feat(schedule): support KOL subscription targets and random time windows
Frontend CI / Frontend (push) Successful in 3m37s
Backend CI / Backend (push) Failing after 6m47s

Extend schedule tasks to dispatch KOL subscription prompts in addition
to prompt rules, add daily/weekly scheduling with fixed or random time
windows, and track dispatch outcomes (last run, status, consecutive
failures).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-05-22 23:51:45 +08:00
parent 9cc6537e68
commit c28c4f1419
20 changed files with 3034 additions and 329 deletions
@@ -3,6 +3,9 @@ package scheduler
import (
"context"
"encoding/json"
"errors"
"fmt"
"strings"
"sync"
"time"
@@ -22,12 +25,14 @@ const (
defaultScheduleDispatchTimeout = 30 * time.Second
defaultScheduleDispatchBatch = 20
scheduleCacheCleanupTimeout = 30 * time.Second
maxInvalidScheduleFailures = 3
)
type ScheduleDispatchWorker struct {
pool *pgxpool.Pool
rabbitMQ *rabbitmq.Client
promptRuleService *tenantapp.PromptRuleGenerationService
kolService *tenantapp.KolGenerationService
cache sharedcache.Cache
logger *zap.Logger
cfg scheduleDispatchRuntimeConfig
@@ -35,23 +40,30 @@ type ScheduleDispatchWorker struct {
}
type dueScheduleTask struct {
ID int64
TenantID int64
WorkspaceID int64
OperatorID *int64
PromptRuleID int64
BrandID *int64
Name string
CronExpr string
AutoPublish bool
PublishAccountIDs []string
CoverAssetURL *string
CoverImageAssetID *int64
EnableWebSearch bool
GenerateCount int
StartAt *time.Time
EndAt *time.Time
ScheduledFor time.Time
ID int64
TenantID int64
WorkspaceID int64
OperatorID *int64
TargetType string
PromptRuleID *int64
SubscriptionPromptID *int64
BrandID *int64
Name string
CronExpr string
ScheduleDays []string
ScheduleTimeMode string
RandomWindowStart string
RandomWindowEnd string
GenerationInput map[string]interface{}
AutoPublish bool
PublishAccountIDs []string
CoverAssetURL *string
CoverImageAssetID *int64
EnableWebSearch bool
GenerateCount int
StartAt *time.Time
EndAt *time.Time
ScheduledFor time.Time
}
type scheduleTaskCacheUpdate struct {
@@ -85,6 +97,11 @@ func NewScheduleDispatchWorker(
}
}
func (w *ScheduleDispatchWorker) WithKolGenerationService(svc *tenantapp.KolGenerationService) *ScheduleDispatchWorker {
w.kolService = svc
return w
}
func (w *ScheduleDispatchWorker) WithConfigProvider(provider config.Provider) *ScheduleDispatchWorker {
w.configProvider = provider
return w
@@ -237,7 +254,8 @@ func (w *ScheduleDispatchWorker) hydrateMissingNextRuns(ctx context.Context, run
defer tx.Rollback(ctx)
rows, err := tx.Query(ctx, `
SELECT id, tenant_id, cron_expr, start_at, end_at, status
SELECT id, tenant_id, cron_expr, schedule_days, schedule_time_mode,
random_window_start::text, random_window_end::text, start_at, end_at, status
FROM schedule_tasks
WHERE deleted_at IS NULL
AND status = 'enabled'
@@ -255,29 +273,35 @@ func (w *ScheduleDispatchWorker) hydrateMissingNextRuns(ctx context.Context, run
updates := make([]struct {
cacheUpdate scheduleTaskCacheUpdate
nextRunAt *time.Time
invalidErr error
}, 0, runtimeCfg.BatchSize)
for rows.Next() {
var (
id int64
tenantID int64
cronExpr string
status string
startAt pgtype.Timestamptz
