agentic-coding-workflow/internal/app/app.go

// Package app 是应用的装配根：把 config、logger、DB 池、各业务模块、HTTP
// 路由按依赖顺序拼起来，并暴露 Run 给入口程序。本包只做组合，不引入新业务。
//
// 启动顺序：
//  1. 用 config.Env 派生 logger（FromEnv 在生产环境默认开启 source）。
//  2. 跑 migrations（先 migrate 再开池，避免应用启动时表还没建好；migrate
//     失败直接返错由调用方记录、退出）。
//  3. 创建 *db.Pool（实际是 *pgxpool.Pool 别名），失败时不再继续。
//  4. 模块装配：audit recorder -> user repo+service（注入 audit）-> notify
//     repo+dispatcher（注入 dummy notifier）-> notify handler（复用 userSvc
//     作为 SessionResolver，user.Service 接口本身满足该契约）。
//  5. 路由：httpx.NewRouter 返回 chi.Router 接口，直接挂 RequestID/Recover/
//     Logger 三个全局中间件，再让模块 Mount 自己的子路由。
//  6. 可选 bootstrap：仅当配置里同时给了 admin 邮箱/密码时尝试。失败仅
//     Warn 不阻塞启动。
//  7. 把所有依赖封装到 *App，由 Run 负责启动 http server 与优雅关停。
package app

import (
	"context"
	"embed"
	"errors"
	"fmt"
	"io/fs"
	"log/slog"
	"net/http"
	"os"
	"path/filepath"
	"time"

	"github.com/google/uuid"
	"github.com/yan1h/agent-coding-workflow/internal/audit"
	"github.com/yan1h/agent-coding-workflow/internal/config"
	"github.com/yan1h/agent-coding-workflow/internal/infra/crypto"
	"github.com/yan1h/agent-coding-workflow/internal/infra/db"
	"github.com/yan1h/agent-coding-workflow/internal/infra/git"
	"github.com/yan1h/agent-coding-workflow/internal/infra/logger"
	"github.com/yan1h/agent-coding-workflow/internal/infra/notify"
	"github.com/yan1h/agent-coding-workflow/internal/project"
	httpx "github.com/yan1h/agent-coding-workflow/internal/transport/http"
	"github.com/yan1h/agent-coding-workflow/internal/transport/http/middleware"
	"github.com/yan1h/agent-coding-workflow/internal/user"
	"github.com/yan1h/agent-coding-workflow/internal/workspace"
)

// App 持有运行期所有需要被关停的资源。字段保持非导出，避免外部直接操作。
type App struct {
	cfg        *config.Config
	log        *slog.Logger
	pool       *db.Pool
	server     *http.Server
	dispatcher *notify.Dispatcher
}

// New 把所有模块按启动顺序拼装好。任何阶段失败都会返回错误，不留半成品资源。
// dist 是 cmd/server 用 //go:embed 注入的前端构建产物，根目录为 web/dist；
// fs.Sub 之后交给 SPA handler 提供静态文件 + index.html 回退。
func New(ctx context.Context, cfg *config.Config, dist embed.FS) (*App, error) {
	log := logger.FromEnv(cfg.Env, nil)

	// 跑 migrations 必须在开池前；否则首次启动连不存在的表会让 ping 失败。
	if err := db.Migrate(ctx, cfg.DB.DSN, "file://migrations"); err != nil {
		return nil, fmt.Errorf("migrate: %w", err)
	}

	pool, err := db.NewPool(ctx, cfg.DB.DSN)
	if err != nil {
		return nil, err
	}

	auditRec := audit.NewPostgresRecorder(pool)

	userRepo := user.NewPostgresRepository(pool)
	userSvc := user.NewService(userRepo, auditRec)

	projectRepo := project.NewPostgresRepository(pool)
	projectSvc := project.NewProjectService(projectRepo, auditRec)
	requirementSvc := project.NewRequirementService(projectRepo, auditRec)
	issueSvc := project.NewIssueService(projectRepo, auditRec)

	notifyRepo := notify.NewPostgresRepository(pool)
	notifyDispatcher := notify.NewDispatcher(notifyRepo, log)
	notifyDispatcher.Register(notify.NewDummyNotifier(log))

	spaSub, err := fs.Sub(dist, "web/dist")
	if err != nil {
		return nil, fmt.Errorf("sub fs: %w", err)
	}
	spaHandler := httpx.NewSPAHandler(spaSub)

	r := httpx.NewRouter(httpx.RouterDeps{
		HealthCheck: func(req *http.Request) error {
			return db.HealthCheck(req.Context(), pool)
		},
		ReadyCheck: func(req *http.Request) error {
			return db.HealthCheck(req.Context(), pool)
		},
		SPAHandler: spaHandler,
	})

