Merge pull request #4128 from zuiho-kai/fix/claude-stream-usage-overwrite

fix: Claude 流式断流时不再整份覆盖 usage，保留 cache 计费字段

relay/channel/claude/relay-claude.go

@@ -809,7 +809,16 @@ func HandleStreamFinalResponse(c *gin.Context, info *relaycommon.RelayInfo, clau
 		if common.DebugEnabled {
 			common.SysLog("claude response usage is not complete, maybe upstream error")
 		}
-		claudeInfo.Usage = service.ResponseText2Usage(c, claudeInfo.ResponseText.String(), info.UpstreamModelName, claudeInfo.Usage.PromptTokens)
+		// 只补缺失字段，不整份覆盖——保留 message_start 已拿到的 cache 字段
+		fallback := service.ResponseText2Usage(c, claudeInfo.ResponseText.String(), info.UpstreamModelName, info.GetEstimatePromptTokens())
+		if claudeInfo.Usage.CompletionTokens == 0 ||
+			(!claudeInfo.Done && fallback.CompletionTokens > claudeInfo.Usage.CompletionTokens) {
+			claudeInfo.Usage.CompletionTokens = fallback.CompletionTokens
+		}
+		if claudeInfo.Usage.PromptTokens == 0 {
+			claudeInfo.Usage.PromptTokens = fallback.PromptTokens
+		}
+		claudeInfo.Usage.TotalTokens = claudeInfo.Usage.PromptTokens + claudeInfo.Usage.CompletionTokens
 	}
 	if claudeInfo.Usage != nil {
 		claudeInfo.Usage.UsageSemantic = "anthropic"