Interface ChatHerokuFields

The model ID to use for completion (e.g., "gpt-oss-120b"). As specified in Heroku API documentation. If not provided, defaults to process.env.INFERENCE_MODEL_ID.

Controls randomness. Lower values make responses more focused. Parameter from Heroku API.

Maximum tokens the model may generate. Maps to max_tokens in Heroku API.

List of strings that stop generation. Parameter from Heroku API.

Whether to stream responses. If true, invoke will still return a complete response. Used by stream(). Heroku API parameter.

Proportion of tokens to consider (cumulative probability). Maps to top_p in Heroku API.

Heroku Inference API Key (INFERENCE_KEY). If not provided, the library will check the environment variable INFERENCE_KEY. Used for authentication.

Heroku Inference API Base URL (INFERENCE_URL). If not provided, checks env var INFERENCE_URL or uses a sensible Heroku default. The endpoint path is /v1/chat/completions.

Maximum number of retries for failed requests. Standard LangChain parameter for resilience.

Timeout for API requests in milliseconds. Standard LangChain parameter for request duration control.

Alias for stream for consistency. Sets default for internal _generate method's streaming behavior.

Allows passing other Heroku-specific parameters not explicitly defined (e.g., extended_thinking). Provides flexibility for future Heroku API additions or less common parameters.