endAt pgtype.Timestamptz
id int64
tenantID int64
cronExpr string
scheduleDaysJSON []byte
scheduleTimeMode string
randomWindowStart string
randomWindowEnd string
status string
startAt pgtype.Timestamptz
endAt pgtype.Timestamptz
)
if err := rows.Scan(&id, &tenantID, &cronExpr, &startAt, &endAt, &status); err != nil {
if err := rows.Scan(&id, &tenantID, &cronExpr, &scheduleDaysJSON, &scheduleTimeMode, &randomWindowStart, &randomWindowEnd, &startAt, &endAt, &status); err != nil {
rows.Close()
return 0, err
}
var nextRunAt *time.Time
var nextErr error
if status == "enabled" {
nextRunAt, err = sharedschedule.ComputeNextRun(cronExpr, timePtrFromTimestamp(startAt), timePtrFromTimestamp(endAt), now)
if err != nil && w.logger != nil {
w.logger.Warn("schedule dispatch skipped invalid cron while hydrating next_run_at",
nextRunAt, nextErr = tenantapp.ComputeScheduleTaskNextRun(status, id, cronExpr, tenantapp.DecodeScheduleDayKeys(scheduleDaysJSON), scheduleTimeMode, randomWindowStart, randomWindowEnd, timePtrFromTimestamp(startAt), timePtrFromTimestamp(endAt), now)
if nextErr != nil && w.logger != nil {
w.logger.Warn("schedule dispatch skipped invalid schedule while hydrating next_run_at",
zap.Int64("schedule_task_id", id),
zap.String("cron_expr", cronExpr),
zap.Error(err),
zap.Error(nextErr),
)
}
}
@@ -285,12 +309,14 @@ func (w *ScheduleDispatchWorker) hydrateMissingNextRuns(ctx context.Context, run
updates = append(updates, struct {
cacheUpdate scheduleTaskCacheUpdate
nextRunAt *time.Time
invalidErr error
}{
cacheUpdate: scheduleTaskCacheUpdate{
id: id,
tenantID: tenantID,
},
nextRunAt: nextRunAt,
nextRunAt: nextRunAt,
invalidErr: nextErr,
})
}
if err := rows.Err(); err != nil {
@@ -300,6 +326,23 @@ func (w *ScheduleDispatchWorker) hydrateMissingNextRuns(ctx context.Context, run
rows.Close()
for _, update := range updates {
if update.invalidErr != nil {
message := truncateDispatchError(update.invalidErr)
if _, err := tx.Exec(ctx, `
UPDATE schedule_tasks
SET next_run_at = NULL,
last_dispatch_status = 'failed',
last_dispatch_error = NULLIF($1, ''),
consecutive_failures = consecutive_failures + 1,
status = CASE WHEN consecutive_failures + 1 >= $2 THEN 'disabled' ELSE status END,
updated_at = NOW()
WHERE id = $3
`, message, maxInvalidScheduleFailures, update.cacheUpdate.id); err != nil {
return 0, err
}
count++
continue
}
if _, err := tx.Exec(ctx, `
UPDATE schedule_tasks
SET next_run_at = $1,
@@ -330,7 +373,9 @@ func (w *ScheduleDispatchWorker) claimDueSchedules(ctx context.Context, runtimeC
defer tx.Rollback(ctx)
rows, err := tx.Query(ctx, `
SELECT id, tenant_id, workspace_id, operator_id, prompt_rule_id, brand_id, name, cron_expr,
SELECT id, tenant_id, workspace_id, operator_id, target_type, prompt_rule_id,
subscription_prompt_id, brand_id, name, cron_expr, schedule_days, schedule_time_mode,
random_window_start::text, random_window_end::text, generation_input_json,
auto_publish, publish_account_ids, cover_asset_url, cover_image_asset_id,
enable_web_search, generate_count, start_at, end_at, next_run_at
FROM schedule_tasks
@@ -351,12 +396,17 @@ func (w *ScheduleDispatchWorker) claimDueSchedules(ctx context.Context, runtimeC
