feat: add river-backed async task queue

apps/api/internal/runner/limits.go

@@ -2,13 +2,49 @@ package runner
 
 import (
 	"context"
+	"errors"
 	"strings"
+	"time"
 
 	"github.com/easyai/easyai-ai-gateway/apps/api/internal/auth"
 	"github.com/easyai/easyai-ai-gateway/apps/api/internal/clients"
 	"github.com/easyai/easyai-ai-gateway/apps/api/internal/store"
 )
 
+type localRateLimitError struct {
+	clientErr  *clients.ClientError
+	cause      error
+	retryAfter time.Duration
+}
+
+func (e *localRateLimitError) Error() string {
+	if e == nil || e.clientErr == nil {
+		return store.ErrRateLimited.Error()
+	}
+	return e.clientErr.Error()
+}
+
+func (e *localRateLimitError) Unwrap() []error {
+	if e == nil || e.clientErr == nil {
+		if e != nil && e.cause != nil {
+			return []error{e.cause}
+		}
+		return []error{store.ErrRateLimited}
+	}
+	if e.cause != nil {
+		return []error{e.clientErr, e.cause}
+	}
+	return []error{e.clientErr, store.ErrRateLimited}
+}
+
+func localRateLimitRetryAfter(err error) time.Duration {
+	var limitErr *localRateLimitError
+	if errors.As(err, &limitErr) && limitErr.retryAfter > 0 {
+		return limitErr.retryAfter
+	}
+	return store.RateLimitRetryAfter(err)
+}
+
 func (s *Service) rateLimitReservations(ctx context.Context, user *auth.User, candidate store.RuntimeModelCandidate, body map[string]any) []store.RateLimitReservation {
 	out := make([]store.RateLimitReservation, 0)
 	out = append(out, reservationsFromPolicy("platform_model", candidate.PlatformModelID, effectiveRateLimitPolicy(candidate), body)...)