gemini-rust

A Rust client library for Google's Gemini 2.0 API.

Features

Complete implementation of the Gemini 2.0 API
Support for system prompts, user prompts
Tools and function calling (including Google Search)
Streaming responses
Async/await API

Installation

Add this to your Cargo.toml:

[dependencies]
gemini-rust = "0.1.0"

Usage

Basic Usage

use gemini_rust::{Gemini, Message, Role, Content};
use tokio;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let api_key = std::env::var("GEMINI_API_KEY")?;
    let client = Gemini::new(&api_key);
    
    let response = client.generate_content()
        .with_system_prompt("You are a helpful assistant.")
        .with_user_message("Hello, how are you?")
        .execute()
        .await?;
    
    println!("Response: {}", response.text());
    
    Ok(())
}

Using Google Search Tool

use gemini_rust::{Gemini, Tool};
use tokio;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let api_key = std::env::var("GEMINI_API_KEY")?;
    let client = Gemini::new(&api_key);
    
    // Create a Google Search tool
    let google_search_tool = Tool::google_search();
    
    let response = client.generate_content()
        .with_user_message("What is the current Google stock price?")
        .with_tool(google_search_tool)
        .execute()
        .await?;
    
    println!("Response: {}", response.text());
    
    Ok(())
}

Text Embedding

use gemini_rust::{Gemini, TaskType};
use tokio;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {    
    let api_key = std::env::var("GEMINI_API_KEY")?;

    // Create client with the default model (gemini-2.0-flash)
    let client = Gemini::with_model(api_key, "models/text-embedding-004".to_string());

    println!("Sending embedding request to Gemini API...");

    // Simple text embedding
    let response = client
        .embed_content()
        .with_text("Hello, this is my text to embed")
        .with_task_type(TaskType::RetrievalDocument)
        .execute()
        .await?;

    println!("Response: {:?}", response.embedding.values);

    Ok(())
}

Configurable Base URL

You can configure a custom base URL for the Gemini API client if you need to use a different endpoint:

use gemini_rust::Gemini;
use tokio;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let api_key = std::env::var("GEMINI_API_KEY")?;
    
    // Using custom base URL
    let client = Gemini::with_base_url(api_key.clone(), "https://custom-api.example.com/v1/".to_string());
    
    // Or with both custom model and base URL
    let client = Gemini::with_model_and_base_url(
        api_key,
        "models/gemini-pro".to_string(),
        "https://custom-api.example.com/v1/".to_string()
    );
    
    let response = client.generate_content()
        .with_user_message("Hello, how are you?")
        .execute()
        .await?;
    
    println!("Response: {}", response.text());
    
    Ok(())
}

Documentation

For more examples and detailed documentation, see docs.rs.

License

MIT

gemini-rust 0.6.0