Text to structured data

This mode is use-case specific. An LLM is good at understand unstructured inputs. Oftentimes, to make use of the LLMs output in a process, you require a structured output. This mode provides just that: It returns a json with a user-defined structure, based on the input.

• Model: You can select any model from OpenAI that supports the chat/completions endpoint and that you have access to. This includes GPT-4, OpenAI's newest model, however, only if you have already been granted access. More info about that supported models can be found here.

• Input: The input field takes a static or dynamic value that contains the text that should be digested by the model and turned into a json.

• Describe the input: A description of what the input is.

• Output fields: Output fields are used to describe the structure of the json that will be returned by the model. Each field has a name and a type. At the moment, we support string, number, boolean, number and array. There needs to be at least one output field.

• Examples: Language models are known to perform best if you provide a list of examples. It learns and understands the patterns and applies them to the input. For every example given, you provide an input and the output values based on the structure that was defined in the previous section. Examples are optional but strongly encouraged.