	// chi.Router 接口本身已经声明 Use，无需类型断言到 *chi.Mux。
	r.Use(middleware.RequestID)
	r.Use(middleware.Recover(log))
	r.Use(middleware.Logger(log))

	// userSvc 直接满足 middleware.SessionResolver（ResolveSession 同名同签名）。
	user.NewHandler(userSvc).Mount(r)
	notify.NewHandler(notifyRepo, userSvc).Mount(r)
	project.NewHandler(projectSvc, requirementSvc, issueSvc, userSvc, userAdminAdapter{svc: userSvc}).Mount(r)

	// ===== workspace 模块装配 =====
	// 依赖图：
	//   git.Runner (cfg) ─┐
	//   crypto.Encryptor ─┼─ workspaceService (clones=nil) ─┐
	//   workspace.Repo ───┘                                 ├─ CloneRunner ─┐
	//                                                       │               │
	//                                       SetScheduler ───┴───────────────┘
	//                                                       │
	//                                       WorktreeService / GitOpsService
	// wsDataDir 是数据根；pathing.MainPath/WorktreePath 内部会再 join "workspaces" 子目录。
	// 这里显式 mkdir 出 ${dataDir}/workspaces 让首个 clone 的父目录已经存在。
	wsDataDir := cfg.Workspace.DataDir
	if err := os.MkdirAll(filepath.Join(wsDataDir, "workspaces"), 0o755); err != nil {
		return nil, fmt.Errorf("mkdir workspace data dir: %w", err)
	}
	tmpDir := filepath.Join(wsDataDir, "tmp")
	// 启动期清掉残留 tmp（可能含上次崩溃留下的 askpass 脚本与 SSH 私钥）
	_ = os.RemoveAll(tmpDir)
	if err := os.MkdirAll(tmpDir, 0o700); err != nil {
		return nil, fmt.Errorf("mkdir workspace tmp: %w", err)
	}

	enc, err := crypto.NewEncryptor(cfg.MasterKey)
	if err != nil {
		return nil, fmt.Errorf("crypto encryptor: %w", err)
	}

	gitr := git.NewDefaultRunner(git.Config{
		Binary:       cfg.Git.Binary,
		TmpDir:       tmpDir,
		CloneTimeout: cfg.Workspace.CloneTimeout,
		FetchTimeout: cfg.Workspace.FetchTimeout,
		PushTimeout:  cfg.Workspace.PushTimeout,
		OpsTimeout:   cfg.Workspace.OpsTimeout,
	})

	wsRepo := workspace.NewPostgresRepository(pool)
	// 进程启动时把卡死的 cloning/syncing 复位为 error，避免行永远卡住
	if err := wsRepo.ResetStuckSyncStatuses(ctx); err != nil {
		log.Warn("reset stuck workspace sync statuses", "err", err.Error())
	}

	pa := projectAccessAdapter{p: projectSvc}
	// 先以 nil scheduler 构造 service，cloneRunner 再回填（环依赖：runner 需要凭据访问器，访问器在 service 上）
	wsSvc := workspace.NewWorkspaceService(wsRepo, auditRec, enc, gitr, pa, nil, wsDataDir)

	// 共享凭据加载器：CloneRunner 与 GitOpsService 都需要。
	// 通过 CredentialAccessor 接口避免直接持有 *workspaceService 私有类型。
	credAccessor, ok := wsSvc.(workspace.CredentialAccessor)
	if !ok {
		return nil, fmt.Errorf("workspace service does not implement CredentialAccessor")
	}
	credLoader := func(ctx context.Context, wsID uuid.UUID) (*git.Credential, error) {
		return credAccessor.LoadDecryptedCredential(ctx, wsID)
	}

	cloneRunner := workspace.NewCloneRunner(wsRepo, auditRec, gitr, log, credLoader, cfg.Workspace.CloneTimeout)
	if setter, ok := wsSvc.(workspace.SchedulerSetter); ok {
		setter.SetScheduler(cloneRunner)
	}

	wtSvc := workspace.NewWorktreeService(wsRepo, auditRec, gitr, pa, wsDataDir)
	gopSvc := workspace.NewGitOpsService(wsRepo, auditRec, gitr, pa, credLoader)
	workspace.NewHandler(wsSvc, wtSvc, gopSvc, userSvc, userAdminAdapter{svc: userSvc}).Mount(r)

	if cfg.Bootstrap.AdminEmail != "" && cfg.Bootstrap.AdminPassword != "" {
		if u, err := userSvc.Bootstrap(ctx, cfg.Bootstrap.AdminEmail, cfg.Bootstrap.AdminPassword); err != nil {
			log.Warn("bootstrap admin failed", "err", err.Error())
		} else if u != nil {
			log.Info("bootstrap admin created", "email", u.Email)
		}
	}

	srv := &http.Server{
		Addr:              cfg.HTTP.Addr,
		Handler:           r,
		ReadHeaderTimeout: 10 * time.Second,
		ReadTimeout:       30 * time.Second,
		WriteTimeout:      60 * time.Second,
		IdleTimeout:       120 * time.Second,
	}
	return &App{cfg: cfg, log: log, pool: pool, server: srv, dispatcher: notifyDispatcher}, nil
}

