github/claude-relay-service
1
0
Fork 0
mirror of https://github.com/Wei-Shaw/claude-relay-service.git synced 2026-01-22 16:43:35 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'upstream/main'

Browse Source 
# Conflicts:
#	src/routes/api.js
This commit is contained in:
於林涛
2025-10-08 19:34:17 +08:00
parent bb95c0b7d5 cffd023239
commit 705bd7611c
13 changed files with 399 additions and 179 deletions
Download Patch File Download Diff File Expand all files Collapse all files

51
src/routes/admin.js
View File

@@ -4162,6 +4162,36 @@ router.get('/accounts/:accountId/usage-history', authenticateAdmin, async (req,
gemini: 'gemini-1.5-flash'
}
// 获取账户信息以获取创建时间
let accountData = null
let accountCreatedAt = null
try {
switch (platform) {
case 'claude':
accountData = await claudeAccountService.getAccount(accountId)
break
case 'claude-console':
accountData = await claudeConsoleAccountService.getAccount(accountId)
break
case 'openai':
accountData = await openaiAccountService.getAccount(accountId)
break
case 'openai-responses':
accountData = await openaiResponsesAccountService.getAccount(accountId)
break
case 'gemini':
accountData = await geminiAccountService.getAccount(accountId)
break
}
if (accountData && accountData.createdAt) {
accountCreatedAt = new Date(accountData.createdAt)
}
} catch (error) {
logger.warn(`Failed to get account data for avgDailyCost calculation: ${error.message}`)
}
const client = redis.getClientSafe()
const fallbackModel = fallbackModelMap[platform] || 'unknown'
const daysCount = Math.min(Math.max(parseInt(days, 10) || 30, 1), 60)
@@ -4281,9 +4311,22 @@ router.get('/accounts/:accountId/usage-history', authenticateAdmin, async (req,
})
}
const avgDailyCost = daysCount > 0 ? totalCost / daysCount : 0
const avgDailyRequests = daysCount > 0 ? totalRequests / daysCount : 0
const avgDailyTokens = daysCount > 0 ? totalTokens / daysCount : 0
// 计算实际使用天数（从账户创建到现在）
let actualDaysForAvg = daysCount
if (accountCreatedAt) {
const now = new Date()
const diffTime = Math.abs(now - accountCreatedAt)
const diffDays = Math.ceil(diffTime / (1000 * 60 * 60 * 24))
// 使用实际使用天数，但不超过请求的天数范围
actualDaysForAvg = Math.min(diffDays, daysCount)
// 至少为1天，避免除零
actualDaysForAvg = Math.max(actualDaysForAvg, 1)
}
// 使用实际天数计算日均值
const avgDailyCost = actualDaysForAvg > 0 ? totalCost / actualDaysForAvg : 0
const avgDailyRequests = actualDaysForAvg > 0 ? totalRequests / actualDaysForAvg : 0
const avgDailyTokens = actualDaysForAvg > 0 ? totalTokens / actualDaysForAvg : 0
const todayData = history.length > 0 ? history[history.length - 1] : null
@@ -4293,6 +4336,8 @@ router.get('/accounts/:accountId/usage-history', authenticateAdmin, async (req,
history,
summary: {
days: daysCount,
actualDaysUsed: actualDaysForAvg, // 实际使用的天数（用于计算日均值）
accountCreatedAt: accountCreatedAt ? accountCreatedAt.toISOString() : null,
totalCost,
totalCostFormatted: CostCalculator.formatCost(totalCost),
totalRequests,