updates := make([]struct {
cacheUpdate scheduleTaskCacheUpdate
nextRun *time.Time
invalidErr error
}, 0, runtimeCfg.BatchSize)
for rows.Next() {
var (
task dueScheduleTask
workspaceID pgtype.Int8
operatorID pgtype.Int8
promptRuleID pgtype.Int8
subscriptionPromptID pgtype.Int8
generationInputJSON []byte
scheduleDaysJSON []byte
publishAccountIDsJSON []byte
startAt pgtype.Timestamptz
endAt pgtype.Timestamptz
@@ -367,10 +417,17 @@ func (w *ScheduleDispatchWorker) claimDueSchedules(ctx context.Context, runtimeC
&task.TenantID,
&workspaceID,
&operatorID,
&task.PromptRuleID,
&task.TargetType,
&promptRuleID,
&subscriptionPromptID,
&task.BrandID,
&task.Name,
&task.CronExpr,
&scheduleDaysJSON,
&task.ScheduleTimeMode,
&task.RandomWindowStart,
&task.RandomWindowEnd,
&generationInputJSON,
&task.AutoPublish,
&publishAccountIDsJSON,
&task.CoverAssetURL,
@@ -391,6 +448,13 @@ func (w *ScheduleDispatchWorker) claimDueSchedules(ctx context.Context, runtimeC
task.WorkspaceID = workspaceID.Int64
}
task.OperatorID = int64PtrFromInt8(operatorID)
task.PromptRuleID = int64PtrFromInt8(promptRuleID)
task.SubscriptionPromptID = int64PtrFromInt8(subscriptionPromptID)
task.ScheduleDays = tenantapp.DecodeScheduleDayKeys(scheduleDaysJSON)
task.GenerationInput = decodeDueScheduleGenerationInput(generationInputJSON)
if task.TargetType == "" {
task.TargetType = tenantapp.ScheduleTargetPromptRule
}
task.PublishAccountIDs = decodeDueSchedulePublishAccountIDs(publishAccountIDsJSON)
if task.GenerateCount <= 0 {
task.GenerateCount = 1
@@ -401,28 +465,29 @@ func (w *ScheduleDispatchWorker) claimDueSchedules(ctx context.Context, runtimeC
}
task.ScheduledFor = *scheduledFor
nextRun, nextErr := sharedschedule.ComputeNextRun(task.CronExpr, task.StartAt, task.EndAt, now.Add(time.Nanosecond))
nextRun, nextErr := tenantapp.ComputeScheduleTaskNextRun("enabled", task.ID, task.CronExpr, task.ScheduleDays, task.ScheduleTimeMode, task.RandomWindowStart, task.RandomWindowEnd, task.StartAt, task.EndAt, now.Add(time.Nanosecond))
if nextErr != nil && w.logger != nil {
w.logger.Warn("schedule dispatch encountered invalid cron while claiming due schedule",
w.logger.Warn("schedule dispatch encountered invalid schedule while claiming due schedule",
zap.Int64("schedule_task_id", task.ID),
zap.String("cron_expr", task.CronExpr),
zap.Error(nextErr),
)
nextRun = nil
}
updates = append(updates, struct {
cacheUpdate scheduleTaskCacheUpdate
nextRun *time.Time
invalidErr error
}{
cacheUpdate: scheduleTaskCacheUpdate{
id: task.ID,
tenantID: task.TenantID,
},
nextRun: nextRun,
nextRun: nextRun,
invalidErr: nextErr,
})
if sharedschedule.IsRunAllowed(task.ScheduledFor, task.StartAt, task.EndAt) {
if nextErr == nil && sharedschedule.IsRunAllowed(task.ScheduledFor, task.StartAt, task.EndAt) {
tasks = append(tasks, task)
}
}
@@ -433,6 +498,22 @@ func (w *ScheduleDispatchWorker) claimDueSchedules(ctx context.Context, runtimeC
rows.Close()
for _, update := range updates {
if update.invalidErr != nil {
message := truncateDispatchError(update.invalidErr)
if _, err := tx.Exec(ctx, `
UPDATE schedule_tasks
SET next_run_at = NULL,
last_dispatch_status = 'failed',
last_dispatch_error = NULLIF($1, ''),
