API Parameters
This page provides a complete list of all parameters supported by the Knox Chat API.
Core Parameters
| Parameter Name | Type | Default Value | Description |
|---|---|---|---|
model | String | None | The model ID to use. If not specified, the user's default setting will be used. |
messages | Array of Objects | None | An array of messages for chat completion requests. Each message should include "role" and "content" fields. |
prompt | String | None | The prompt text for text completion requests. |
Shared Parameters
The following parameters can be used for both completion and chat completion requests:
| Parameter Name | Type | Default Value | Description |
|---|---|---|---|
max_tokens | Integer | None | Maximum number of tokens to generate (range: [1, context_length)). |
temperature | Double | 1.0 | Sampling temperature (range: [0, 2]). Higher values produce more diverse results. |
top_p | Double | 1.0 | Top-p sampling value (range: (0, 1]). Controls the probability space of vocabulary considered by the model. |
top_k | Integer | None | Top-k sampling value (range: [1, Infinity)). Only considers the k most probable tokens. |
stream | Boolean | false | Enable streaming of results. |
seed | Integer | None | Seed value for deterministic output. |
frequency_penalty | Double | 0.0 | Frequency penalty (range: [-2, 2]). Decreases the probability of tokens that have been used frequently. |
presence_penalty | Double | 0.0 | Presence penalty (range: [-2, 2]). Decreases the probability of tokens that have already appeared. |
repetition_penalty | Double | 1.0 | Repetition penalty (range: (0, 2]). Decreases the probability of repeated words. |
logit_bias | Object | None | A mapping from token IDs to bias values for adjusting the probability of specific tokens. |
min_p | Double | 0.0 | Minimum probability threshold (range: [0, 1]). Filters out tokens below the threshold. |
top_a | Double | None | Alternative top sampling parameter (range: [0, 1]). |
top_logprobs | Integer | None | Number of top log probabilities to return. |
Knox Chat Specific Parameters
The following parameters are specific to Knox Chat:
| Parameter Name | Type | Default Value | Description |
|---|---|---|---|
models | Array of Strings | None | List of alternative models for routing override. |
provider | Object | None | Provider preferences for routing. |
transforms | Array of Strings | None | List of prompt transformations (Knox Chat only). |
reasoning | Object | None | Configuration for model reasoning/thinking tokens. |
usage | Object | None | Whether to include usage information in the response. |
Context Length Limitations
Each model has different context length limitations, which determine the total number of input and output tokens. You can query the model list via the /v1/models endpoint to view the context_length value for each model.