scenarios/workload-genai/policies/fragments/rate-limiting/rate-limiting-workaround.xml

<fragment>
    <!-- This is an alternate approach for scenarios where the `azure-openai-token-limit` policy can't be used 
    ex. DallE Endpoints where the image size from the request can be used to increment the counter.
    Non Azure openAI endpoints, where the response structure containing the token is different.-->

    <!-- Rate limit configuration for global tokens. 
         Allows 500 tokens per 60 seconds. 
         The count is incremented if the response status code is between 200 and 400 by any request -->
    
    <rate-limit-by-key calls="500"
        renewal-period="60"
        counter-key="GlobalTokensLimit"
        increment-condition="@(context.Response.StatusCode >= 200 && context.Response.StatusCode < 400)"
        increment-count="@(context.Response.Body.As<JObject>(true).SelectToken("usage.total_tokens").ToObject<int>())"
        remaining-calls-variable-name="globalRemainingTokens"
        remaining-calls-header-name="x-apim-global-remaining-tokens"/>
</fragment>