The Crowdin API adds a List AI Request Logs method for auditing AI usage:
- crowdin.com:
GET /users/{userId}/ai/request-logs
- Crowdin Enterprise:
GET /ai/request-logs
- operationId:
api.ai.requestLogs.getMany
Optional query filters: requestId, projectId, userId, aiProviderId, model, sourceAction (enum: ai_proxy, ai_gateway, ai_translate_strings, ai_file_translate, ai_prompt_completion, pre_translate:manual, pre_translate:workflow, ai_alignment, qa_check, ai_suggestion, advisor), promptAction, statuses (comma-separated: pending, success, error, timeout), systemCredentials (boolean), isAutoTriggered (boolean), tokenName, oauthClientId, createdAfter/createdBefore (date-time), plus limit/offset.
The response model (AiRequestLog) includes: id, requestId, createdAt, status, httpStatus, model, sourceAction, promptAction, systemCredentials, isAutoTriggered, durationMs, inputTokens, outputTokens, totalCost, userId, projectId, promptId, aiProviderId, tokenName, oauthClientId, oauthClientName, ip, userAgent, and error.
Client libraries should add this method so consumers can monitor AI request activity and costs via the SDK.
References:
The Crowdin API adds a List AI Request Logs method for auditing AI usage:
GET /users/{userId}/ai/request-logsGET /ai/request-logsapi.ai.requestLogs.getManyOptional query filters:
requestId,projectId,userId,aiProviderId,model,sourceAction(enum:ai_proxy,ai_gateway,ai_translate_strings,ai_file_translate,ai_prompt_completion,pre_translate:manual,pre_translate:workflow,ai_alignment,qa_check,ai_suggestion,advisor),promptAction,statuses(comma-separated:pending,success,error,timeout),systemCredentials(boolean),isAutoTriggered(boolean),tokenName,oauthClientId,createdAfter/createdBefore(date-time), pluslimit/offset.The response model (
AiRequestLog) includes:id,requestId,createdAt,status,httpStatus,model,sourceAction,promptAction,systemCredentials,isAutoTriggered,durationMs,inputTokens,outputTokens,totalCost,userId,projectId,promptId,aiProviderId,tokenName,oauthClientId,oauthClientName,ip,userAgent, anderror.Client libraries should add this method so consumers can monitor AI request activity and costs via the SDK.
References: