OpenAI-compatible models

Use any OpenAI-compatible endpoint with Kern agents.

Many providers support the OpenAI API format. Use the OpenAILike model to access them by replacing the base_url.

Example

1from os import getenv
2from kern.agent import Agent
3from kern.models.openai.like import OpenAILike
4
5agent = Agent(
6    model=OpenAILike(
7        id="mistralai/Mixtral-8x7B-Instruct-v0.1",
8        api_key=getenv("TOGETHER_API_KEY"),
9        base_url="https://api.together.xyz/v1",
10    )
11)
12
13# Print the response in the terminal
14agent.print_response("Share a 2 sentence horror story.")

Parameters

Parameter	Type	Default	Description
`id`	`str`	`"not-provided"`	The id of the model to use
`name`	`str`	`"OpenAILike"`	The name of the model
`provider`	`str`	`"OpenAILike"`	The provider of the model
`api_key`	`Optional[str]`	`"not-provided"`	The API key for authentication
`base_url`	`Optional[str]`	`None`	The base URL for the API service
`collect_metrics_on_completion`	`bool`	`False`	Collect token metrics only from the final streaming chunk (for providers with cumulative token counts)

OpenAILike extends the OpenAI-compatible interface and supports all parameters from OpenAIChat. Simply change the base_url and api_key to point to your preferred OpenAI-compatible service.

Responses API

For providers that implement the Open Responses API specification, use OpenResponses:

1from kern.agent import Agent
2from kern.models.openai import OpenResponses
3
4agent = Agent(
5    model=OpenResponses(
6        id="your-model-id",
7        base_url="https://your-provider.com/v1",
8        api_key="your-api-key",
9    ),
10)
11
12agent.print_response("Share a 2 sentence horror story.")

The Responses API is stateless by default. Each request is independent with no previous_response_id chaining.

For specific providers, use the dedicated classes:

OllamaResponses for Ollama (v0.13.3+)
OpenRouterResponses for OpenRouter

See OpenResponses reference for full parameters.