217
src/routes/api.js
View File

@@ -6,10 +6,8 @@ const ccrRelayService = require('../services/ccrRelayService')
const bedrockAccountService = require('../services/bedrockAccountService')
const unifiedClaudeScheduler = require('../services/unifiedClaudeScheduler')
const apiKeyService = require('../services/apiKeyService')
const pricingService = require('../services/pricingService')
const { authenticateApiKey } = require('../middleware/auth')
const logger = require('../utils/logger')
const redis = require('../models/redis')
const { getEffectiveModel, parseVendorPrefixedModel } = require('../utils/modelHelper')
const sessionHelper = require('../utils/sessionHelper')
const openaiToClaude = require('../services/openaiToClaude')
@@ -34,6 +32,33 @@ function detectBackendFromModel(modelName) {
return 'claude' // 默认使用 Claude
}
const { updateRateLimitCounters } = require('../utils/rateLimitHelper')
const router = express.Router()
function queueRateLimitUpdate(rateLimitInfo, usageSummary, model, context = '') {
if (!rateLimitInfo) {
return Promise.resolve({ totalTokens: 0, totalCost: 0 })
}
const label = context ? ` (${context})` : ''
return updateRateLimitCounters(rateLimitInfo, usageSummary, model)
.then(({ totalTokens, totalCost }) => {
if (totalTokens > 0) {
logger.api(`📊 Updated rate limit token count${label}: +${totalTokens} tokens`)
}
if (typeof totalCost === 'number' && totalCost > 0) {
logger.api(`💰 Updated rate limit cost count${label}: +$${totalCost.toFixed(6)}`)
}
return { totalTokens, totalCost }
})
.catch((error) => {
logger.error(`❌ Failed to update rate limit counters${label}:`, error)
return { totalTokens: 0, totalCost: 0 }
})
}
// 🔧 共享的消息处理函数
async function handleMessagesRequest(req, res) {
try {
@@ -210,35 +235,17 @@ async function handleMessagesRequest(req, res) {
logger.error('❌ Failed to record stream usage:', error)
})
// 更新时间窗口内的token计数和费用
if (req.rateLimitInfo) {
const totalTokens = inputTokens + outputTokens + cacheCreateTokens + cacheReadTokens
// 更新Token计数（向后兼容）
redis
.getClient()
.incrby(req.rateLimitInfo.tokenCountKey, totalTokens)
.catch((error) => {
logger.error('❌ Failed to update rate limit token count:', error)
})
logger.api(`📊 Updated rate limit token count: +${totalTokens} tokens`)
// 计算并更新费用计数（新功能）
if (req.rateLimitInfo.costCountKey) {
const costInfo = pricingService.calculateCost(usageData, model)
if (costInfo.totalCost > 0) {
redis
.getClient()
.incrbyfloat(req.rateLimitInfo.costCountKey, costInfo.totalCost)
.catch((error) => {
logger.error('❌ Failed to update rate limit cost count:', error)
})
logger.api(
`💰 Updated rate limit cost count: +$${costInfo.totalCost.toFixed(6)}`
)
}
}
}
queueRateLimitUpdate(
req.rateLimitInfo,
{
inputTokens,
outputTokens,
cacheCreateTokens,
cacheReadTokens
},
model,
'claude-stream'
)
usageDataCaptured = true
logger.api(
@@ -319,35 +326,17 @@ async function handleMessagesRequest(req, res) {
logger.error('❌ Failed to record stream usage:', error)
})
// 更新时间窗口内的token计数和费用
if (req.rateLimitInfo) {
const totalTokens = inputTokens + outputTokens + cacheCreateTokens + cacheReadTokens
// 更新Token计数（向后兼容）
redis
.getClient()
.incrby(req.rateLimitInfo.tokenCountKey, totalTokens)
.catch((error) => {
logger.error('❌ Failed to update rate limit token count:', error)
})
logger.api(`📊 Updated rate limit token count: +${totalTokens} tokens`)
// 计算并更新费用计数（新功能）
if (req.rateLimitInfo.costCountKey) {
const costInfo = pricingService.calculateCost(usageData, model)
if (costInfo.totalCost > 0) {
redis
.getClient()
