Skip to main content

multimodal/
multimodal.rs

1use gemini_client_api::gemini::ask::Gemini;
2use gemini_client_api::gemini::types::sessions::Session;
3use gemini_client_api::gemini::utils::MarkdownToParts;
4use std::env;
5
6#[tokio::test]
7async fn raw_multimodal() {
8    use gemini_client_api::gemini::types::request::InlineData;
9
10    let mut session = Session::new(6);
11    let api_key = std::env::var("GEMINI_API_KEY").expect("GEMINI_API_KEY must be set");
12    let ai = Gemini::new(api_key, "gemini-2.5-flash", None);
13
14    session.ask("What is there in this pdf")
15        .ask(InlineData::from_url("https://bitmesra.ac.in/UploadedDocuments/admingo/files/221225_List%20of%20Holiday_2026_26.pdf").await.unwrap());
16
17    let response = ai.ask(&mut session).await.unwrap();
18    println!("\nGemini: {}", response.get_chat().get_text_no_think(""));
19}
20#[tokio::main]
21async fn main() {
22    let mut session = Session::new(6);
23    let api_key = env::var("GEMINI_API_KEY").expect("GEMINI_API_KEY must be set");
24    let ai = Gemini::new(api_key, "gemini-2.5-flash", None);
25
26    println!("--- Multimodal (Images/Files) Example ---");
27
28    // Use MarkdownToParts to easily parse a string with image/file markers
29    // It supports both URLs and local file paths!
30    let content = "Describe this image: ![image](https://www.google.com/images/branding/googlelogo/1x/googlelogo_color_272x92dp.png)";
31    println!("Processing: {}", content);
32
33    let parts = MarkdownToParts::new(content, |_| mime::IMAGE_PNG)
34        .await
35        .process();
36
37    let response = ai.ask(session.ask_parts(parts)).await.unwrap();
38
39    println!("\nGemini: {}", response.get_chat().get_text_no_think(""));
40}