feat(channel): 缓存扁平化 + 网关映射集成 + 计费模式统一 + 模型限制
- 缓存重构为 O(1) 哈希结构 (pricingByGroupModel, mappingByGroupModel) - 渠道模型映射接入网关流程 (Forward 前应用, a→b→c 映射链) - 新增 billing_model_source 配置 (请求模型/最终模型计费) - usage_logs 新增 channel_id, model_mapping_chain, billing_tier 字段 - 每种计费模式统一支持默认价格 + 区间定价 - 渠道模型限制开关 (restrict_models) - 分组按平台分类展示 + 彩色图标 - 必填字段红色星号 + 模型映射 UI - 去除模型通配符支持
This commit is contained in:
erio
@@ -7413,6 +7413,12 @@ type RecordUsageInput struct {
|
||||
RequestPayloadHash string // 请求体语义哈希，用于降低 request_id 误复用时的静默误去重风险
|
||||
ForceCacheBilling bool // 强制缓存计费：将 input_tokens 转为 cache_read 计费（用于粘性会话切换）
|
||||
APIKeyService APIKeyQuotaUpdater // 可选：用于更新API Key配额
|
||||
|
||||
// 渠道映射信息（由 handler 在 Forward 前解析）
|
||||
ChannelID int64 // 渠道 ID（0 = 无渠道）
|
||||
OriginalModel string // 用户原始请求模型（渠道映射前）
|
||||
BillingModelSource string // 计费模型来源："requested" / "upstream"
|
||||
ModelMappingChain string // 映射链描述，如 "a→b→c"
|
||||
}
|
||||
|
||||
// APIKeyQuotaUpdater defines the interface for updating API Key quota and rate limit usage
|
||||
@@ -7732,7 +7738,17 @@ func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInpu
|
||||
}
|
||||
|
||||
var cost *CostBreakdown
|
||||
// 确定计费模型
|
||||
billingModel := forwardResultBillingModel(result.Model, result.UpstreamModel)
|
||||
if input.BillingModelSource == BillingModelSourceRequested && input.OriginalModel != "" {
|
||||
billingModel = input.OriginalModel
|
||||
}
|
||||
|
||||
// 确定 RequestedModel（渠道映射前的原始模型）
|
||||
requestedModel := result.Model
|
||||
if input.OriginalModel != "" {
|
||||
requestedModel = input.OriginalModel
|
||||
}
|
||||
|
||||
// 根据请求类型选择计费方式
|
||||
if result.MediaType == "image" || result.MediaType == "video" {
|
||||
@@ -7815,7 +7831,7 @@ func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInpu
|
||||
AccountID: account.ID,
|
||||
RequestID: requestID,
|
||||
Model: result.Model,
|
||||
RequestedModel: result.Model,
|
||||
RequestedModel: requestedModel,
|
||||
UpstreamModel: optionalNonEqualStringPtr(result.UpstreamModel, result.Model),
|
||||
ReasoningEffort: result.ReasoningEffort,
|
||||
InboundEndpoint: optionalTrimmedStringPtr(input.InboundEndpoint),
|
||||
@@ -7842,6 +7858,8 @@ func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInpu
|
||||
ImageSize: imageSize,
|
||||
MediaType: mediaType,
|
||||
CacheTTLOverridden: cacheTTLOverridden,
|
||||
ChannelID: optionalInt64Ptr(input.ChannelID),
|
||||
ModelMappingChain: optionalTrimmedStringPtr(input.ModelMappingChain),
|
||||
CreatedAt: time.Now(),
|
||||
}
|
||||
|
||||
@@ -7909,6 +7927,12 @@ type RecordUsageLongContextInput struct {
|
||||
LongContextMultiplier float64 // 超出阈值部分的倍率（如 2.0）
|
||||
ForceCacheBilling bool // 强制缓存计费：将 input_tokens 转为 cache_read 计费（用于粘性会话切换）
|
||||
APIKeyService APIKeyQuotaUpdater // API Key 配额服务（可选）
|
||||
|
||||
// 渠道映射信息（由 handler 在 Forward 前解析）
|
||||
ChannelID int64 // 渠道 ID（0 = 无渠道）
|
||||
OriginalModel string // 用户原始请求模型（渠道映射前）
|
||||
BillingModelSource string // 计费模型来源："requested" / "upstream"
