fix gateway loopback validation chains
This commit is contained in:
@@ -50,8 +50,8 @@ func (s *Service) ExecuteStream(ctx context.Context, task store.GatewayTask, use
|
||||
}
|
||||
|
||||
func (s *Service) execute(ctx context.Context, task store.GatewayTask, user *auth.User, onDelta clients.StreamDelta) (Result, error) {
|
||||
modelType := modelTypeFromKind(task.Kind)
|
||||
body := normalizeRequest(task.Kind, task.Request)
|
||||
modelType := modelTypeFromKind(task.Kind, body)
|
||||
if err := validateRequest(task.Kind, body); err != nil {
|
||||
failed, finishErr := s.failTask(ctx, task.ID, "bad_request", err.Error(), task.RunMode == "simulation", err)
|
||||
if finishErr != nil {
|
||||
@@ -102,6 +102,17 @@ func (s *Service) execute(ctx context.Context, task store.GatewayTask, user *aut
|
||||
if finishErr != nil {
|
||||
return Result{}, finishErr
|
||||
}
|
||||
if settleErr := s.store.SettleTaskBilling(ctx, finished); settleErr != nil {
|
||||
return Result{}, settleErr
|
||||
}
|
||||
if finished.FinalChargeAmount > 0 {
|
||||
if err := s.emit(ctx, task.ID, "task.billing.settled", "succeeded", "billing", 0.98, "task billing settled", map[string]any{
|
||||
"amount": finished.FinalChargeAmount,
|
||||
"currency": stringFromAny(record.BillingSummary["currency"]),
|
||||
}, isSimulation(task, candidate)); err != nil {
|
||||
return Result{}, err
|
||||
}
|
||||
}
|
||||
if err := s.emit(ctx, task.ID, "task.completed", "succeeded", "completed", 1, "task completed", map[string]any{
|
||||
"result": response.Result,
|
||||
"billings": billings,
|
||||
@@ -226,6 +237,7 @@ func (s *Service) runCandidate(ctx context.Context, task store.GatewayTask, user
|
||||
ErrorMessage: err.Error(),
|
||||
})
|
||||
_ = s.emit(ctx, task.ID, "task.attempt.failed", "running", "attempt_failed", 0.45, err.Error(), map[string]any{"attempt": attemptNo, "retryable": retryable, "requestId": requestID, "metrics": metrics}, simulated)
|
||||
s.applyCandidateFailurePolicies(ctx, task.ID, candidate, err, simulated)
|
||||
return clients.Response{}, err
|
||||
}
|
||||
uploadedResult, err := s.uploadGeneratedAssets(ctx, response.Result)
|
||||
@@ -317,7 +329,7 @@ func (s *Service) emit(ctx context.Context, taskID string, eventType string, sta
|
||||
return nil
|
||||
}
|
||||
|
||||
func modelTypeFromKind(kind string) string {
|
||||
func modelTypeFromKind(kind string, body map[string]any) string {
|
||||
switch kind {
|
||||
case "chat.completions", "responses":
|
||||
return "text_generate"
|
||||
@@ -327,12 +339,30 @@ func modelTypeFromKind(kind string) string {
|
||||
}
|
||||
return "image_generate"
|
||||
case "videos.generations":
|
||||
if videoRequestHasReferenceImage(body) {
|
||||
return "image_to_video"
|
||||
}
|
||||
return "video_generate"
|
||||
default:
|
||||
return "task"
|
||||
}
|
||||
}
|
||||
|
||||
func videoRequestHasReferenceImage(body map[string]any) bool {
|
||||
if body == nil {
|
||||
return false
|
||||
}
|
||||
for _, key := range []string{
|
||||
"image", "images", "image_url", "imageUrl", "image_urls", "imageUrls",
|
||||
"reference_image", "referenceImage", "first_frame", "firstFrame", "last_frame", "lastFrame",
|
||||
} {
|
||||
if hasAnyString(body, key) {
|
||||
return true
|
||||
}
|
||||
}
|
||||
return false
|
||||
}
|
||||
|
||||
func isTextGenerationKind(kind string) bool {
|
||||
return kind == "chat.completions" || kind == "responses"
|
||||
}
|
||||
@@ -345,10 +375,8 @@ func isSimulation(task store.GatewayTask, candidate store.RuntimeModelCandidate)
|
||||
}
|
||||
|
||||
func retryEnabled(candidate store.RuntimeModelCandidate) bool {
|
||||
if enabled, ok := candidate.ModelRetryPolicy["enabled"].(bool); ok {
|
||||
return enabled
|
||||
}
|
||||
if enabled, ok := candidate.PlatformRetryPolicy["enabled"].(bool); ok {
|
||||
policy := effectiveRetryPolicy(candidate)
|
||||
if enabled, ok := policy["enabled"].(bool); ok {
|
||||
return enabled
|
||||
}
|
||||
return true
|
||||
@@ -360,10 +388,7 @@ func maxAttemptsForCandidates(candidates []store.RuntimeModelCandidate) int {
|
||||
}
|
||||
maxAttempts := len(candidates)
|
||||
for _, candidate := range candidates {
|
||||
if value := intFromPolicy(candidate.ModelRetryPolicy, "maxAttempts"); value > 0 && value < maxAttempts {
|
||||
maxAttempts = value
|
||||
}
|
||||
if value := intFromPolicy(candidate.PlatformRetryPolicy, "maxAttempts"); value > 0 && value < maxAttempts {
|
||||
if value := intFromPolicy(effectiveRetryPolicy(candidate), "maxAttempts"); value > 0 && value < maxAttempts {
|
||||
maxAttempts = value
|
||||
}
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user