// Run 启动 HTTP server 并阻塞直到 ctx 取消（Shutdown）或服务器报错。
// 关停路径上保证关闭 db pool；返回 nil 表示由 ctx 取消触发的正常关停。
func (a *App) Run(ctx context.Context) error {
	a.log.Info("server starting", "addr", a.cfg.HTTP.Addr, "env", a.cfg.Env)

	errCh := make(chan error, 1)
	go func() {
		if err := a.server.ListenAndServe(); err != nil && !errors.Is(err, http.ErrServerClosed) {
			errCh <- err
		}
	}()

	select {
	case <-ctx.Done():
		a.log.Info("shutdown signal received, draining...")
		shutCtx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
		defer cancel()
		shutErr := a.server.Shutdown(shutCtx)
		if shutErr != nil {
			a.log.Error("server shutdown", "err", shutErr.Error())
		}
		// server 已停止接收新请求；等异步通知投递结束再关 pool
		if err := a.dispatcher.Shutdown(shutCtx); err != nil {
			a.log.Warn("notify dispatcher drain", "err", err.Error())
		}
		a.pool.Close()
		return shutErr
	case err := <-errCh:
		drainCtx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
		defer cancel()
		_ = a.dispatcher.Shutdown(drainCtx)
		a.pool.Close()
		return err
	}
}

// userAdminAdapter 把 user.Service 适配为 project.AdminLookup（窄接口）。
// 复用 user.Service.Get：返回的 *User 直接读 IsAdmin。
// workspace.AdminLookup 与 project.AdminLookup 同形（IsAdmin 同名同签名），
// 因此本类型同时满足两者，可在两个 handler 复用。
type userAdminAdapter struct{ svc user.Service }

func (a userAdminAdapter) IsAdmin(ctx context.Context, id uuid.UUID) (bool, error) {
	u, err := a.svc.Get(ctx, id)
	if err != nil {
		return false, err
	}
	return u.IsAdmin, nil
}

// projectAccessAdapter 把 project.ProjectService.Get / GetByID 桥接成 workspace.ProjectAccess。
//   - Resolve     = by slug：用于 POST /projects/{slug}/workspaces 等路径
//   - ResolveByID = by uuid：用于 wsID 拿到 ws 后再校验 project 权限
//
// ACL 语义复刻 project 包内部 assertReadable / assertWritable：
//
//	read   private:  owner + admin
//	read   internal: 任意已登录
//	write  *:        owner + admin
//
// 注意：project.assertReadable 在 visibility=private 且非 owner+admin 时返回 NotFound
// 防探测；这里 ProjectService.Get/GetByID 已替我们兜住，到本适配层时 pj 一定可读。
type projectAccessAdapter struct {
	p project.ProjectService
}

func (a projectAccessAdapter) Resolve(ctx context.Context, callerID uuid.UUID, isAdmin bool, projectSlug string) (uuid.UUID, bool, bool, error) {
	pj, err := a.p.Get(ctx, project.Caller{UserID: callerID, IsAdmin: isAdmin}, projectSlug)
	if err != nil {
		return uuid.Nil, false, false, err
	}
	canRead, canWrite := projectACL(pj, callerID, isAdmin)
	return pj.ID, canRead, canWrite, nil
}

func (a projectAccessAdapter) ResolveByID(ctx context.Context, callerID uuid.UUID, isAdmin bool, projectID uuid.UUID) (bool, bool, error) {
	pj, err := a.p.GetByID(ctx, project.Caller{UserID: callerID, IsAdmin: isAdmin}, projectID)
	if err != nil {
		return false, false, err
	}
	canRead, canWrite := projectACL(pj, callerID, isAdmin)
	return canRead, canWrite, nil
}

// projectACL 复制 project 包内部 assertReadable/assertWritable 的语义：
//   - admin 永远可读可写
//   - owner 可读可写
//   - internal 可见性：所有登录用户可读
//
// 注：写操作还需考虑 archived 状态；workspace 的写场景会在 service 层用 ProjectAccess
// 拿到 canWrite 后再下沉，archived 检查由 project service 在 GetByID 已可见。
func projectACL(pj *project.Project, callerID uuid.UUID, isAdmin bool) (canRead, canWrite bool) {
	owned := pj.OwnerID == callerID
	canWrite = isAdmin || owned
	canRead = canWrite || pj.Visibility == project.VisibilityInternal
	return canRead, canWrite
}