From 77f80ef1f4d35acf6bf63d24401c36c2e29a8eb0 Mon Sep 17 00:00:00 2001 From: shaw Date: Sun, 17 Aug 2025 00:41:16 +0800 Subject: [PATCH] =?UTF-8?q?fix:=20claude=20token=E7=BB=9F=E8=AE=A1?= =?UTF-8?q?=E9=97=AE=E9=A2=98?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/routes/openaiClaudeRoutes.js | 20 +++----- src/services/claudeRelayService.js | 74 ++++++++++++++++++++---------- 2 files changed, 55 insertions(+), 39 deletions(-) diff --git a/src/routes/openaiClaudeRoutes.js b/src/routes/openaiClaudeRoutes.js index 63308a72..b2c43ed9 100644 --- a/src/routes/openaiClaudeRoutes.js +++ b/src/routes/openaiClaudeRoutes.js @@ -250,19 +250,13 @@ async function handleChatCompletion(req, res, apiKeyData) { (usage) => { // 记录使用统计 if (usage && usage.input_tokens !== undefined && usage.output_tokens !== undefined) { - const inputTokens = usage.input_tokens || 0 - const outputTokens = usage.output_tokens || 0 - const cacheCreateTokens = usage.cache_creation_input_tokens || 0 - const cacheReadTokens = usage.cache_read_input_tokens || 0 const model = usage.model || claudeRequest.model + // 使用新的 recordUsageWithDetails 方法来支持详细的缓存数据 apiKeyService - .recordUsage( + .recordUsageWithDetails( apiKeyData.id, - inputTokens, - outputTokens, - cacheCreateTokens, - cacheReadTokens, + usage, // 直接传递整个 usage 对象，包含可能的 cache_creation 详细数据 model, accountId ) @@ -328,13 +322,11 @@ async function handleChatCompletion(req, res, apiKeyData) { // 记录使用统计 if (claudeData.usage) { const { usage } = claudeData + // 使用新的 recordUsageWithDetails 方法来支持详细的缓存数据 apiKeyService - .recordUsage( + .recordUsageWithDetails( apiKeyData.id, - usage.input_tokens || 0, - usage.output_tokens || 0, - usage.cache_creation_input_tokens || 0, - usage.cache_read_input_tokens || 0, + usage, // 直接传递整个 usage 对象，包含可能的 cache_creation 详细数据 claudeRequest.model, accountId ) diff --git a/src/services/claudeRelayService.js b/src/services/claudeRelayService.js index 50fb739c..fa6c39b4 100644 --- a/src/services/claudeRelayService.js +++ b/src/services/claudeRelayService.js @@ -279,16 +279,10 @@ class ClaudeRelayService { if (responseBody && responseBody.usage) { const { usage } = responseBody - logger.info('📊 === Non-Stream Request Usage Summary ===', { - request_model: requestBody.model, - response_model: responseBody.model || requestBody.model, - input_tokens: usage.input_tokens || 0, - output_tokens: usage.output_tokens || 0, - cache_creation_tokens: usage.cache_creation_input_tokens || 0, - cache_read_tokens: usage.cache_read_input_tokens || 0, - api_key: apiKeyData.name, - account_id: accountId - }) + // 打印原始usage数据为JSON字符串 + logger.info( + `📊 === Non-Stream Request Usage Summary === Model: ${requestBody.model}, Usage: ${JSON.stringify(usage)}` + ) } else { // 如果没有usage数据，使用估算值 const inputTokens = requestBody.messages @@ -1008,13 +1002,15 @@ class ClaudeRelayService { // 如果已经收集到了input数据和output数据，这是一个完整的usage if (currentUsageData.input_tokens !== undefined) { - logger.info( + logger.debug( '🎯 Complete usage data collected for model:', - currentUsageData.model + currentUsageData.model, + '- Input:', + currentUsageData.input_tokens, + 'Output:', + currentUsageData.output_tokens ) - // 触发回调记录这个usage - usageCallback(currentUsageData) - // 保存到列表中 + // 保存到列表中，但不立即触发回调 allUsageData.push({ ...currentUsageData }) // 重置当前数据，准备接收下一个 currentUsageData = {} @@ -1080,7 +1076,6 @@ class ClaudeRelayService { if (currentUsageData.output_tokens === undefined) { currentUsageData.output_tokens = 0 // 如果没有output，设为0 } - usageCallback(currentUsageData) allUsageData.push(currentUsageData) } @@ -1104,16 +1099,45 @@ class ClaudeRelayService { {} ) - logger.info('📊 === Stream Request Usage Summary ===', { - request_body_model: body.model, - total_input_tokens: totalUsage.input_tokens, - total_output_tokens: totalUsage.output_tokens, - total_cache_creation: totalUsage.cache_creation_input_tokens, - total_cache_read: totalUsage.cache_read_input_tokens, - models_used: [...new Set(totalUsage.models)], - usage_events_count: allUsageData.length, - detailed_usage: allUsageData + // 打印原始的usage数据为JSON字符串，避免嵌套问题 + logger.info( + `📊 === Stream Request Usage Summary === Model: ${body.model}, Total Events: ${allUsageData.length}, Usage Data: ${JSON.stringify(allUsageData)}` + ) + + // 一般一个请求只会使用一个模型，即使有多个usage事件也应该合并 + // 计算总的usage + const finalUsage = { + input_tokens: totalUsage.input_tokens, + output_tokens: totalUsage.output_tokens, + cache_creation_input_tokens: totalUsage.cache_creation_input_tokens, + cache_read_input_tokens: totalUsage.cache_read_input_tokens, + model: allUsageData[allUsageData.length - 1].model || body.model // 使用最后一个模型或请求模型 + } + + // 如果有详细的cache_creation数据，合并它们 + let totalEphemeral5m = 0 + let totalEphemeral1h = 0 + allUsageData.forEach((usage) => { + if (usage.cache_creation && typeof usage.cache_creation === 'object') { + totalEphemeral5m += usage.cache_creation.ephemeral_5m_input_tokens || 0 + totalEphemeral1h += usage.cache_creation.ephemeral_1h_input_tokens || 0 + } }) + + // 如果有详细的缓存数据，添加到finalUsage + if (totalEphemeral5m > 0 || totalEphemeral1h > 0) { + finalUsage.cache_creation = { + ephemeral_5m_input_tokens: totalEphemeral5m, + ephemeral_1h_input_tokens: totalEphemeral1h + } + logger.info( + '📊 Detailed cache creation breakdown:', + JSON.stringify(finalUsage.cache_creation) + ) + } + + // 调用一次usageCallback记录合并后的数据 + usageCallback(finalUsage) } // 处理限流状态