I’m getting sick of all large language models. Ouch, sorry, what is wrong with you? It’s not wrong with me, it’s Richard GPD. And all these models, they are not stable at all in their output. For example, I’m trying hard to get a structured output out of these models, like JSON format, but each time it keeps adding some random stuff to the answer, like extra commas, dots, and quotation marks. I’m done, man. There’s no way to get a stable output out of these models. But you don’t need to manually keep changing your prompt. There are some solutions developed that will help you to get a structured output from these large language models. And you don’t need to manually parse the output of these models. In this article, we will explore two options: using open AI functions and using parsing with the pedantic library. Open AI functions allow you to call functions to extract specific entities from the response of large language models. This is useful when the models natively support it. On the other hand, parsing with pedantic allows you to enforce a structured output by specifying a schema and validating the output against it. Both options provide a way to get structured output from large language models, eliminating the need for manual parsing and ensuring stability in the output. By using these techniques, you can easily extract the information you need and use it in your applications or tools. So, if you’re tired of unstructured output from large language models, give these options a try and enjoy the benefits of structured data.
