Workshop Part 3

track usage and cost across applications you have a working amberflo unified interface and real traffic flowing through it now you’ll see how it scales in this section, you will add a second workload generate traffic from multiple applications view usage and cost attribution by workload and model this is where the governance value becomes clear add another workload workloads represent independent applications or systems go to workloads and create a new one example name marketing helper leave the generated id unchanged select available models simple model latest model save this workload now has access to the same models but will be tracked independently generate a new workload key in workload keys create a new key select marketing helper name it marketing helper key save and copy the key value remember the key determines attribution send traffic from multiple workloads use your existing test scripts or html example for marketing related prompts, use marketing helper key latest model for support related prompts, continue using customer support bot key either model alias example marketing prompts write a product description paragraph rewrite this headline to be more compelling example support prompts respond to an angry customer explain a delay in resolving an issue you are now generating traffic from multiple workloads through the same unified interface review usage and attribution return to ai spend and refresh you will now see separate usage totals per workload model level breakdown within each workload provider level aggregation cost accumulating in real time navigate to workload showbacks and select each workload individually you can see which models were used how many tokens were consumed total spend per workload all traffic flows through one interface attribution remains clean and isolated what this demonstrates you now have a single, unified entry point for ai traffic consistent api structure across providers centralized governance clear attribution by workload real time visibility into spend without this structure, ai usage fragments quickly across teams and api keys with it, you have control