fix(billing): extend quantity validation and saturating conversions to remaining paths

Bound max-tokens fields across all relay format validators, saturate
tiered-expression rounding and audio/tool/task token conversions, and
route legacy remix ratios through the guarded setter.

relay/helper/valid_request.go

@@ -114,6 +114,20 @@ func GetAndValidateEmbeddingRequest(c *gin.Context, relayMode int) (*dto.Embeddi
 	return embeddingRequest, nil
 }
 
+// maxTokensLimit bounds user-supplied max token fields. These values feed
+// pre-consume quota math (preConsumedTokens * ratio); an unbounded value can
+// overflow the conversion and corrupt billing.
+const maxTokensLimit = math.MaxInt32 / 2
+
+func exceedsMaxTokensLimit(values ...*uint) bool {
+	for _, v := range values {
+		if lo.FromPtrOr(v, uint(0)) > maxTokensLimit {
+			return true
+		}
+	}
+	return false
+}
+
 func GetAndValidateResponsesRequest(c *gin.Context) (*dto.OpenAIResponsesRequest, error) {
 	request := &dto.OpenAIResponsesRequest{}
 	err := common.UnmarshalBodyReusable(c, request)
@@ -126,6 +140,9 @@ func GetAndValidateResponsesRequest(c *gin.Context) (*dto.OpenAIResponsesRequest
 	if request.Input == nil {
 		return nil, errors.New("input is required")
 	}
+	if exceedsMaxTokensLimit(request.MaxOutputTokens) {
+		return nil, errors.New("max_output_tokens is invalid")
+	}
 	return request, nil
 }
 
@@ -259,6 +276,9 @@ func GetAndValidateClaudeRequest(c *gin.Context) (textRequest *dto.ClaudeRequest
 	if textRequest.Model == "" {
 		return nil, errors.New("field model is required")
 	}
+	if exceedsMaxTokensLimit(textRequest.MaxTokens, textRequest.MaxTokensToSample) {
+		return nil, errors.New("max_tokens is invalid")
+	}
 
 	//if textRequest.Stream {
 	//	relayInfo.IsStream = true
@@ -281,7 +301,7 @@ func GetAndValidateTextRequest(c *gin.Context, relayMode int) (*dto.GeneralOpenA
 		textRequest.Model = c.Param("model")
 	}
 
-	if lo.FromPtrOr(textRequest.MaxTokens, uint(0)) > math.MaxInt32/2 {
+	if exceedsMaxTokensLimit(textRequest.MaxTokens, textRequest.MaxCompletionTokens) {
 		return nil, errors.New("max_tokens is invalid")
 	}
 	if textRequest.Model == "" {
@@ -334,6 +354,9 @@ func GetAndValidateGeminiRequest(c *gin.Context) (*dto.GeminiChatRequest, error)
 	if len(request.Contents) == 0 && len(request.Requests) == 0 {
 		return nil, errors.New("contents is required")
 	}
+	if exceedsMaxTokensLimit(request.GenerationConfig.MaxOutputTokens) {
+		return nil, errors.New("maxOutputTokens is invalid")
+	}
 
 	//if c.Query("alt") == "sse" {
 	//	relayInfo.IsStream = true