fix: recognize non-standard cache tokens from OpenAI-compatible channels

dto/openai_response.go

@@ -251,7 +251,7 @@ type OpenAIVideoResponse struct {
 
 type InputTokenDetails struct {
 	CachedTokens         int `json:"cached_tokens"`
-	CachedCreationTokens int `json:"-"`
+	CachedCreationTokens int `json:"cache_creation_input_tokens"`
 	TextTokens           int `json:"text_tokens"`
 	AudioTokens          int `json:"audio_tokens"`
 	ImageTokens          int `json:"image_tokens"`

relay/channel/openai/relay-openai.go

@@ -627,6 +627,19 @@ func applyUsagePostProcessing(info *relaycommon.RelayInfo, usage *dto.Usage, res
 				usage.PromptTokensDetails.CachedTokens = usage.PromptCacheHitTokens
 			}
 		}
+	default:
+		if usage.PromptTokensDetails.CachedTokens == 0 {
+			if usage.PromptCacheHitTokens > 0 {
+				usage.PromptTokensDetails.CachedTokens = usage.PromptCacheHitTokens
+			} else if cachedTokens, ok := extractCachedTokensFromBody(responseBody); ok {
+				usage.PromptTokensDetails.CachedTokens = cachedTokens
+			}
+		}
+		if usage.PromptTokensDetails.CachedCreationTokens == 0 {
+			if cacheCreationTokens, ok := extractCacheCreationTokensFromBody(responseBody); ok {
+				usage.PromptTokensDetails.CachedCreationTokens = cacheCreationTokens
+			}
+		}
 	}
 }
 
@@ -661,6 +674,29 @@ func extractCachedTokensFromBody(body []byte) (int, bool) {
 	return 0, false
 }
 
+func extractCacheCreationTokensFromBody(body []byte) (int, bool) {
+	if len(body) == 0 {
+		return 0, false
+	}
+
+	var payload struct {
+		Usage struct {
+			PromptTokensDetails struct {
+				CacheCreationInputTokens *int `json:"cache_creation_input_tokens"`
+			} `json:"prompt_tokens_details"`
+		} `json:"usage"`
+	}
+
+	if err := common.Unmarshal(body, &payload); err != nil {
+		return 0, false
+	}
+
+	if payload.Usage.PromptTokensDetails.CacheCreationInputTokens != nil {
+		return *payload.Usage.PromptTokensDetails.CacheCreationInputTokens, true
+	}
+	return 0, false
+}
+
 // extractMoonshotCachedTokensFromBody 从Moonshot的非标准位置提取cached_tokens
 // Moonshot的流式响应格式: {"choices":[{"usage":{"cached_tokens":111}}]}
 func extractMoonshotCachedTokensFromBody(body []byte) (int, bool) {

