ollama.wit

package ollama:llm;

interface llm {
    record message {
        role: string,
        content: string
    }

    record chat-request {
        model: string,
        messages: list<message>,
        with-stream: option<bool>,
        format: string,
        options: list<tuple<string, string>>
    }

    record chat-response {
        model: string,
        created-at: u64,
        message: message,
        done: bool,
        total-duration: u64,
        load-duration: u64,
        prompt-eval-count: u32,
        prompt-eval-duration: u32,
        eval-count: u32,
        eval-duration: u32
    }

    record status-error {
        status-code: u32,
        status: string,
        error: string
    }

    record show-request {
        name: string,
        model: string,
        system: string,
        template: string,
        options: list<tuple<string, string>>
    }

    record show-response {
        license: string,
        modelfile: string,
        parameters: string,
        template: string,
        system: string,
        details: model-details
    }

    record model-details {
        format: string,
        family: string,
        families: list<string>,
        parameter-size: string,
        quantization-level: string
    }

    record model-response {
        name: string,
        modified-at: u64,
        size: u64,
        digest: string,
        details: model-details
    }

    record list-response {
        models: list<model-response>
    }

    chat: func(request: chat-request) -> result<chat-response, status-error>;

    show: func(request: show-request) -> result<show-response, status-error>;

    %list: func() -> result<list-response, status-error>;
}