consecutive_failures = consecutive_failures + 1,
status = CASE WHEN consecutive_failures + 1 >= $2 THEN 'disabled' ELSE status END,
updated_at = NOW()
WHERE id = $3
`, message, maxInvalidScheduleFailures, update.cacheUpdate.id); err != nil {
return nil, err
}
continue
}
if _, err := tx.Exec(ctx, `
UPDATE schedule_tasks
SET next_run_at = $1,
@@ -460,31 +541,20 @@ func (w *ScheduleDispatchWorker) dispatch(parent context.Context, task dueSchedu
generateCount = 1
}
failures := make([]error, 0)
for idx := 0; idx < generateCount; idx++ {
ctx, cancel := w.stageContext(parent)
resp, err := w.promptRuleService.EnqueueScheduledGeneration(ctx, tenantapp.SchedulePromptRuleGenerationInput{
ScheduleTaskID: task.ID,
OperatorID: task.OperatorID,
TenantID: task.TenantID,
PromptRuleID: task.PromptRuleID,
BrandID: task.BrandID,
Name: task.Name,
WorkspaceID: task.WorkspaceID,
AutoPublish: task.AutoPublish,
PublishAccountIDs: task.PublishAccountIDs,
CoverAssetURL: task.CoverAssetURL,
CoverImageAssetID: task.CoverImageAssetID,
EnableWebSearch: task.EnableWebSearch,
GenerateCount: generateCount,
ScheduledFor: task.ScheduledFor,
})
resp, err := w.enqueueDueGeneration(ctx, task, generateCount)
cancel()
if err != nil {
failures = append(failures, err)
if w.logger != nil {
w.logger.Warn("schedule dispatch enqueue generation failed",
zap.Int64("schedule_task_id", task.ID),
zap.Int64("tenant_id", task.TenantID),
zap.Int64("prompt_rule_id", task.PromptRuleID),
zap.String("target_type", task.TargetType),
zap.Int64("prompt_rule_id", derefInt64(task.PromptRuleID)),
zap.Int64("subscription_prompt_id", derefInt64(task.SubscriptionPromptID)),
zap.Int("generate_index", idx+1),
zap.Int("generate_count", generateCount),
zap.Time("scheduled_for", task.ScheduledFor.UTC()),
@@ -498,7 +568,9 @@ func (w *ScheduleDispatchWorker) dispatch(parent context.Context, task dueSchedu
w.logger.Info("schedule dispatch enqueued generation",
zap.Int64("schedule_task_id", task.ID),
zap.Int64("tenant_id", task.TenantID),
zap.Int64("prompt_rule_id", task.PromptRuleID),
zap.String("target_type", task.TargetType),
zap.Int64("prompt_rule_id", derefInt64(task.PromptRuleID)),
zap.Int64("subscription_prompt_id", derefInt64(task.SubscriptionPromptID)),
zap.Int64("article_id", resp.ArticleID),
zap.Int64("generation_task_id", resp.TaskID),
zap.Int("generate_index", idx+1),
@@ -507,6 +579,141 @@ func (w *ScheduleDispatchWorker) dispatch(parent context.Context, task dueSchedu
)
}
}
w.recordDispatchResult(context.Background(), task, generateCount, failures)
}
func (w *ScheduleDispatchWorker) enqueueDueGeneration(ctx context.Context, task dueScheduleTask, generateCount int) (*tenantapp.GenerateFromRuleResponse, error) {
switch task.TargetType {
case "", tenantapp.ScheduleTargetPromptRule:
if task.PromptRuleID == nil || *task.PromptRuleID <= 0 {
return nil, errors.New("schedule prompt_rule target missing prompt_rule_id")
}
return w.promptRuleService.EnqueueScheduledGeneration(ctx, tenantapp.SchedulePromptRuleGenerationInput{
ScheduleTaskID: task.ID,
OperatorID: task.OperatorID,
TenantID: task.TenantID,
PromptRuleID: *task.PromptRuleID,
BrandID: task.BrandID,
Name: task.Name,
