Function llm_chat_vision_with_usage 

pub async fn llm_chat_vision_with_usage(
    llm: &LlmConfig,
    system: &str,
    user: &str,
    image_data_urls: Option<&[String]>,
) -> Result<LlmResponse, String>

Sends a vision-enabled chat completion request.

The user message carries the text prompt and one or more screenshots (JPEG/PNG data URLs, ordered from the top of the page down) as OpenAI-compatible image_url content parts.

Retries and error reporting behave like llm_chat_with_usage.

Errors

Returns the last underlying error as a human-readable string when every attempt fails (transport error, non-success HTTP status, response that is not valid JSON, or a response missing choices[0].message.content).