openai4rs
简体中文 | English
一个基于 tokio 和 reqwest 的异步 Rust crate，用于与遵循 OpenAI 规范的大模型供应商进行交互。
✨ 特性
🗨️ Chat 聊天
- ✅ 流式响应
- ✅ 工具调用
- ✅ 多轮对话
- ✅ 视觉（Vision）API（如果模型支持）
📝 Completions 文本补全 (Legacy)
- ✅ 非流式响应
- ✅ 流式响应
🤖 Models 模型管理
- ✅ 获取模型列表
- ✅ 获取单个模型信息
🔄 HTTP 请求控制
- ✅ 可配置的重试次数
- ✅ 可配置的请求超时
- ✅ 可配置的连接超时
- ✅ HTTP 代理支持
- ✅ 自定义 User-Agent
- ✅ 自定义 Headers
🚀 快速开始
安装
添加依赖到你的 Cargo.toml：
[dependencies]
openai4rs = "0.1.5"
tokio = { version = "1.45.1", features = ["full"] }
futures = "0.3.31"
dotenvy = "0.15"
或使用 cargo 命令：
cargo add openai4rs
基础使用
use dotenvy::dotenv;
use openai4rs::*;
#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
dotenv().ok();
let client = OpenAI::from_env()?;
let model = "Qwen/Qwen3-Coder-480B-A35B-Instruct";
let messages = vec![
system!(content: "You are a helpful assistant."),
user!(content: "Introduce the Rust programming language in one sentence."),
];
let request = chat_request(model, &messages);
let response = client.chat().create(request).await?;
if let Some(content) = response.content() {
println!("\nResponse:\n{}", content);
} else {
println!("\nNo content in response.");
}
Ok(())
}
📚 核心用法
🗨️ Chat 聊天
流式聊天
实时接收响应内容，适合需要逐步显示的场景：
use std::io::{self, Write};
use dotenvy::dotenv;
use futures::StreamExt;
use openai4rs::*;
#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
dotenv().ok();
let client = OpenAI::from_env()?;
let model = "Qwen/Qwen3-Coder-480B-A35B-Instruct";
let messages = vec![
system!(content: "You are a helpful assistant."),
user!(content: "Introduce the Rust programming language in one sentence."),
];
let request = chat_request(model, &messages).build()?;
println!("Sending request to model: {}...", model);
let mut stream = client.chat().create_stream(request).await?;
let mut first_content = true;
while let Some(chunk_result) = stream.next().await {
match chunk_result {
Ok(chunk) => {
if chunk.has_content() {
if first_content {
println!("\n========Response========");
first_content = false;
}
if let Some(content) = chunk.content() {
print!("{}", content);
io::stdout().flush()?;
}
}
}
Err(e) => {
eprintln!("\nAn error occurred during streaming: {}", e);
break;
}
}
}
println!();
Ok(())
}
🔧 工具调用
让模型能够调用外部工具来增强功能：
use dotenvy::dotenv;
use openai4rs::*;
use serde_json::json;
// Mock function to get weather data
fn get_current_weather(location: &str, unit: Option<&str>) -> String {
let unit = unit.unwrap_or("celsius");
format!(
"The current weather in {} is 22 degrees {}.",
location, unit
)
}
#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
dotenv().ok();
let client = OpenAI::from_env()?;
let model = "Qwen/Qwen3-Coder-480B-A35B-Instruct";
// 1. Define the tool (function)
let weather_tool_params = Parameters::object()
.property(
"location",
Parameters::string()
.description("The city and state, e.g. San Francisco, CA")
.build(),
)
.property(
"unit",
Parameters::string()
.description("The unit of temperature, e.g. celsius or fahrenheit")
.build(),
)
.require("location")
.build()?;
let weather_tool = ChatCompletionToolParam::function(
"get_current_weather",
"Get the current weather in a given location",
weather_tool_params,
);
// 2. Create the initial message and request
let messages = vec![
system!(content: "You are a helpful assistant."),
user!(content: "What's the weather like in Boston today?"),
];
let request = chat_request(model, &messages)
.tools(vec![weather_tool])
.tool_choice(ToolChoice::Auto)
.build()?;
let response = client.chat().create(request).await?;
// 3. Check if the model wants to call a tool
if response.has_tool_calls() {
let tool_calls = response.tool_calls().unwrap();
// For simplicity, we'll only handle the first tool call
if let Some(tool_call) = tool_calls.first() {