.incrbyfloat(req.rateLimitInfo.costCountKey, costInfo.totalCost)
.catch((error) => {
logger.error('❌ Failed to update rate limit cost count:', error)
})
logger.api(
`💰 Updated rate limit cost count: +$${costInfo.totalCost.toFixed(6)}`
)
}
}
}
queueRateLimitUpdate(
req.rateLimitInfo,
{
inputTokens,
outputTokens,
cacheCreateTokens,
cacheReadTokens
},
model,
'claude-console-stream'
)
usageDataCaptured = true
logger.api(
@@ -387,33 +376,17 @@ async function handleMessagesRequest(req, res) {
logger.error('❌ Failed to record Bedrock stream usage:', error)
})
// 更新时间窗口内的token计数和费用
if (req.rateLimitInfo) {
const totalTokens = inputTokens + outputTokens
// 更新Token计数（向后兼容）
redis
.getClient()
.incrby(req.rateLimitInfo.tokenCountKey, totalTokens)
.catch((error) => {
logger.error('❌ Failed to update rate limit token count:', error)
})
logger.api(`📊 Updated rate limit token count: +${totalTokens} tokens`)
// 计算并更新费用计数（新功能）
if (req.rateLimitInfo.costCountKey) {
const costInfo = pricingService.calculateCost(result.usage, result.model)
if (costInfo.totalCost > 0) {
redis
.getClient()
.incrbyfloat(req.rateLimitInfo.costCountKey, costInfo.totalCost)
.catch((error) => {
logger.error('❌ Failed to update rate limit cost count:', error)
})
logger.api(`💰 Updated rate limit cost count: +$${costInfo.totalCost.toFixed(6)}`)
}
}
}
queueRateLimitUpdate(
req.rateLimitInfo,
{
inputTokens,
outputTokens,
cacheCreateTokens: 0,
cacheReadTokens: 0
},
result.model,
'bedrock-stream'
)
usageDataCaptured = true
logger.api(
@@ -488,35 +461,17 @@ async function handleMessagesRequest(req, res) {
logger.error('❌ Failed to record CCR stream usage:', error)
})
// 更新时间窗口内的token计数和费用
if (req.rateLimitInfo) {
const totalTokens = inputTokens + outputTokens + cacheCreateTokens + cacheReadTokens
// 更新Token计数（向后兼容）
redis
.getClient()
.incrby(req.rateLimitInfo.tokenCountKey, totalTokens)
.catch((error) => {
logger.error('❌ Failed to update rate limit token count:', error)
})
logger.api(`📊 Updated rate limit token count: +${totalTokens} tokens`)
// 计算并更新费用计数（新功能）
if (req.rateLimitInfo.costCountKey) {
const costInfo = pricingService.calculateCost(usageData, model)
if (costInfo.totalCost > 0) {
redis
.getClient()
.incrbyfloat(req.rateLimitInfo.costCountKey, costInfo.totalCost)
.catch((error) => {
logger.error('❌ Failed to update rate limit cost count:', error)
})
logger.api(
`💰 Updated rate limit cost count: +$${costInfo.totalCost.toFixed(6)}`
)
}
}
}
queueRateLimitUpdate(
req.rateLimitInfo,
{
inputTokens,
outputTokens,
cacheCreateTokens,
cacheReadTokens
},
model,
'ccr-stream'
)
usageDataCaptured = true
logger.api(
@@ -704,25 +659,17 @@ async function handleMessagesRequest(req, res) {
responseAccountId
)
// 更新时间窗口内的token计数和费用
if (req.rateLimitInfo) {
const totalTokens = inputTokens + outputTokens + cacheCreateTokens + cacheReadTokens
// 更新Token计数（向后兼容）
await redis.getClient().incrby(req.rateLimitInfo.tokenCountKey, totalTokens)
logger.api(`📊 Updated rate limit token count: +${totalTokens} tokens`)
// 计算并更新费用计数（新功能）
if (req.rateLimitInfo.costCountKey) {
const costInfo = pricingService.calculateCost(jsonData.usage, model)
if (costInfo.totalCost > 0) {
await redis
.getClient()
.incrbyfloat(req.rateLimitInfo.costCountKey, costInfo.totalCost)
logger.api(`💰 Updated rate limit cost count: +$${costInfo.totalCost.toFixed(6)}`)
}
}
}
await queueRateLimitUpdate(
req.rateLimitInfo,
{
inputTokens,
outputTokens,
cacheCreateTokens,
cacheReadTokens
},
model,
'claude-non-stream'
)
usageRecorded = true
logger.api(

