Cost control of ChatGPT in Notes applications
When you integrate ChatGPT into your Notes applications, you get great opportunities to improve them. Both in terms of generated text and recommendations, and an intelligence that can get an overview of very large amounts of data.
When you integrate ChatGPT into your Notes applications, be aware that the price is usage based. You pay for the tokens you send and receive. A token is roughly four characters. That also means it pays to spend a little time on prompt engineering, that is, optimising what is sent back and forth.
At the time of writing, GPT-4o (omni) is the latest model. Its advantage is that it is the most advanced, but also that it is twice as fast as earlier models and costs half as much. You can read more about GPT-4o at OpenAI.
Ongoing calculation of ChatGPT usage
When we develop the integration between HCL Notes and ChatGPT, we also give you an overview where you can follow usage over time.
We do that through the reply ChatGPT sends back. It shows how many tokens were sent in and how many were returned. With the requests you send, you can also set a maximum number of tokens that may be used.
From the token counts we calculate what a call and a reply cost, so you know how much it costs you to analyse your data.
With our integration we strongly recommend that the integration keeps track of whether new data is being sent to ChatGPT, so you do not needlessly ask for information you already have. That could be reports generated on the same data as before, where you would expect the same report back. We control that through logic in HCL Notes that keeps track of it.
Rate limits in ChatGPT
When you get your API keys, you also set a monthly maximum on OpenAI's platform. To begin with the maximum is 120 USD, which you can set lower if you wish. As your usage grows, you can move up through what are called tiers. Higher tiers give you a higher maximum and more freedom to send large amounts of data.
Conclusion
With ChatGPT's platform you can set limits so the budget does not run away. At the same time we build a usage overview that we use to improve the prompts we send to ChatGPT, to keep the cost per call as low as possible.
We advise on use and options, and we can build a proof of concept for you using OpenAI's sandbox, an area where we can test what comes back.
You are welcome to contact Christian Gravgaard on +45 70 230 249 or by email at cg@bizperformer.com for advice on your options for integrating your Notes applications with ChatGPT.