Function invoke 

pub async fn invoke(
    client: &Client,
    base_url: &str,
    api_key: &str,
    extra_headers: &[(String, String)],
    upstream_model: &str,
    default_max_tokens: Option<u64>,
    request: CompletionRequest,
) -> Result<ModelStream, CompletionError>

Stream one completion from an OpenAI-compatible endpoint.

Errors

Returns CompletionError::provider when the endpoint rejects the request (with the provider’s retry classification) and CompletionError::ResponseError when a live stream breaks mid-flight. Individual stream items may also yield errors.