdict ?? '（未解析到）'}`);
      return { summary: rv.summary, verdict: rv.verdict };
    } catch (e) {
      const reMsg = (e as Error).message;
      if (reviewRunId) {
        try { store.finishRun(reviewRunId, 'failed', { error: reMsg }); } catch { /* 收尾失败不影响主流程 */ }
      }
      log.error(`任务 ${task.id} ${label} 失败（不挡任务，照常进 exec_review）：${reMsg}`);
      return { summary: `自动复审失败：${reMsg}`, verdict: null };
    }
  }

  /** 单任务全流程：executing → worktree → run → verify → code review run → 安全审计 run（均失败不挡）→ setResult(四字段) → exec_review / failed(重试) */
  async function executeTask(project: Project, task: Task): Promise<void> {
    let runId: string | null = null;
    let runClosed = false;
    try {
      store.transition(task.id, 'executing', { by: 'orchestrator' });
      const wt = await d.createWorktree(project.repoPath, task.id, project.defaultBranch);
      const run = store.startRun(task.id, 'executor', { worktree: wt.dir, branch: wt.branch });
      runId = run.id;
      log.info(`执行任务 ${task.id}「${task.title}」 run=${run.id} worktree=${wt.dir}`);

      const rr = await d.runner(task, project, wt, run.id);
      if (!rr.ok) {
        store.finishRun(run.id, 'failed', {
          error: rr.error ?? '执行失败',
          transcriptRef: rr.transcriptRef ?? undefined,
          claudeSessionId: rr.sessionId ?? undefined,
        });
        runClosed = true;
        throw new Error(rr.error ?? '执行失败');
      }

      const vr = await d.verify(project, wt.dir, run.id);
      if (!vr.ok) {
        store.finishRun(run.id, 'failed', {
          error: vr.error ?? 'verify 失败',
          transcriptRef: rr.transcriptRef ?? undefined,
          claudeSessionId: rr.sessionId ?? undefined,
        });
        runClosed = true;
        throw new Error(vr.error ?? 'verify 失败');
      }

      const diff = await d.worktreeDiff(project.repoPath, wt.dir, wt.branch, project.defaultBranch);

      // 双复审（顺序）：code review run（kind=reviewer）→ 安全审计 run（kind=security）。任一失败不挡任务。
      const code = await runOneReview('reviewer', d.reviewCode, task, project, wt, rr.finalText ?? '');
      const sec = await runOneReview('security', d.reviewSecurity, task, project, wt, rr.finalText ?? '');

      store.setResult(task.id, {
        branch: wt.branch, worktree: wt.dir,
        diffSummary: diff.diffSummary, commits: diff.commits, prUrl: null,
        summary: code.summary, verdict: code.verdict,
        securitySummary: sec.summary, securityVerdict: sec.verdict,
      });
      store.transition(task.id, 'exec_review', { by: 'orchestrator', runId: run.id });
      store.finishRun(run.id, 'succeeded', {
        transcriptRef: rr.transcriptRef ?? undefined,
        claudeSessionId: rr.sessionId ?? undefined,
      });
      log.info(`任务 ${task.id} 执行完成 → exec_review（${diff.commits.length} commits）`);
    } catch (e) {
      const msg = (e as Error).message;
      log.error(`任务 ${task.id} 执行失败：${msg}`);
      try {
        if (runId && !runClosed) {
          store.finishRun(runId, 'failed', { error: msg });                 // worktree 创建后抛错（diff 等）时收尾
        } else if (!runId) {
          // run 还没建（如 createWorktree 失败）：补记一条 failed run，保证重试计数不漏
          const r = store.startRun(task.id, 'executor');
          store.finishRun(r.id, 'failed', { error: msg });
        }
        store.transition(task.id, 'failed', { by: 'orchestrator', error: msg });
        const allFailed = store.listRuns(task.id).filter((r) => r.kind === 'executor' && r.status === 'failed').length;
        // 相对于上次手动重投基线的净失败次数（retryBaseline 在 requeueTask 时设置）
        const netFailed = Math.max(0, allFailed - task.retryBaseline);
        const priorNetFailed = Math.max(0, netFailed - 1);                  // 不含本次
        const maxRetries = project.maxRetries;
        if (priorNetFailed < maxRetries) {
          const attempt = priorNetFailed + 1;                               // 本次是第几次重试
          const backoffMs = computeBackoffMs(attempt);
          backoffUntil.set(task.id, d.nowMs() + backoffMs);
          store.transition(task.id, 'queued', { by: 'orchestrator', retry: attempt, backoffMs });
          log.info(`任务 ${task.id} 重新入队（第 ${attempt} 次重试，退避 ${backoffMs}ms 后可领）`);
        } else {
          backoffUntil.delete(task.id);
          store.transition(task.id, 'needs_attention', { by: 'orchestrator', allFailed, netFailed });
          log.error(`任务 ${task.id} 净失败 ${netFailed} 次（上限 ${maxRetries}）→ needs_attention`);
        }
      } catch (e2) {
        log.error(`任务 ${task.id} 失败收尾出错：${(e2 as Error).message}`);
      }
    }
  }

  function tick(): void {
    try {
      for (const p of store.listProjects()) {
        if (p.status !== 'active' || p.autonomy === 'manual') continue;
        let active = 0;
        for (const pid of inflight.values()) if (pid === p.id) active++;
        if (active >= p.concurrency) continue;

        for (const { task, score } of claimable(p)) {
          if (active >= p.concurrency) break;
          try {
            if (task.status === 'ready') store.transition(task.id, 'queued', { by: 'orchestrator', score });
          } catch (e) {
            log.error(`任务 ${task.id} 入队失败：${(e as Error).message}`);
            continue;
          }
          log.info(`领取任务 ${task.id}「${task.title}」score=${score} model=${pickModel(task, p, 'executor')}`);
          inflight.set(task.id, p.id);
          active++;
          const job: Promise<void> = executeTask(p, { ...task, status: 'queued' })
            .catch((e) => log.error(`任务 ${task.id} 执行异常：${(e as Error).message}`))
            .finally(() => {
              inflight.delete(task.id);
              pending.delete(job);
            });
          pending.add(job);
        }
      }
    } catch (e) {
      log.error(`编排器轮询失败：${(e as Error).message}`);
    }
  }

  async function drain(): Promise<void> {
    while (pending.size > 0) await Promise.allSettled([...pending]);
  }

  return { tick, drain, inflight };
}

/**
 * 接线入口：MAESTRO_ORCH_INTERVAL（秒）控制轮询间隔，默认 15，0=关闭。
 * 返回 timer 供 shutdown 时 clearInterval。
 */
export function startOrchestrator(
  store: Store,
  app: { log: OrchestratorLogger },
  deps: Partial<OrchestratorDeps> = {},
): NodeJS.Timeout | null {
  const intervalSec = Number(process.env.MAESTRO_ORCH_INTERVAL ?? 15);
  if (!Number.isFinite(intervalSec) || intervalSec <= 0) {
    app.log.info('编排器已关闭（MAESTRO_ORCH_INTERVAL=0）');
    return null;
  }
  const orch = createOrchestrator(store, app.log, deps);
  const timer = setInterval(() => orch.tick(), intervalSec * 1000);
  timer.unref();
  app.log.info(`编排器已启用：每 ${intervalSec}s 一轮领取（autonomy≠manual 的 active 项目）`);
  return timer;
}