feat: image output token billing, channel-mapped billing source, credits balance precheck

- Parse candidatesTokensDetails from Gemini API to separate image/text output tokens
- Add image_output_tokens and image_output_cost to usage_log (migration 089)
- Support per-image-token pricing via output_cost_per_image_token from model pricing data
- Channel pricing ImageOutputPrice override works in token billing mode
- Auto-fill image_output_price in channel pricing form from model defaults
- Add "channel_mapped" billing model source as new default (migration 088)
- Bills by model name after channel mapping, before account mapping
- Fix channel cache error TTL sign error (115s → 5s)
- Fix Update channel only invalidating new groups, not removed groups
- Fix frontend model_mapping clearing sending undefined instead of {}
- Credits balance precheck via shared AccountUsageService cache before injection
- Skip credits injection for accounts with insufficient balance
- Don't mark credits exhausted for "exhausted your capacity on this model" 429s

backend/internal/service/model_pricing_resolver.go

@@ -134,6 +134,9 @@ func (r *ModelPricingResolver) applyTokenOverrides(chPricing *ChannelModelPricin
 		resolved.BasePricing.CacheReadPricePerToken = *chPricing.CacheReadPrice
 		resolved.BasePricing.CacheReadPricePerTokenPriority = *chPricing.CacheReadPrice
 	}
+	if chPricing.ImageOutputPrice != nil {
+		resolved.BasePricing.ImageOutputPricePerToken = *chPricing.ImageOutputPrice
+	}
 }
 
 // applyRequestTierOverrides 应用按次/图片模式的渠道覆盖

backend/internal/service/openai_gateway_service.go

@@ -204,6 +204,7 @@ type OpenAIUsage struct {
 	OutputTokens             int `json:"output_tokens"`
 	CacheCreationInputTokens int `json:"cache_creation_input_tokens,omitempty"`
 	CacheReadInputTokens     int `json:"cache_read_input_tokens,omitempty"`
+	ImageOutputTokens        int `json:"image_output_tokens,omitempty"`
 }
 
 // OpenAIForwardResult represents the result of forwarding
@@ -4177,6 +4178,7 @@ func (s *OpenAIGatewayService) RecordUsage(ctx context.Context, input *OpenAIRec
 		OutputTokens:        result.Usage.OutputTokens,
 		CacheCreationTokens: result.Usage.CacheCreationInputTokens,
 		CacheReadTokens:     result.Usage.CacheReadInputTokens,
+		ImageOutputTokens:   result.Usage.ImageOutputTokens,
 	}
 
 	// Get rate multiplier
@@ -4195,6 +4197,9 @@ func (s *OpenAIGatewayService) RecordUsage(ctx context.Context, input *OpenAIRec
 	if result.BillingModel != "" {
 		billingModel = strings.TrimSpace(result.BillingModel)
 	}
+	if input.BillingModelSource == BillingModelSourceChannelMapped && input.ChannelMappedModel != "" {
+		billingModel = input.ChannelMappedModel
+	}
 	if input.BillingModelSource == "requested" && input.OriginalModel != "" {
 		billingModel = input.OriginalModel
 	}
@@ -4255,8 +4260,10 @@ func (s *OpenAIGatewayService) RecordUsage(ctx context.Context, input *OpenAIRec
 		OutputTokens:          result.Usage.OutputTokens,
 		CacheCreationTokens:   result.Usage.CacheCreationInputTokens,
 		CacheReadTokens:       result.Usage.CacheReadInputTokens,
+		ImageOutputTokens:     result.Usage.ImageOutputTokens,
 		InputCost:             cost.InputCost,
 		OutputCost:            cost.OutputCost,
+		ImageOutputCost:       cost.ImageOutputCost,
 		CacheCreationCost:     cost.CacheCreationCost,
 		CacheReadCost:         cost.CacheReadCost,
 		TotalCost:             cost.TotalCost,

backend/internal/service/pricing_service.go

@@ -71,6 +71,7 @@ type LiteLLMModelPricing struct {
 	Mode                                string  `json:"mode"`
 	SupportsPromptCaching               bool    `json:"supports_prompt_caching"`
 	OutputCostPerImage                  float64 `json:"output_cost_per_image"`       // 图片生成模型每张图片价格
+	OutputCostPerImageToken             float64 `json:"output_cost_per_image_token"` // 图片输出 token 价格
 }
 
 // PricingRemoteClient 远程价格数据获取接口
@@ -94,6 +95,7 @@ type LiteLLMRawEntry struct {
 	Mode                                string   `json:"mode"`
 	SupportsPromptCaching               bool     `json:"supports_prompt_caching"`
 	OutputCostPerImage                  *float64 `json:"output_cost_per_image"`
+	OutputCostPerImageToken             *float64 `json:"output_cost_per_image_token"`
 }
 
 // PricingService 动态价格服务
@@ -408,6 +410,9 @@ func (s *PricingService) parsePricingData(body []byte) (map[string]*LiteLLMModel
 		if entry.OutputCostPerImage != nil {
 			pricing.OutputCostPerImage = *entry.OutputCostPerImage
 		}
+		if entry.OutputCostPerImageToken != nil {
+			pricing.OutputCostPerImageToken = *entry.OutputCostPerImageToken
+		}
 
 		result[modelName] = pricing
 	}

backend/internal/service/usage_log.go

@@ -134,6 +134,9 @@ type UsageLog struct {
 	CacheCreation5mTokens int `gorm:"column:cache_creation_5m_tokens"`
 	CacheCreation1hTokens int `gorm:"column:cache_creation_1h_tokens"`
 