let function_name = &tool_call.function.name;
let arguments_str = &tool_call.function.arguments;
if function_name == "get_current_weather" {
let args: serde_json::Value = serde_json::from_str(arguments_str)?;
let location = args["location"].as_str().unwrap_or("Unknown");
let unit = args["unit"].as_str();
// 4. Call the function and get the result
let function_result = get_current_weather(location, unit);
// 5. Send the function result back to the model
let mut new_messages = messages.clone();
new_messages.push(response.first_choice_message().unwrap().clone().into());
new_messages.push(tool!(
tool_call_id = tool_call.function.id.clone(),
content = function_result
));
let follow_up_request = chat_request(model, &new_messages).build()?;
let final_response = client.chat().create(follow_up_request).await?;
if let Some(content) = final_response.content() {
println!("\nFinal Assistant Response:\n{}", content);
}
}
}
} else {
// If no tool call, just print the content
if let Some(content) = response.content() {
println!("\nAssistant Response:\n{}", content);
}
}
Ok(())
}
🧠 多轮对话
维护一个具有上下文的多轮对话：
use dotenvy::dotenv;
use openai4rs::*;
use std::io::{stdin, stdout, Write};
#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
dotenv().ok();
let client = OpenAI::from_env()?;
let model = "Qwen/Qwen3-Coder-480B-A35B-Instruct";
let mut messages = vec![system!(content: "You are a helpful assistant.")];
loop {
print!("You: ");
stdout().flush()?;
let mut user_input = String::new();
stdin().read_line(&mut user_input)?;
let user_input = user_input.trim();
if user_input.eq_ignore_ascii_case("exit") {
println!("Goodbye!");
break;
}
messages.push(user!(content: user_input));
let request = chat_request(model, &messages);
let response = client.chat().create(request).await?;
if let Some(content) = response.content() {
println!("Assistant: {}\n", content);
messages.push(assistant!(content));
} else {
println!("Assistant: No response.\n");
}
}
Ok(())
}
🔧 高级配置
客户端配置
use dotenvy::dotenv;
use openai4rs::*;
use std::collections::HashMap;
#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
dotenv().ok();
// Get the API key and base URL from the environment
let api_key = std::env::var("OPENAI_API_KEY")?;
let base_url = std::env::var("OPENAI_BASE_URL")?;
// 1. Basic client with default settings
let basic_client = OpenAI::new(&api_key, &base_url);
// 2. Client with a custom base URL (e.g., for a proxy or a different provider)
let _custom_base_url_client = Config::builder()
.api_key(api_key.clone())
.base_url(base_url.clone()) // Replace with your custom base URL
.build_openai()?;
// 3. Client with a proxy
let _proxy_client = Config::builder()
.api_key(api_key.clone())
.base_url(base_url.clone())
.proxy("http://proxy.example.com:8080".to_string()) // Replace with your proxy URL
.build_openai()?;
// 4. Client with custom headers
let mut custom_headers = HashMap::new();
custom_headers.insert("X-Custom-Header".to_string(), "custom_value".into());
let _custom_headers_client = Config::builder()
.api_key(api_key.clone())
.base_url(base_url.clone())
.extra_headers(custom_headers)
.build_openai()?;
// 5. Client with custom timeout
let _timeout_client = Config::builder()
.api_key(api_key)
.base_url(base_url.clone())
.timeout_seconds(120) // 2 minutes
.build_openai()?;
Ok(())
}
📖 运行示例
查看 examples 目录获取更多使用示例：
- 01. 基础聊天
- 02. 流式响应
- 03. 多轮对话
- 04. 工具调用
- 05. 客户端配置
- 06. 视觉（Vision）API (如果模型支持)
- 07. 思维模型（Thinking Model） (如果模型支持复杂推理)
你可以通过以下命令运行示例：
# 设置环境变量
export OPENAI_API_KEY=your_api_key
export OPENAI_BASE_URL=your_base_url # 可选, 默认为 https://api.openai.com/v1
# 运行示例
cargo run --example 01_simple_chat
cargo run --example 02_streaming_chat
# ... 其他示例
📄 许可证
本项目采用 Apache-2.0 许可证。