WorkspaceID: task.WorkspaceID,
AutoPublish: task.AutoPublish,
PublishAccountIDs: task.PublishAccountIDs,
CoverAssetURL: task.CoverAssetURL,
CoverImageAssetID: task.CoverImageAssetID,
EnableWebSearch: task.EnableWebSearch,
GenerateCount: generateCount,
ScheduledFor: task.ScheduledFor,
})
case tenantapp.ScheduleTargetKolSubscriptionPrompt:
if w.kolService == nil {
return nil, errors.New("kol generation service is not configured")
}
if task.SubscriptionPromptID == nil || *task.SubscriptionPromptID <= 0 {
return nil, errors.New("schedule kol target missing subscription_prompt_id")
}
variables, _ := task.GenerationInput["variables"].(map[string]interface{})
kolResp, err := w.kolService.EnqueueScheduledGeneration(ctx, tenantapp.ScheduleKolGenerationInput{
ScheduleTaskID: task.ID,
OperatorID: task.OperatorID,
TenantID: task.TenantID,
WorkspaceID: task.WorkspaceID,
BrandID: task.BrandID,
SubscriptionPromptID: *task.SubscriptionPromptID,
Name: task.Name,
Variables: variables,
EnableWebSearch: scheduleGenerationInputBool(task.GenerationInput, "enable_web_search", task.EnableWebSearch),
KnowledgeGroupIDs: scheduleGenerationInputInt64s(task.GenerationInput, "knowledge_group_ids"),
AutoPublish: task.AutoPublish,
PublishAccountIDs: task.PublishAccountIDs,
CoverAssetURL: task.CoverAssetURL,
CoverImageAssetID: task.CoverImageAssetID,
GenerateCount: generateCount,
ScheduledFor: task.ScheduledFor,
})
if err != nil {
return nil, err
}
return &tenantapp.GenerateFromRuleResponse{
ArticleID: kolResp.ArticleID,
TaskID: kolResp.GenerationTaskID,
}, nil
default:
return nil, fmt.Errorf("unsupported schedule target type: %s", task.TargetType)
}
}
func (w *ScheduleDispatchWorker) recordDispatchResult(ctx context.Context, task dueScheduleTask, generateCount int, failures []error) {
if w == nil || w.pool == nil {
return
}
failedCount := len(failures)
result := aggregateDispatchResult(generateCount, failedCount)
if result.status == "success" {
_, _ = w.pool.Exec(ctx, `
UPDATE schedule_tasks
SET last_run_at = $1,
last_dispatch_status = $2,
last_dispatch_error = NULL,
consecutive_failures = 0,
updated_at = NOW()
WHERE id = $3 AND tenant_id = $4
`, task.ScheduledFor, result.status, task.ID, task.TenantID)
return
}
message := truncateDispatchErrors(failures)
_, _ = w.pool.Exec(ctx, `
UPDATE schedule_tasks
SET last_run_at = $1,
last_dispatch_status = $2,
last_dispatch_error = NULLIF($3, ''),
consecutive_failures = CASE WHEN $4 THEN consecutive_failures + 1 ELSE 0 END,
updated_at = NOW()
WHERE id = $5 AND tenant_id = $6
`, task.ScheduledFor, result.status, message, result.countAsFailure, task.ID, task.TenantID)
}
type dispatchAggregateResult struct {
status string
countAsFailure bool
}
func aggregateDispatchResult(generateCount, failedCount int) dispatchAggregateResult {
if generateCount <= 0 {
generateCount = 1
}
if failedCount <= 0 {
return dispatchAggregateResult{status: "success"}
}
if failedCount >= generateCount {
return dispatchAggregateResult{status: "failed", countAsFailure: true}
}
return dispatchAggregateResult{status: "partial"}
}
func truncateDispatchError(err error) string {
if err == nil {
return ""
}
message := err.Error()
if len(message) > 1000 {
