talk/internal/daemon/daemon.go

// Package daemon is talk's background process: HTTP API on 127.0.0.1, a
// serialized announcement queue with priorities and the now jump,
// persistent mute state, the anti-spam laws (rate limit, quiet hours), the
// CORS door and the MCP endpoint. There is exactly one binary: `talk daemon
// run` runs the daemon in the foreground, and the CLI starts it in the
// background itself via Spawn when needed (daemon.autostart).
package daemon

import (
	"context"
	"fmt"
	"io"
	"log"
	"net"
	"net/http"
	"os"
	"path/filepath"
	"strconv"
	"strings"
	"sync"
	"time"

	"code.m3ta.dev/m3tam3re/talk/internal/api"
	"code.m3ta.dev/m3tam3re/talk/internal/config"
	"code.m3ta.dev/m3tam3re/talk/internal/notify"
	"code.m3ta.dev/m3tam3re/talk/internal/provider"
	"code.m3ta.dev/m3tam3re/talk/internal/queue"
	"code.m3ta.dev/m3tam3re/talk/internal/speak"
)

// Files of the daemon in the config directory.
const (
	// LogFileName is the log of a daemon started in the background (Spawn
	// redirects stdout/stderr there; fresh per start).
	LogFileName = "daemon.log"

	// PIDFileName holds the process ID of the running daemon (for
	// diagnostics and as an emergency brake when the HTTP API does not
	// answer).
	PIDFileName = "daemon.pid"
)

// Options are the dependencies of Run.
type Options struct {
	// Config is the loaded configuration (port, laws, sink, ...).
	Config config.Config

	// Getenv supplies keys and endpoint overrides (nil = os.Getenv).
	Getenv func(string) string

	// Version is the daemon version for /status and MCP serverInfo.
	Version string

	// Log receives the log output (foreground: stderr; background: the log
	// file). nil = os.Stderr.
	Log io.Writer
}

// Run runs the daemon until ctx ends (signal), POST /shutdown arrives or
// the HTTP server fails. A playing announcement is cancelled on shutdown;
// the daemon should react promptly. A port in use is a clear error ("is
// another daemon running?").
func Run(ctx context.Context, opts Options) error {
	getenv := opts.Getenv
	if getenv == nil {
		getenv = os.Getenv
	}
	out := opts.Log
	if out == nil {
		out = os.Stderr
	}
	logger := log.New(out, "talk daemon: ", log.LstdFlags)
	lw := lineWriter{logger}
	cfg := opts.Config

	engine, err := speak.New(cfg, getenv, lw)
	if err != nil {
		return err
	}

	// Anti-spam law: config.Load has validated quiet_hours; an error here
	// is practically impossible but must never be silent.
	quiet, err := queue.ParseQuietHours(cfg.QuietHours)
	if err != nil {
		return err
	}
	// Desktop side: the icon next to the config, Windows additionally the
	// toast identity and the talk: protocol handler for the Cancel button.
	if dir, err := config.Dir(getenv); err == nil {
		exe, _ := os.Executable()
		if err := notify.Setup(dir, exe); err != nil {
			logger.Printf("warning: %v; notifications fall back to the plain identity or lose the Cancel button", err)
		}
	}

	// Desktop notification per announcement (notify.on_speak): shown when the
	// announcement starts, with a Cancel button where the platform has one,
	// closed when it ends. Best effort: the first failure is logged, later
	// ones are silent, speech never waits for a notification problem.
	var q *queue.Queue
	var notifyFailed sync.Once
	announceNotification := func(ctx context.Context, item queue.Item, lang string) func() {
		if !cfg.Notify.OnSpeak {
			return func() {}
		}
		label := "Cancel"
		if lang == "de" {
			label = "Abbrechen"
		}
		closeFn, err := notify.Announce(ctx, notify.Announcement{
			Text:        item.Text,
			CancelLabel: label,
			CancelURI:   fmt.Sprintf("talk:cancel?id=%s&port=%d", item.ID, cfg.Server.Port),
			Tag:         item.ID,
		}, func() {
			if q.CancelItem(item.ID) {
				logger.Printf("announcement %s cancelled from its notification", item.ID)
			}
		})
		if err != nil {
			notifyFailed.Do(func() {
				logger.Printf("warning: notifications unavailable (%v); announcements play without one. Set [notify] on_speak = false to silence this", err)
			})
		}
		return closeFn
	}

	q = queue.New(queue.Laws{
		RateLimit:      cfg.Queue.RateLimit,
		AllowHighBurst: cfg.Queue.AllowHighBurst,
		Quiet:          quiet,
		QuietMode:      cfg.QuietMode,
		MaxText:        cfg.Queue.MaxText,
		MaxAge:         cfg.Queue.MaxAge,
	}, func(ctx context.Context, item queue.Item) error {
		req := speak.Request{Text: item.Text, Lang: item.Lang, Provider: item.Provider, Voice: item.Voice}
		closeNotification := announceNotification(ctx, item, engine.Lang(req))
		defer closeNotification()
		out, err := engine.Speak(ctx, req)
		if err == nil && !out.Discarded {
			logger.Printf("announcement %s via %s (synthesis %s, playback %s)", item.ID, out.Provider,
				out.Synthesis.Round(time.Millisecond), out.Playback.Round(time.Millisecond))
		}
		return err
	}, lw)

