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/pkg/antigravity/claude_types.go

@@ -125,6 +125,7 @@ type ClaudeUsage 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"`
 }
 
 // ClaudeError Claude 错误响应

backend/internal/pkg/antigravity/gemini_types.go

@@ -149,13 +149,31 @@ type GeminiCandidate struct {
 	GroundingMetadata *GeminiGroundingMetadata `json:"groundingMetadata,omitempty"`
 }
 
+// GeminiTokenDetail Gemini token 详情（按模态分类）
+type GeminiTokenDetail struct {
+	Modality   string `json:"modality"`
+	TokenCount int    `json:"tokenCount"`
+}
+
 // GeminiUsageMetadata Gemini 用量元数据
 type GeminiUsageMetadata struct {
-	PromptTokenCount        int `json:"promptTokenCount,omitempty"`
-	CandidatesTokenCount    int `json:"candidatesTokenCount,omitempty"`
-	CachedContentTokenCount int `json:"cachedContentTokenCount,omitempty"`
-	TotalTokenCount         int `json:"totalTokenCount,omitempty"`
-	ThoughtsTokenCount      int `json:"thoughtsTokenCount,omitempty"` // thinking tokens（按输出价格计费）
+	PromptTokenCount        int                 `json:"promptTokenCount,omitempty"`
+	CandidatesTokenCount    int                 `json:"candidatesTokenCount,omitempty"`
+	CachedContentTokenCount int                 `json:"cachedContentTokenCount,omitempty"`
+	TotalTokenCount         int                 `json:"totalTokenCount,omitempty"`
+	ThoughtsTokenCount      int                 `json:"thoughtsTokenCount,omitempty"` // thinking tokens（按输出价格计费）
+	CandidatesTokensDetails []GeminiTokenDetail `json:"candidatesTokensDetails,omitempty"`
+	PromptTokensDetails     []GeminiTokenDetail `json:"promptTokensDetails,omitempty"`
+}
+
+// ImageOutputTokens 从 CandidatesTokensDetails 中提取 IMAGE 模态的 token 数
+func (m *GeminiUsageMetadata) ImageOutputTokens() int {
+	for _, d := range m.CandidatesTokensDetails {
+		if d.Modality == "IMAGE" {
+			return d.TokenCount
+		}
+	}
+	return 0
 }
 
 // GeminiGroundingMetadata Gemini grounding 元数据（Web Search）

backend/internal/pkg/antigravity/response_transformer.go

@@ -284,6 +284,7 @@ func (p *NonStreamingProcessor) buildResponse(geminiResp *GeminiResponse, respon
 		usage.InputTokens = geminiResp.UsageMetadata.PromptTokenCount - cached
 		usage.OutputTokens = geminiResp.UsageMetadata.CandidatesTokenCount + geminiResp.UsageMetadata.ThoughtsTokenCount
 		usage.CacheReadInputTokens = cached
+		usage.ImageOutputTokens = geminiResp.UsageMetadata.ImageOutputTokens()
 	}
 
 	// 生成响应 ID

backend/internal/pkg/antigravity/stream_transformer.go

@@ -32,9 +32,10 @@ type StreamingProcessor struct {
 	groundingChunks   []GeminiGroundingChunk
 
 	// 累计 usage
-	inputTokens     int
-	outputTokens    int
-	cacheReadTokens int
+	inputTokens       int
+	outputTokens      int
+	cacheReadTokens   int
+	imageOutputTokens int
 }
 
 // NewStreamingProcessor 创建流式响应处理器
@@ -45,6 +46,28 @@ func NewStreamingProcessor(originalModel string) *StreamingProcessor {
 	}
 }
 
+// SetUsageMapHook sets an optional hook that modifies usage maps before they are emitted.
+func (p *StreamingProcessor) SetUsageMapHook(fn UsageMapHook) {
+	p.usageMapHook = fn
+}
+
+func usageToMap(u ClaudeUsage) map[string]any {
+	m := map[string]any{
+		"input_tokens":  u.InputTokens,
+		"output_tokens": u.OutputTokens,
+	}
+	if u.CacheCreationInputTokens > 0 {
+		m["cache_creation_input_tokens"] = u.CacheCreationInputTokens
+	}
+	if u.CacheReadInputTokens > 0 {
+		m["cache_read_input_tokens"] = u.CacheReadInputTokens
+	}
+	if u.ImageOutputTokens > 0 {
+		m["image_output_tokens"] = u.ImageOutputTokens
+	}
+	return m
+}
+
 // ProcessLine 处理 SSE 行，返回 Claude SSE 事件
 func (p *StreamingProcessor) ProcessLine(line string) []byte {
 	line = strings.TrimSpace(line)
@@ -87,6 +110,7 @@ func (p *StreamingProcessor) ProcessLine(line string) []byte {
 		p.inputTokens = geminiResp.UsageMetadata.PromptTokenCount - cached
 		p.outputTokens = geminiResp.UsageMetadata.CandidatesTokenCount + geminiResp.UsageMetadata.ThoughtsTokenCount
 		p.cacheReadTokens = cached
+		p.imageOutputTokens = geminiResp.UsageMetadata.ImageOutputTokens()
 	}
 
 	// 处理 parts
@@ -127,6 +151,7 @@ func (p *StreamingProcessor) Finish() ([]byte, *ClaudeUsage) {
 		InputTokens:          p.inputTokens,
 		OutputTokens:         p.outputTokens,
 		CacheReadInputTokens: p.cacheReadTokens,
+		ImageOutputTokens:    p.imageOutputTokens,
 	}
 
 	if !p.messageStartSent {
@@ -158,6 +183,7 @@ func (p *StreamingProcessor) emitMessageStart(v1Resp *V1InternalResponse) []byte
 		usage.InputTokens = v1Resp.Response.UsageMetadata.PromptTokenCount - cached
 		usage.OutputTokens = v1Resp.Response.UsageMetadata.CandidatesTokenCount + v1Resp.Response.UsageMetadata.ThoughtsTokenCount
 		usage.CacheReadInputTokens = cached
+		usage.ImageOutputTokens = v1Resp.Response.UsageMetadata.ImageOutputTokens()
 	}
 
 	responseID := v1Resp.ResponseID
@@ -485,6 +511,7 @@ func (p *StreamingProcessor) emitFinish(finishReason string) []byte {
 		InputTokens:          p.inputTokens,
 		OutputTokens:         p.outputTokens,
 		CacheReadInputTokens: p.cacheReadTokens,
+		ImageOutputTokens:    p.imageOutputTokens,
 	}
 
 	deltaEvent := map[string]any{