relay/channel/openai/relay_openai_test.go

@@ -0,0 +1,115 @@
+package openai
+
+import (
+	"testing"
+
+	"github.com/QuantumNous/new-api/constant"
+	"github.com/QuantumNous/new-api/dto"
+	relaycommon "github.com/QuantumNous/new-api/relay/common"
+)
+
+func newRelayInfo(channelType int) *relaycommon.RelayInfo {
+	return &relaycommon.RelayInfo{
+		ChannelMeta: &relaycommon.ChannelMeta{ChannelType: channelType},
+	}
+}
+
+func TestApplyUsagePostProcessing_DefaultCase_PromptCacheHitTokens(t *testing.T) {
+	info := newRelayInfo(0)
+	usage := &dto.Usage{
+		PromptCacheHitTokens: 200,
+	}
+	applyUsagePostProcessing(info, usage, nil)
+	if usage.PromptTokensDetails.CachedTokens != 200 {
+		t.Errorf("CachedTokens = %d, want 200", usage.PromptTokensDetails.CachedTokens)
+	}
+}
+
+func TestApplyUsagePostProcessing_DefaultCase_StepFunCachedTokens(t *testing.T) {
+	body := []byte(`{"usage":{"cached_tokens":150}}`)
+	info := newRelayInfo(0)
+	usage := &dto.Usage{}
+	applyUsagePostProcessing(info, usage, body)
+	if usage.PromptTokensDetails.CachedTokens != 150 {
+		t.Errorf("CachedTokens = %d, want 150", usage.PromptTokensDetails.CachedTokens)
+	}
+}
+
+func TestApplyUsagePostProcessing_DefaultCase_PrefersPromptCacheHitTokens(t *testing.T) {
+	body := []byte(`{"usage":{"cached_tokens":100}}`)
+	info := newRelayInfo(0)
+	usage := &dto.Usage{
+		PromptCacheHitTokens: 200,
+	}
+	applyUsagePostProcessing(info, usage, body)
+	if usage.PromptTokensDetails.CachedTokens != 200 {
+		t.Errorf("CachedTokens = %d, want 200 (should prefer PromptCacheHitTokens over body)", usage.PromptTokensDetails.CachedTokens)
+	}
+}
+
+func TestApplyUsagePostProcessing_DefaultCase_CacheCreationTokens(t *testing.T) {
+	body := []byte(`{"usage":{"prompt_tokens_details":{"cache_creation_input_tokens":300}}}`)
+	info := newRelayInfo(0)
+	usage := &dto.Usage{}
+	applyUsagePostProcessing(info, usage, body)
+	if usage.PromptTokensDetails.CachedCreationTokens != 300 {
+		t.Errorf("CachedCreationTokens = %d, want 300", usage.PromptTokensDetails.CachedCreationTokens)
+	}
+}
+
+func TestApplyUsagePostProcessing_DefaultCase_AlreadyPopulated(t *testing.T) {
+	body := []byte(`{"usage":{"cached_tokens":999,"prompt_tokens_details":{"cache_creation_input_tokens":999}}}`)
+	info := newRelayInfo(0)
+	usage := &dto.Usage{
+		PromptTokensDetails: dto.InputTokenDetails{
+			CachedTokens:         100,
+			CachedCreationTokens: 50,
+		},
+	}
+	applyUsagePostProcessing(info, usage, body)
+	if usage.PromptTokensDetails.CachedTokens != 100 {
+		t.Errorf("CachedTokens = %d, want 100 (should not overwrite)", usage.PromptTokensDetails.CachedTokens)
+	}
+	if usage.PromptTokensDetails.CachedCreationTokens != 50 {
+		t.Errorf("CachedCreationTokens = %d, want 50 (should not overwrite)", usage.PromptTokensDetails.CachedCreationTokens)
+	}
+}
+
+func TestApplyUsagePostProcessing_DeepSeek_Unaffected(t *testing.T) {
+	info := newRelayInfo(constant.ChannelTypeDeepSeek)
+	usage := &dto.Usage{
+		PromptCacheHitTokens: 500,
+	}
+	applyUsagePostProcessing(info, usage, nil)
+	if usage.PromptTokensDetails.CachedTokens != 500 {
+		t.Errorf("CachedTokens = %d, want 500", usage.PromptTokensDetails.CachedTokens)
+	}
+}
+
+func TestExtractCacheCreationTokensFromBody(t *testing.T) {
+	body := []byte(`{"usage":{"prompt_tokens_details":{"cache_creation_input_tokens":42}}}`)
+	tokens, ok := extractCacheCreationTokensFromBody(body)
+	if !ok {
+		t.Fatal("expected ok=true")
+	}
+	if tokens != 42 {
+		t.Errorf("tokens = %d, want 42", tokens)
+	}
+}
+
+func TestExtractCacheCreationTokensFromBody_Empty(t *testing.T) {
+	tokens, ok := extractCacheCreationTokensFromBody(nil)
+	if ok {
+		t.Errorf("expected ok=false for nil body, got tokens=%d", tokens)
+	}
+
+	tokens, ok = extractCacheCreationTokensFromBody([]byte(`{}`))
+	if ok {
+		t.Errorf("expected ok=false for empty JSON, got tokens=%d", tokens)
+	}
+
+	tokens, ok = extractCacheCreationTokensFromBody([]byte(`invalid`))
+	if ok {
+		t.Errorf("expected ok=false for invalid JSON, got tokens=%d", tokens)
+	}
+}