50
src/routes/geminiRoutes.js
View File

@@ -8,6 +8,7 @@ const crypto = require('crypto')
const sessionHelper = require('../utils/sessionHelper')
const unifiedGeminiScheduler = require('../services/unifiedGeminiScheduler')
const apiKeyService = require('../services/apiKeyService')
const { updateRateLimitCounters } = require('../utils/rateLimitHelper')
// const { OAuth2Client } = require('google-auth-library'); // OAuth2Client is not used in this file
// 生成会话哈希
@@ -49,6 +50,31 @@ function ensureGeminiPermission(req, res) {
return false
}
async function applyRateLimitTracking(req, usageSummary, model, context = '') {
if (!req.rateLimitInfo) {
return
}
const label = context ? ` (${context})` : ''
try {
const { totalTokens, totalCost } = await updateRateLimitCounters(
req.rateLimitInfo,
usageSummary,
model
)
if (totalTokens > 0) {
logger.api(`📊 Updated rate limit token count${label}: +${totalTokens} tokens`)
}
if (typeof totalCost === 'number' && totalCost > 0) {
logger.api(`💰 Updated rate limit cost count${label}: +$${totalCost.toFixed(6)}`)
}
} catch (error) {
logger.error(`❌ Failed to update rate limit counters${label}:`, error)
}
}
// Gemini 消息处理端点
router.post('/messages', authenticateApiKey, async (req, res) => {
const startTime = Date.now()
@@ -679,6 +705,18 @@ async function handleGenerateContent(req, res) {
logger.info(
`📊 Recorded Gemini usage - Input: ${usage.promptTokenCount}, Output: ${usage.candidatesTokenCount}, Total: ${usage.totalTokenCount}`
)
await applyRateLimitTracking(
req,
{
inputTokens: usage.promptTokenCount || 0,
outputTokens: usage.candidatesTokenCount || 0,
cacheCreateTokens: 0,
cacheReadTokens: 0
},
model,
'gemini-non-stream'
)
} catch (error) {
logger.error('Failed to record Gemini usage:', error)
}
@@ -935,6 +973,18 @@ async function handleStreamGenerateContent(req, res) {
logger.info(
`📊 Recorded Gemini stream usage - Input: ${totalUsage.promptTokenCount}, Output: ${totalUsage.candidatesTokenCount}, Total: ${totalUsage.totalTokenCount}`
)
await applyRateLimitTracking(
req,
{
inputTokens: totalUsage.promptTokenCount || 0,
outputTokens: totalUsage.candidatesTokenCount || 0,
cacheCreateTokens: 0,
cacheReadTokens: 0
},
model,
'gemini-stream'
)
} catch (error) {
logger.error('Failed to record Gemini usage:', error)
}

58
src/routes/openaiClaudeRoutes.js
View File

@@ -15,6 +15,7 @@ const apiKeyService = require('../services/apiKeyService')
const unifiedClaudeScheduler = require('../services/unifiedClaudeScheduler')
const claudeCodeHeadersService = require('../services/claudeCodeHeadersService')
const sessionHelper = require('../utils/sessionHelper')
const { updateRateLimitCounters } = require('../utils/rateLimitHelper')
// 加载模型定价数据
let modelPricingData = {}
@@ -33,6 +34,27 @@ function checkPermissions(apiKeyData, requiredPermission = 'claude') {
return permissions === 'all' || permissions === requiredPermission
}
function queueRateLimitUpdate(rateLimitInfo, usageSummary, model, context = '') {
if (!rateLimitInfo) {
return
}
const label = context ? ` (${context})` : ''
updateRateLimitCounters(rateLimitInfo, usageSummary, model)
.then(({ totalTokens, totalCost }) => {
if (totalTokens > 0) {
logger.api(`📊 Updated rate limit token count${label}: +${totalTokens} tokens`)
}
if (typeof totalCost === 'number' && totalCost > 0) {
logger.api(`💰 Updated rate limit cost count${label}: +$${totalCost.toFixed(6)}`)
}
})
.catch((error) => {
logger.error(`❌ Failed to update rate limit counters${label}:`, error)
})
}
// 📋 OpenAI 兼容的模型列表端点
router.get('/v1/models', authenticateApiKey, async (req, res) => {
try {
@@ -263,6 +285,12 @@ async function handleChatCompletion(req, res, apiKeyData) {
// 记录使用统计
if (usage && usage.input_tokens !== undefined && usage.output_tokens !== undefined) {
const model = usage.model || claudeRequest.model
const cacheCreateTokens =
(usage.cache_creation && typeof usage.cache_creation === 'object'
? (usage.cache_creation.ephemeral_5m_input_tokens || 0) +
(usage.cache_creation.ephemeral_1h_input_tokens || 0)
: usage.cache_creation_input_tokens || 0) || 0
const cacheReadTokens = usage.cache_read_input_tokens || 0
// 使用新的 recordUsageWithDetails 方法来支持详细的缓存数据
apiKeyService
@@ -275,6 +303,18 @@ async function handleChatCompletion(req, res, apiKeyData) {
.catch((error) => {
logger.error('❌ Failed to record usage:', error)
})
queueRateLimitUpdate(
req.rateLimitInfo,
{
inputTokens: usage.input_tokens || 0,
outputTokens: usage.output_tokens || 0,
cacheCreateTokens,
cacheReadTokens
},
model,
'openai-claude-stream'
)
}
},
// 流转换器
@@ -334,6 +374,12 @@ async function handleChatCompletion(req, res, apiKeyData) {
// 记录使用统计
if (claudeData.usage) {
const { usage } = claudeData
const cacheCreateTokens =
(usage.cache_creation && typeof usage.cache_creation === 'object'
? (usage.cache_creation.ephemeral_5m_input_tokens || 0) +
(usage.cache_creation.ephemeral_1h_input_tokens || 0)
: usage.cache_creation_input_tokens || 0) || 0
const cacheReadTokens = usage.cache_read_input_tokens || 0
// 使用新的 recordUsageWithDetails 方法来支持详细的缓存数据
apiKeyService
.recordUsageWithDetails(
@@ -345,6 +391,18 @@ async function handleChatCompletion(req, res, apiKeyData) {
.catch((error) => {
logger.error('❌ Failed to record usage:', error)
})
queueRateLimitUpdate(
req.rateLimitInfo,
{
inputTokens: usage.input_tokens || 0,
outputTokens: usage.output_tokens || 0,
cacheCreateTokens,
cacheReadTokens
},
claudeRequest.model,
'openai-claude-non-stream'
)
}
// 返回 OpenAI 格式响应

