To ensure optimal outcomes from the model, we acknowledge the limitations of our current version. Please note that we are actively working on iterating and improving the model, contact us if any of them blocks your use case:
- Nested function calling: This refers to scenarios where the result of one function depends on another, for example, getWeather(getCurrentLocation()). The model does not currently specialize in handling nested function calls.
- Number of functions: We optimized the model for up to 10 functions in the
"tool_call"
, expect some performance degrade if there are more than 10 functions.- Deeply Nested Argument Schema: We have optimized the model for up to one layer of nesting in the arguments. This means if your arguments include an object or dictionary, we recommend that its value be of a scalar type (number, string, boolean, etc.). Expect some performance degradation if the arguments contain objects within objects.
- Set a very low temperature as possible for the best performance.
tools
parameter and optionally set the tool_choice
parameter to set the mode.tools
parametertool_choice
parameter"auto"
: This means the model can choose between generating a message or calling a function."none"
: The model will not call a function and will instead generate a message."any"
: The model is compelled to trigger functions, even when it may not be relevant. In such cases, the most relevant function available will be activated.