return message[:1000]
}
return message
}
func truncateDispatchErrors(errs []error) string {
parts := make([]string, 0, len(errs))
for _, err := range errs {
if err == nil {
continue
}
parts = append(parts, err.Error())
}
message := strings.Join(parts, "; ")
if len(message) > 1000 {
return message[:1000]
}
return message
}
func (w *ScheduleDispatchWorker) dispatchBatch(parent context.Context, tasks []dueScheduleTask) {
@@ -662,6 +869,24 @@ func int64PtrFromInt8(value pgtype.Int8) *int64 {
return &result
}
func derefInt64(value *int64) int64 {
if value == nil {
return 0
}
return *value
}
func decodeDueScheduleGenerationInput(raw []byte) map[string]interface{} {
if len(raw) == 0 {
return map[string]interface{}{}
}
var payload map[string]interface{}
if err := json.Unmarshal(raw, &payload); err != nil || payload == nil {
return map[string]interface{}{}
}
return payload
}
func decodeDueSchedulePublishAccountIDs(raw []byte) []string {
if len(raw) == 0 {
return []string{}
@@ -684,3 +909,65 @@ func decodeDueSchedulePublishAccountIDs(raw []byte) []string {
}
return result
}
func scheduleGenerationInputBool(input map[string]interface{}, key string, fallback bool) bool {
if len(input) == 0 {
return fallback
}
raw, ok := input[key]
if !ok || raw == nil {
return fallback
}
switch typed := raw.(type) {
case bool:
return typed
case string:
return typed == "true" || typed == "1" || typed == "yes"
case float64:
return typed != 0
default:
return fallback
}
}
func scheduleGenerationInputInt64s(input map[string]interface{}, key string) []int64 {
if len(input) == 0 {
return nil
}
raw, ok := input[key]
if !ok || raw == nil {
return nil
}
var ids []int64
switch typed := raw.(type) {
case []int64:
ids = typed
case []interface{}:
ids = make([]int64, 0, len(typed))
for _, item := range typed {
switch value := item.(type) {
case int64:
ids = append(ids, value)
case int:
ids = append(ids, int64(value))
case float64:
ids = append(ids, int64(value))
}
}
default:
return nil
}
seen := make(map[int64]struct{}, len(ids))
result := make([]int64, 0, len(ids))
for _, id := range ids {
if id <= 0 {
continue
}
if _, ok := seen[id]; ok {
continue
}
seen[id] = struct{}{}
result = append(result, id)
}
return result
}
@@ -0,0 +1,30 @@
package scheduler
import "testing"
func TestAggregateDispatchResult(t *testing.T) {
tests := []struct {
name string
generateCount int
failedCount int
wantStatus string
wantFailureCount bool
}{
{name: "all success", generateCount: 3, failedCount: 0, wantStatus: "success"},
{name: "partial success", generateCount: 3, failedCount: 1, wantStatus: "partial"},
{name: "all failed", generateCount: 3, failedCount: 3, wantStatus: "failed", wantFailureCount: true},
{name: "zero generate count defaults to one", generateCount: 0, failedCount: 1, wantStatus: "failed", wantFailureCount: true},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
got := aggregateDispatchResult(tt.generateCount, tt.failedCount)
if got.status != tt.wantStatus {
t.Fatalf("status = %q, want %q", got.status, tt.wantStatus)
}
if got.countAsFailure != tt.wantFailureCount {
t.Fatalf("countAsFailure = %v, want %v", got.countAsFailure, tt.wantFailureCount)
}
})
}
}