Retrieve agent configuration

Determines the level of randomness in the generated output. Can take values between 0 and 1.0, with higher values like 0.8 resulting in more randomness, and lower values like 0.2 making the output more predictable. See original definition.

Only sample from the top K options for each subsequent token. Used to remove "long tail" low probability responses. Defaults to -1, which disables it. See original definition. See technical details.

An alternative to the temperature option. Uses a method called nucleus sampling, which involves considering only the tokens with the highest probability mass. It specifies the proportion of the probability mass to consider, with a value of 0.1 indicating that only the tokens comprising the top 10% probability mass will be taken into account. In essence, top_p determines the size of the set of tokens to choose from, based on their probabilities. Defaults to -1, which disables it. See original definition.

Can take values between -2.0 and 2.0, with positive values penalizing the model for generating new tokens that have already appeared frequently in the generated text so far. This reduces the likelihood of the model repeating the same line or phrase verbatim. In essence, a higher positive value for frequency_penalty encourages the model to use less frequently used tokens in the generated text. See original definition.

Can take values between -2.0 and 2.0, with positive values penalizing the model for generating new tokens that have already appeared in the generated text. This encourages the model to generate text about new topics and increases the likelihood of generating more diverse content. In essence, a higher positive value for presence_penalty will lead to more avoidance of repeating previously used tokens in the generated text. See original definition.

Determines the level of randomness in the generated output. Can take values between 0 and 2.0, with higher values like 0.8 resulting in more randomness, and lower values like 0.2 making the output more predictable. See original definition.

An alternative to the temperature option. Uses a method called nucleus sampling, which involves considering only the tokens with the highest probability mass. It specifies the proportion of the probability mass to consider, with a value of 0.1 indicating that only the tokens comprising the top 10% probability mass will be taken into account. In essence, top_p determines the size of the set of tokens to choose from, based on their probabilities. See original definition.