50
src/routes/openaiRoutes.js
View File

@@ -11,6 +11,7 @@ const openaiResponsesRelayService = require('../services/openaiResponsesRelaySer
const apiKeyService = require('../services/apiKeyService')
const crypto = require('crypto')
const ProxyHelper = require('../utils/proxyHelper')
const { updateRateLimitCounters } = require('../utils/rateLimitHelper')
// 创建代理 Agent（使用统一的代理工具）
function createProxyAgent(proxy) {
@@ -67,6 +68,31 @@ function extractCodexUsageHeaders(headers) {
return hasData ? snapshot : null
}
async function applyRateLimitTracking(req, usageSummary, model, context = '') {
if (!req.rateLimitInfo) {
return
}
const label = context ? ` (${context})` : ''
try {
const { totalTokens, totalCost } = await updateRateLimitCounters(
req.rateLimitInfo,
usageSummary,
model
)
if (totalTokens > 0) {
logger.api(`📊 Updated rate limit token count${label}: +${totalTokens} tokens`)
}
if (typeof totalCost === 'number' && totalCost > 0) {
logger.api(`💰 Updated rate limit cost count${label}: +$${totalCost.toFixed(6)}`)
}
} catch (error) {
logger.error(`❌ Failed to update rate limit counters${label}:`, error)
}
}
// 使用统一调度器选择 OpenAI 账户
async function getOpenAIAuthToken(apiKeyData, sessionId = null, requestedModel = null) {
try {
@@ -579,6 +605,18 @@ const handleResponses = async (req, res) => {
logger.info(
`📊 Recorded OpenAI non-stream usage - Input: ${totalInputTokens}(actual:${actualInputTokens}+cached:${cacheReadTokens}), Output: ${outputTokens}, Total: ${usageData.total_tokens || totalInputTokens + outputTokens}, Model: ${actualModel}`
)
await applyRateLimitTracking(
req,
{
inputTokens: actualInputTokens,
outputTokens,
cacheCreateTokens: 0,
cacheReadTokens
},
actualModel,
'openai-non-stream'
)
}
// 返回响应
@@ -700,6 +738,18 @@ const handleResponses = async (req, res) => {
`📊 Recorded OpenAI usage - Input: ${totalInputTokens}(actual:${actualInputTokens}+cached:${cacheReadTokens}), Output: ${outputTokens}, Total: ${usageData.total_tokens || totalInputTokens + outputTokens}, Model: ${modelToRecord} (actual: ${actualModel}, requested: ${requestedModel})`
)
usageReported = true
await applyRateLimitTracking(
req,
{
inputTokens: actualInputTokens,
outputTokens,
cacheCreateTokens: 0,
cacheReadTokens
},
modelToRecord,
'openai-stream'
)
} catch (error) {
logger.error('Failed to record OpenAI usage:', error)
}

