Amberflo Playground Overview

the amberflo playground lets you interact with real ai models through a simple chat interface while automatically capturing usage and cost data it is designed as the fastest way to understand how amberflo tracks ai usage, attributes it to workloads, and calculates cost in real time using the playground, you can send prompts to models from providers such as openai, azure openai, aws bedrock, and others using your own credentials every request made through the interface is captured, metered, and priced so you can immediately see how ai usage translates into cost this allows you to experiment with models and prompts while also seeing how amberflo organizes usage data across workloads and customers how the playground works when you send a prompt from the playground the request is sent to the ai provider using your configured provider credentials amberflo captures usage details such as input tokens, output tokens, request timing, and metadata the request is automatically attributed to the selected workload amberflo calculates the cost using model pricing and records the result within moments, this data becomes visible in the ai spend dashboards , allowing you to see exactly how usage is translated into cost this gives you a safe environment to explore how amberflo tracks ai usage before integrating it into applications or production workflows core concepts the playground operates using the same underlying governance and attribution model used throughout the amberflo platform four core concepts define how access is configured and how usage is tracked providers providers define how amberflo connects to external ai services for each provider you want to use, you create a credential and supply the required authentication information such as an api key or cloud credentials supported providers include openai azure openai aws bedrock google ai studio anthropic these credentials are stored securely and used when requests are sent from the playground models models represent the specific ai models you want to make available each model is associated with a provider credential and maps to a specific model offered by that provider for example gpt models from openai claude models from anthropic via bedrock other provider specific models defining models allows amberflo to track usage by model calculate costs accurately control which models workloads are allowed to use workloads workloads represent the logical consumers of ai within your system a workload might correspond to an application a product feature a team a service a customer facing capability every request sent through the playground is attributed to a workload this allows amberflo to organize usage and cost data around how ai is actually used inside your organization workloads are the primary unit of attribution for usage and spend virtual keys virtual keys control access to models through a specific workload each key is associated with a workload and inherits that workload’s allowed models when requests are made using the key, amberflo automatically attributes all usage and cost to the associated workload in the playground, you select a workload and virtual key before sending prompts this ensures that all activity generated through the playground is tracked consistently with how production usage would be tracked how it all fits together conceptually, the process looks like this you configure providers to connect to ai services such as openai or aws bedrock you define models that map to specific ai models from those providers you create workloads to represent applications, teams, or product features you generate virtual keys that control access and attribution you use the playground to send prompts and generate real ai usage every request made through the playground produces usage and cost data that is automatically attributed and visible inside amberflo what happens next after sending a few prompts through the playground, you can explore how amberflo organizes and analyzes the resulting data navigate to spend to see cost attributed to workloads trends in ai usage and spend which workloads and customers are driving ai cost this is where the full power of amberflo becomes clear the same usage signals generated by ai interactions can be used to understand cost, optimize usage, and power customer billing if you want, i can also help you create a much stronger doc structure around the playground (right now it’s still a bit conceptual) the ideal flow would be playground overview send your first prompt understanding attribution (workloads, customers) viewing cost in spend that sequence matches the actual first user experience much better