story/internal/callmemory/volume.go

package callmemory

import (
	"context"
	"encoding/json"
	"errors"
	"regexp"
	"strings"
	"unicode"

	"data-terminal-cloud/internal/volcsubtitle"
	"github.com/google/uuid"
	"github.com/jackc/pgx/v5"
)

// Whole, short requests only. Substring matching would execute quoted commands,
// negations, and unrelated stories. No extra model or transcript forwarding.
var spokenVolume = regexp.MustCompile(`^(?:(?:请|麻烦|帮我|给我|能不能|可以|你|您|把|再|稍微))*((?:声音|音量)?(?:调|调得|开|开得|放)?(?:大|高|响|小|低|轻)(?:一?点儿?)(?:声|声音|音量)?|(?:调|开)(?:大|高|小|低)(?:声音|音量)|(?:声音|音量)(?:调|开)(?:大|高|小|低)|(?:声音|音量)(?:有点|太)(?:小|低|大|高))(?:一点)?(?:好吗|好不好|可以吗|行吗|吧|呀|啊|呢|吗|了)?$`)

func spokenVolumeAction(text string) string {
	text = strings.Map(func(r rune) rune {
		if unicode.IsSpace(r) {
			return -1
		}
		return r
	}, text)
	text = strings.Trim(text, "。！？，,.!?")
	if len([]rune(text)) > 40 || !spokenVolume.MatchString(text) {
		return ""
	}
	if !strings.Contains(text, "声") && !strings.Contains(text, "音量") {
		return ""
	}
	if strings.Contains(text, "有点小") || strings.Contains(text, "太小") || strings.Contains(text, "有点低") || strings.Contains(text, "太低") {
		return "up"
	}
	if strings.Contains(text, "有点大") || strings.Contains(text, "太大") || strings.Contains(text, "有点高") || strings.Contains(text, "太高") {
		return "down"
	}
	if strings.ContainsAny(text, "大高响") {
		return "up"
	}
	return "down"
}

// Called inside the authenticated subtitle transaction while its call row is
// locked. A callback and its volume change commit together or retry together.
func (s *Service) applySpokenVolume(ctx context.Context, tx pgx.Tx, sid string, part volcsubtitle.Segment) error {
	action := spokenVolumeAction(part.Text)
	if action == "" || !part.Definite || part.RoundID == nil {
		return nil
	}
	var device, tenant string
	err := tx.QueryRow(ctx, `SELECT d.id::text,d.tenant_id::text FROM call_session c JOIN device d ON d.id=c.device_id
 JOIN device_binding b ON b.device_id=d.id AND b.generation=d.binding_generation AND b.subject_id=c.subject_id AND b.unbound_at IS NULL
 WHERE c.id=$1::uuid AND c.binding_generation=d.binding_generation AND c.budget_snapshot->>'protocol'='DIRECT-CHAT-1'
 AND c.call_state IN ('STARTING','TALKING') AND c.ended_at IS NULL AND c.expires_at>now() AND c.lease_expires_at>now()
 AND NOT c.local_audio_off AND NOT c.sdk_stopped`, sid).Scan(&device, &tenant)
	if errors.Is(err, pgx.ErrNoRows) {
		return nil
	}
	if err != nil {
		return err
	}
	// One operation per user's speech round, even if final subtitle segments or
	// provider retries use different sequence numbers.
	key := uuid.NewSHA1(uuid.NameSpaceOID, jsonBytes([]any{sid, part.UserID, *part.RoundID})).String()
	actor := "spoken-volume:" + device
	var exists bool
	if err = tx.QueryRow(ctx, `SELECT EXISTS(SELECT 1 FROM call_request_result WHERE actor=$1 AND request_id=$2::uuid)`, actor, key).Scan(&exists); err != nil {
		return err
	}
	if exists {
		return nil
	}
	var desired []byte
	var revision int64
	if err = tx.QueryRow(ctx, `SELECT desired_config,desired_revision FROM device_twin WHERE device_id=$1::uuid AND tenant_id=$2::uuid FOR UPDATE`, device, tenant).Scan(&desired, &revision); err != nil {
		return err
	}
	var config map[string]any
	if json.Unmarshal(desired, &config) != nil {
		return errors.New("invalid device volume configuration")
	}
	volume, _ := config["volume"].(map[string]any)
	number, ok := volume["speaker"].(float64)
	// Older devices without a volume setting are not silently given a default.
	if !ok || number < 0 || number > 100 || number != float64(int(number)) {
		return nil
	}
	target := int(number)
	if action == "up" {
		target += 10
	} else {
		target -= 10
	}
	if target < 0 {
		target = 0
	}
	if target > 100 {
		target = 100
	}
	if target != int(number) {
		config["volume"] = map[string]any{"speaker": target}
		revision++
		if _, err = tx.Exec(ctx, `UPDATE device_twin SET desired_config=$3::jsonb,desired_revision=$4,updated_at=now() WHERE device_id=$1::uuid AND tenant_id=$2::uuid`, device, tenant, jsonBytes(config), revision); err != nil {
			return err
		}
	}
	result := map[string]any{"session_id": sid, "action": action, "speaker": target, "revision": revision, "status": "requested"}
	_, err = tx.Exec(ctx, `INSERT INTO call_request_result(actor,request_id,path,body_digest,response) VALUES($1,$2::uuid,'spoken-volume',$3,$4::jsonb)`, actor, key, hash(jsonBytes([]any{sid, key, action})), jsonBytes(result))
	return err
}