Recipes

A recipe is a combination of AI instructions. It can be described as code written in natural language that you can use as automation, part of an automation or as part of a customer facing application. A recipe is your way of defining a blueprint of AI instructions that the LLM follows to solve your specific problem.

In a recipe, instructions can be combined together to create more advanced functionality. Once created, a recipe can be deployed through an API. You can use our SDK or our Rest API to integrate the tool into your codebase. You can also share a preview of the recipe using our shareable ui feature.

A recipe can contain chained instructions. But, unless you chain the instructions by referring to the other instructions, then they the default behavior is that they run independently.

LLMs can produce even better results if they solve a specific problem at a time instead of many problems at the same time. That's why we recommend you to split your solution into several instructions and to make your instructions as precise as possible.

Default behavior of instructions in a recipe

Instructions inside a recipe are independent of each other by default, they are not part of a chat. If you want to use the output of an instruction in another instruction, you can chain them