|
||||
ModelMappingChain string // 映射链描述，如 "a→b→c"
|
||||
}
|
||||
|
||||
// RecordUsageWithLongContext 记录使用量并扣费，支持长上下文双倍计费（用于 Gemini）
|
||||
@@ -7946,7 +7970,17 @@ func (s *GatewayService) RecordUsageWithLongContext(ctx context.Context, input *
|
||||
}
|
||||
|
||||
var cost *CostBreakdown
|
||||
// 确定计费模型
|
||||
billingModel := forwardResultBillingModel(result.Model, result.UpstreamModel)
|
||||
if input.BillingModelSource == BillingModelSourceRequested && input.OriginalModel != "" {
|
||||
billingModel = input.OriginalModel
|
||||
}
|
||||
|
||||
// 确定 RequestedModel（渠道映射前的原始模型）
|
||||
requestedModel := result.Model
|
||||
if input.OriginalModel != "" {
|
||||
requestedModel = input.OriginalModel
|
||||
}
|
||||
|
||||
// 根据请求类型选择计费方式
|
||||
if result.ImageCount > 0 {
|
||||
@@ -8008,7 +8042,7 @@ func (s *GatewayService) RecordUsageWithLongContext(ctx context.Context, input *
|
||||
AccountID: account.ID,
|
||||
RequestID: requestID,
|
||||
Model: result.Model,
|
||||
RequestedModel: result.Model,
|
||||
RequestedModel: requestedModel,
|
||||
UpstreamModel: optionalNonEqualStringPtr(result.UpstreamModel, result.Model),
|
||||
ReasoningEffort: result.ReasoningEffort,
|
||||
InboundEndpoint: optionalTrimmedStringPtr(input.InboundEndpoint),
|
||||
@@ -8034,6 +8068,8 @@ func (s *GatewayService) RecordUsageWithLongContext(ctx context.Context, input *
|
||||
ImageCount: result.ImageCount,
|
||||
ImageSize: imageSize,
|
||||
CacheTTLOverridden: cacheTTLOverridden,
|
||||
ChannelID: optionalInt64Ptr(input.ChannelID),
|
||||
ModelMappingChain: optionalTrimmedStringPtr(input.ModelMappingChain),
|
||||
CreatedAt: time.Now(),
|
||||
}
|
||||
|
||||
@@ -8085,6 +8121,27 @@ func (s *GatewayService) RecordUsageWithLongContext(ctx context.Context, input *
|
||||
return nil
|
||||
}
|
||||
|
||||
// ResolveChannelMapping 委托渠道服务解析模型映射
|
||||
func (s *GatewayService) ResolveChannelMapping(ctx context.Context, groupID int64, model string) ChannelMappingResult {
|
||||
if s.channelService == nil {
|
||||
return ChannelMappingResult{MappedModel: model}
|
||||
}
|
||||
return s.channelService.ResolveChannelMapping(ctx, groupID, model)
|
||||
}
|
||||
|
||||
// ReplaceModelInBody 替换请求体中的模型名（导出供 handler 使用）
|
||||
func (s *GatewayService) ReplaceModelInBody(body []byte, newModel string) []byte {
|
||||
return s.replaceModelInBody(body, newModel)
|
||||
}
|
||||
|
||||
// IsModelRestricted 检查模型是否被渠道限制
|
||||
func (s *GatewayService) IsModelRestricted(ctx context.Context, groupID int64, model string) bool {
|
||||
if s.channelService == nil {
|
||||
return false
|
||||
}
|
||||
return s.channelService.IsModelRestricted(ctx, groupID, model)
|
||||
}
|
||||
|
||||
// ForwardCountTokens 转发 count_tokens 请求到上游 API
|
||||
// 特点：不记录使用量、仅支持非流式响应
|
||||
func (s *GatewayService) ForwardCountTokens(ctx context.Context, c *gin.Context, account *Account, parsed *ParsedRequest) error {
|
||||
|
||||
Reference in New Issue
Block a user