+	ImageOutputTokens int
+	ImageOutputCost   float64
+
 	InputCost         float64
 	OutputCost        float64
 	CacheCreationCost float64

backend/migrations/088_channel_billing_model_source_channel_mapped.sql

+-- Change default billing_model_source for new channels to 'channel_mapped'
+-- Existing channels keep their current setting (no UPDATE on existing rows)
+ALTER TABLE channels ALTER COLUMN billing_model_source SET DEFAULT 'channel_mapped';

backend/migrations/089_usage_log_image_output_tokens.sql

+ALTER TABLE usage_logs ADD COLUMN IF NOT EXISTS image_output_tokens INTEGER NOT NULL DEFAULT 0;
+ALTER TABLE usage_logs ADD COLUMN IF NOT EXISTS image_output_cost DECIMAL(20, 10) NOT NULL DEFAULT 0;

frontend/src/api/admin/channels.ts

@@ -134,6 +134,7 @@ export interface ModelDefaultPricing {
   output_price?: number
   cache_write_price?: number
   cache_read_price?: number
+  image_output_price?: number
 }
 
 export async function getModelDefaultPricing(model: string): Promise<ModelDefaultPricing> {

frontend/src/components/admin/channel/PricingEntryCard.vue

@@ -328,6 +328,7 @@ async function onModelsUpdate(newModels: string[]) {
         output_price: perTokenToMTok(result.output_price ?? null),
         cache_write_price: perTokenToMTok(result.cache_write_price ?? null),
         cache_read_price: perTokenToMTok(result.cache_read_price ?? null),
+        image_output_price: perTokenToMTok(result.image_output_price ?? null),
       })
     }
   } catch {

frontend/src/i18n/locales/en.ts

@@ -1800,6 +1800,7 @@ export default {
         mappingSource: 'Source model',
         mappingTarget: 'Target model',
         billingModelSource: 'Billing Model',
+        billingModelSourceChannelMapped: 'Bill by channel-mapped model',
         billingModelSourceRequested: 'Bill by requested model',
         billingModelSourceUpstream: 'Bill by final upstream model',
         billingModelSourceHint: 'Controls which model name is used for pricing lookup',

frontend/src/i18n/locales/zh.ts

@@ -1880,6 +1880,7 @@ export default {
         mappingSource: '源模型',
         mappingTarget: '目标模型',
         billingModelSource: '计费基准',
+        billingModelSourceChannelMapped: '以渠道映射后的模型计费',
         billingModelSourceRequested: '以请求模型计费',
         billingModelSourceUpstream: '以最终模型计费',
         billingModelSourceHint: '控制使用哪个模型名称进行定价查找',

frontend/src/views/admin/ChannelsView.vue

@@ -471,6 +471,7 @@ const statusEditOptions = computed(() => [
 ])
 
 const billingModelSourceOptions = computed(() => [
+  { value: 'channel_mapped', label: t('admin.channels.form.billingModelSourceChannelMapped', 'Bill by channel-mapped model') },
   { value: 'requested', label: t('admin.channels.form.billingModelSourceRequested', 'Bill by requested model') },
   { value: 'upstream', label: t('admin.channels.form.billingModelSourceUpstream', 'Bill by final upstream model') }
 ])
@@ -504,7 +505,7 @@ const form = reactive({
   description: '',
   status: 'active',
   restrict_models: false,
-  billing_model_source: 'requested' as string,
+  billing_model_source: 'channel_mapped' as string,
   platforms: [] as PlatformSection[]
 })
 
@@ -819,7 +820,7 @@ function resetForm() {
   form.description = ''
   form.status = 'active'
   form.restrict_models = false
-  form.billing_model_source = 'requested'
+  form.billing_model_source = 'channel_mapped'
   form.platforms = []
   activeTab.value = 'basic'
 }
@@ -837,7 +838,7 @@ async function openEditDialog(channel: Channel) {
   form.description = channel.description || ''
   form.status = channel.status
   form.restrict_models = channel.restrict_models || false
-  form.billing_model_source = channel.billing_model_source || 'requested'
+  form.billing_model_source = channel.billing_model_source || 'channel_mapped'
   // Must load groups first so apiToForm can map groupID → platform
   await loadGroups()
   form.platforms = apiToForm(channel)
@@ -932,7 +933,7 @@ async function handleSubmit() {
         status: form.status,
         group_ids,
         model_pricing,
-        model_mapping: Object.keys(model_mapping).length > 0 ? model_mapping : undefined,
+        model_mapping: Object.keys(model_mapping).length > 0 ? model_mapping : {},
         billing_model_source: form.billing_model_source,
         restrict_models: form.restrict_models
       }
@@ -944,7 +945,7 @@ async function handleSubmit() {
         description: form.description.trim() || undefined,
         group_ids,
         model_pricing,
-        model_mapping: Object.keys(model_mapping).length > 0 ? model_mapping : undefined,
+        model_mapping: Object.keys(model_mapping).length > 0 ? model_mapping : {},
         billing_model_source: form.billing_model_source,
         restrict_models: form.restrict_models
       }