31
src/services/claudeRelayService.js
View File

@@ -39,17 +39,8 @@ class ClaudeRelayService {
}
// 🔍 判断是否是真实的 Claude Code 请求
isRealClaudeCodeRequest(requestBody, clientHeaders) {
// 使用 claudeCodeValidator 来进行完整的验证
// 注意：claudeCodeValidator.validate() 需要一个完整的 req 对象
// 我们需要构造一个最小化的 req 对象来满足验证器的需求
const mockReq = {
headers: clientHeaders || {},
body: requestBody,
path: '/api/v1/messages'
}
return ClaudeCodeValidator.validate(mockReq)
isRealClaudeCodeRequest(requestBody) {
return ClaudeCodeValidator.hasClaudeCodeSystemPrompt(requestBody)
}
// 🚀 转发请求到Claude API
@@ -151,8 +142,7 @@ class ClaudeRelayService {
// 获取有效的访问token
const accessToken = await claudeAccountService.getValidAccessToken(accountId)
// 处理请求体（传递 clientHeaders 以判断是否需要设置 Claude Code 系统提示词）
const processedBody = this._processRequestBody(requestBody, clientHeaders, account)
const processedBody = this._processRequestBody(requestBody, account)
// 获取代理配置
const proxyAgent = await this._getProxyAgent(accountId)
@@ -397,7 +387,7 @@ class ClaudeRelayService {
if (
clientHeaders &&
Object.keys(clientHeaders).length > 0 &&
this.isRealClaudeCodeRequest(requestBody, clientHeaders)
this.isRealClaudeCodeRequest(requestBody)
) {
await claudeCodeHeadersService.storeAccountHeaders(accountId, clientHeaders)
}
@@ -444,7 +434,7 @@ class ClaudeRelayService {
}
// 🔄 处理请求体
_processRequestBody(body, clientHeaders = {}, account = null) {
_processRequestBody(body, account = null) {
if (!body) {
return body
}
@@ -459,7 +449,7 @@ class ClaudeRelayService {
this._stripTtlFromCacheControl(processedBody)
// 判断是否是真实的 Claude Code 请求
const isRealClaudeCode = this.isRealClaudeCodeRequest(processedBody, clientHeaders)
const isRealClaudeCode = this.isRealClaudeCodeRequest(processedBody)
// 如果不是真实的 Claude Code 请求，需要设置 Claude Code 系统提示词
if (!isRealClaudeCode) {
@@ -760,7 +750,7 @@ class ClaudeRelayService {
const filteredHeaders = this._filterClientHeaders(clientHeaders)
// 判断是否是真实的 Claude Code 请求
const isRealClaudeCode = this.isRealClaudeCodeRequest(body, clientHeaders)
const isRealClaudeCode = this.isRealClaudeCodeRequest(body)
// 如果不是真实的 Claude Code 请求，需要使用从账户获取的 Claude Code headers
const finalHeaders = { ...filteredHeaders }
@@ -1007,8 +997,7 @@ class ClaudeRelayService {
// 获取有效的访问token
const accessToken = await claudeAccountService.getValidAccessToken(accountId)
// 处理请求体（传递 clientHeaders 以判断是否需要设置 Claude Code 系统提示词）
const processedBody = this._processRequestBody(requestBody, clientHeaders, account)
const processedBody = this._processRequestBody(requestBody, account)
// 获取代理配置
const proxyAgent = await this._getProxyAgent(accountId)
@@ -1065,7 +1054,7 @@ class ClaudeRelayService {
const filteredHeaders = this._filterClientHeaders(clientHeaders)
// 判断是否是真实的 Claude Code 请求
const isRealClaudeCode = this.isRealClaudeCodeRequest(body, clientHeaders)
const isRealClaudeCode = this.isRealClaudeCodeRequest(body)
// 如果不是真实的 Claude Code 请求，需要使用从账户获取的 Claude Code headers
const finalHeaders = { ...filteredHeaders }
@@ -1595,7 +1584,7 @@ class ClaudeRelayService {
if (
clientHeaders &&
Object.keys(clientHeaders).length > 0 &&
this.isRealClaudeCodeRequest(body, clientHeaders)
this.isRealClaudeCodeRequest(body)
) {
await claudeCodeHeadersService.storeAccountHeaders(accountId, clientHeaders)
}