	// Quiet-hours toast: desktop notification instead of audio (quiet_mode
	// = "toast"); best effort, errors only in the log. Runs asynchronously
	// so a hanging notification daemon never blocks the queue.
	if cfg.QuietMode == queue.QuietModeToast {
		if notify.Available() {
			q.OnToast(func(item queue.Item) {
				go func() {
					if err := notify.Show(context.Background(), item.Text); err != nil {
						logger.Printf("toast: %v", err)
					}
				}()
			})
		} else {
			logger.Printf("warning: quiet_mode = toast, but %s; quiet-hours announcements are only logged", notify.Describe())
		}
	}

	// Persistent mute state and PID file in the config dir. Without a
	// config dir the daemon keeps running without both (warning) instead
	// of failing.
	var mutePath, pidPath string
	if dir, err := config.Dir(getenv); err == nil {
		mutePath = filepath.Join(dir, queue.MuteFileName)
		pidPath = filepath.Join(dir, PIDFileName)
	} else {
		logger.Printf("warning: config directory unknown; mute state and PID file disabled (%v)", err)
	}
	mute := queue.MuteStore{Path: mutePath}

	shutdown := make(chan struct{})
	var once sync.Once
	handler := api.NewHandler(api.HandlerConfig{
		Queue:          q,
		Mute:           mute,
		Version:        opts.Version,
		KnownProviders: provider.Names(),
		AllowedOrigins: cfg.Server.AllowedOrigins,
		Provider: func() string {
			if cur, ok := q.Current(); ok {
				return engine.ActiveProvider(cur.Provider)
			}
			if last := engine.LastProvider(); last != "" {
				return last
			}
			return engine.ActiveProvider("")
		},
		Shutdown: func() { once.Do(func() { close(shutdown) }) },
	})

	addr := fmt.Sprintf("127.0.0.1:%d", cfg.Server.Port)
	ln, err := net.Listen("tcp", addr)
	if err != nil {
		q.Close()
		return fmt.Errorf("bind %s: %w; is another talk daemon running?", addr, err)
	}
	if pidPath != "" {
		writePID(pidPath, logger)
		// The PID file disappears last: `talk daemon stop` waits for that
		// and then knows the process is done, not just HTTP.
		defer os.Remove(pidPath)
	}

	srv := &http.Server{Handler: handler, ReadHeaderTimeout: 5 * time.Second}
	errCh := make(chan error, 1)
	go func() { errCh <- srv.Serve(ln) }()

	corsLog := "closed (no CORS)"
	if len(cfg.Server.AllowedOrigins) > 0 {
		corsLog = fmt.Sprintf("%v", cfg.Server.AllowedOrigins)
	}
	logger.Printf("listening on http://%s (version %s, pid %d, audio sink %s, mute=%v, notify=%v, rate-limit %s, high-burst=%v, max-text %d, max-age %s, quiet-hours=%q mode %s, cors=%s, mcp=/mcp)",
		addr, opts.Version, os.Getpid(), engine.SinkName(), mute.Muted(), cfg.Notify.OnSpeak,
		cfg.Queue.RateLimit, cfg.Queue.AllowHighBurst, cfg.Queue.MaxText, cfg.Queue.MaxAge, cfg.QuietHours, cfg.QuietMode, corsLog)

	select {
	case err := <-errCh:
		q.Close()
		return fmt.Errorf("http server: %w", err)
	case <-ctx.Done():
		logger.Printf("signal received; shutting down (the playing announcement is cancelled)")
	case <-shutdown:
		logger.Printf("shutdown requested (talk daemon stop); shutting down")
	}

	// Order: close HTTP first (no new requests), then close the queue
	// (cancel the playing announcement, stop the worker and wait for it).
	// That way no announcement ever keeps playing after the daemon ends.
	shutdownCtx, cancel := context.WithTimeout(context.Background(), 3*time.Second)
	defer cancel()
	if err := srv.Shutdown(shutdownCtx); err != nil {
		logger.Printf("http shutdown: %v", err)
	}
	q.Close()
	return nil
}

// writePID writes our own process ID (best effort; an error is only a log
// line, the daemon works without a PID file too).
func writePID(path string, logger *log.Logger) {
	if err := os.MkdirAll(filepath.Dir(path), 0o755); err != nil {
		logger.Printf("warning: PID file: %v", err)
		return
	}
	if err := os.WriteFile(path, []byte(strconv.Itoa(os.Getpid())+"\n"), 0o644); err != nil {
		logger.Printf("warning: PID file: %v", err)
	}
}

// ReadPID reads the PID file; 0 if no (readable) file exists.
func ReadPID(dir string) int {
	b, err := os.ReadFile(filepath.Join(dir, PIDFileName))
	if err != nil {
		return 0
	}
	pid, err := strconv.Atoi(strings.TrimSpace(string(b)))
	if err != nil || pid <= 0 {
		return 0
	}
	return pid
}

// lineWriter forwards lines to the logger (timestamp + prefix) instead of
// writing them raw: provider chain, queue and null sink write line by line,
// which yields one uniform log.
type lineWriter struct{ l *log.Logger }

func (w lineWriter) Write(p []byte) (int, error) {
	for _, line := range strings.Split(strings.TrimRight(string(p), "\n"), "\n") {
		if line != "" {
			w.l.Print(line)
		}
	}
	return len(p), nil
}