5
src/utils/contents.js
View File

@@ -74,6 +74,11 @@ const PROMPT_DEFINITIONS = {
title: 'Claude Agent SDK System Prompt',
text: "You are a Claude agent, built on Anthropic's Claude Agent SDK."
},
claudeOtherSystemPrompt4: {
category: 'system',
title: 'Claude Code Compact System Prompt Agent SDK2',
text: "You are Claude Code, Anthropic's official CLI for Claude, running within the Claude Agent SDK."
},
claudeOtherSystemPromptCompact: {
category: 'system',
title: 'Claude Code Compact System Prompt',

71
src/utils/rateLimitHelper.js Normal file
View File

@@ -0,0 +1,71 @@
const redis = require('../models/redis')
const pricingService = require('../services/pricingService')
const CostCalculator = require('./costCalculator')
function toNumber(value) {
const num = Number(value)
return Number.isFinite(num) ? num : 0
}
async function updateRateLimitCounters(rateLimitInfo, usageSummary, model) {
if (!rateLimitInfo) {
return { totalTokens: 0, totalCost: 0 }
}
const client = redis.getClient()
if (!client) {
throw new Error('Redis 未连接，无法更新限流计数')
}
const inputTokens = toNumber(usageSummary.inputTokens)
const outputTokens = toNumber(usageSummary.outputTokens)
const cacheCreateTokens = toNumber(usageSummary.cacheCreateTokens)
const cacheReadTokens = toNumber(usageSummary.cacheReadTokens)
const totalTokens = inputTokens + outputTokens + cacheCreateTokens + cacheReadTokens
if (totalTokens > 0 && rateLimitInfo.tokenCountKey) {
await client.incrby(rateLimitInfo.tokenCountKey, Math.round(totalTokens))
}
let totalCost = 0
const usagePayload = {
input_tokens: inputTokens,
output_tokens: outputTokens,
cache_creation_input_tokens: cacheCreateTokens,
cache_read_input_tokens: cacheReadTokens
}
try {
const costInfo = pricingService.calculateCost(usagePayload, model)
const { totalCost: calculatedCost } = costInfo || {}
if (typeof calculatedCost === 'number') {
totalCost = calculatedCost
}
} catch (error) {
// 忽略此处错误，后续使用备用计算
totalCost = 0
}
if (totalCost === 0) {
try {
const fallback = CostCalculator.calculateCost(usagePayload, model)
const { costs } = fallback || {}
if (costs && typeof costs.total === 'number') {
totalCost = costs.total
}
} catch (error) {
totalCost = 0
}
}
if (totalCost > 0 && rateLimitInfo.costCountKey) {
await client.incrbyfloat(rateLimitInfo.costCountKey, totalCost)
}
return { totalTokens, totalCost }
}
module.exports = {
updateRateLimitCounters
}

11
src/validators/clients/claudeCodeValidator.js
View File

@@ -74,16 +74,7 @@ class ClaudeCodeValidator {
const userAgent = req.headers['user-agent'] || ''
const path = req.path || ''
// 1. 先检查是否是 Claude Code 的 User-Agent
// 支持的格式:
// - claude-cli/1.0.86 (external, cli) - 原有 CLI 格式
// - claude-cli/2.0.0 (external, claude-vscode) - VSCode 插件格式
// - claude-cli/x.x.x (external, sdk-py) - Python SDK 格式
// - claude-cli/x.x.x (external, sdk-js) - JavaScript SDK 格式
// - 其他 (external, claude-xxx) 或 (external, sdk-xxx) 格式
const claudeCodePattern =
/^claude-cli\/[\d.]+(?:[-\w]*)?\s+\(external,\s*(?:cli|claude-[\w-]+|sdk-[\w-]+)\)$/i
const claudeCodePattern = /^claude-cli\/\d+\.\d+\.\d+/i;
if (!claudeCodePattern.test(userAgent)) {
// 不是 Claude Code 的请求，此验证器不处理