聊天与流式处理

本节展示核心API:chat_completion、流式变体、取消、批处理、快速助手和模型列表。请始终对照docs.rs上的crate版本进行验证。

基本聊天完成

直接提供商

use ai_lib::prelude::*;

#[tokio::main]
async fn main() -> Result<(), AiLibError> {
    let client = AiClient::new(Provider::OpenAI)?;
    let req = ChatCompletionRequest::new(
        "gpt-4o".to_string(),
        vec![Message { 
            role: Role::User, 
            content: Content::from_text("简洁地总结Rust所有权。".to_string()), 
            function_call: None 
        }]
    );
    let resp = client.chat_completion(req).await?;
    if let Some(first) = resp.choices.first() {
        println!("回答: {}", first.message.content.as_text());
    }
    Ok(())
}

网关型提供商

使用OpenRouter等网关时,需要使用provider/model格式的模型名:

use ai_lib::prelude::*;

#[tokio::main]
async fn main() -> Result<(), AiLibError> {
    let client = AiClient::new(Provider::OpenRouter)?;
    let req = ChatCompletionRequest::new(
        "openai/gpt-4o-mini".to_string(), // 注意使用provider前缀
        vec![Message { 
            role: Role::User, 
            content: Content::from_text("简洁地总结Rust所有权。".to_string()), 
            function_call: None 
        }]
    );
    let resp = client.chat_completion(req).await?;
    if let Some(first) = resp.choices.first() {
        println!("回答: {}", first.message.content.as_text());
    }
    Ok(())
}

流式令牌

方法假设:chat_completion_stream(request)返回Result<ChatCompletionChunk, AiLibError>的异步流。

use ai_lib::prelude::*;
use futures::StreamExt;

#[tokio::main]
async fn main() -> Result<(), AiLibError> {
    let client = AiClient::new(Provider::Groq)?;
    let req = ChatCompletionRequest::new(
        "llama3-8b-8192".to_string(),
        vec![Message { 
            role: Role::User, 
            content: Content::from_text("流式输出一首关于并发的俳句。".to_string()), 
            function_call: None 
        }]
    );
    let mut stream = client.chat_completion_stream(req).await?;
    while let Some(chunk) = stream.next().await {
        match chunk {
            Ok(c) => {
                if let Some(choice) = c.choices.first() {
                    if let Some(content) = &choice.delta.content {
                        print!("{}", content);
                    }
                }
            }
            Err(e) => { 
                eprintln!("流式错误: {e}"); 
                break; 
            }
        }
    }
    Ok(())
}

流式处理 + 取消

假设助手:chat_completion_stream_with_cancel(req)(impl Stream, CancelHandle)

use ai_lib::prelude::*;
use futures::StreamExt;
use tokio::time::{sleep, Duration};

#[tokio::main]
async fn main() -> Result<(), AiLibError> {
    let client = AiClient::new(Provider::OpenAI)?;
    let req = ChatCompletionRequest::new(
        "gpt-4o".to_string(),
        vec![Message { 
            role: Role::User, 
            content: Content::from_text("慢慢解释借用检查器。".to_string()), 
            function_call: None 
        }]
    );
    let (mut stream, handle) = client.chat_completion_stream_with_cancel(req).await?;
    tokio::select! {
        _ = async {
            while let Some(chunk) = stream.next().await {
                if let Ok(c) = chunk { /* print!("{}", c.delta_text()); */ }
            }
        } => {},
        _ = sleep(Duration::from_millis(400)) => {
            handle.cancel();
            eprintln!("400毫秒后取消");
        }
    }
    Ok(())
}

批处理请求

两种模式(假设名称):

  1. chat_completion_batch(Vec<ChatCompletionRequest>) – 并发执行,返回结果向量。
  2. chat_completion_batch_smart – 可能应用内部启发式/路由。
use ai_lib::{AiClient, Provider, ChatCompletionRequest, Message, Role};
use ai_lib::types::common::Content;

fn prompt(p: &str) -> ChatCompletionRequest {
    ChatCompletionRequest::new(
        "gpt-4o".to_string(),
        vec![Message { 
            role: Role::User, 
            content: Content::from_text(p.to_string()), 
            function_call: None 
        }]
    )
}

#[tokio::main]
async fn main() -> Result<(), AiLibError> {
    let client = AiClient::new(Provider::OpenAI)?;
    let batch = vec![
        prompt("定义RAII"), 
        prompt("用一句话说明生命周期"), 
        prompt("解释Send vs Sync")
    ];
    let results = client.chat_completion_batch(batch, None).await?;
    for (i, r) in results.iter().enumerate() {
        if let Ok(response) = r {
            if let Some(c) = response.choices.first() { 
                println!("{}: {}", i, c.message.content.as_text()); 
            }
        }
    }
    Ok(())
}

如果存在更智能的变体:

// let results = client.chat_completion_batch_smart(batch).await?;

快速助手

一些crate暴露了符合人体工程学的快捷方式,如quick_chat_text(model, prompt)返回String

// let text = client.quick_chat_text("gpt-4o", "什么是所有权?").await?;
// println!("{text}");

列出模型

// let models = client.list_models().await?;
// for m in models { println!("{}", m.name); }

多模态内容

ai-lib支持文本、图像和音频内容:

use ai_lib::prelude::*;

// 文本消息
let text_msg = Message {
    role: Role::User,
    content: Content::from_text("描述这张图片".to_string()),
    function_call: None,
};

// 图像消息(从文件)
let image_msg = Message {
    role: Role::User,
    content: Content::from_image_file("path/to/image.jpg"),
    function_call: None,
};

// 图像消息(从URL)
let image_url_msg = Message {
    role: Role::User,
    content: Content::new_image(
        Some("https://example.com/image.jpg".to_string()),
        Some("image/jpeg".to_string()),
        Some("image.jpg".to_string()),
    ),
    function_call: None,
};

// 音频消息(从文件)
let audio_msg = Message {
    role: Role::User,
    content: Content::from_audio_file("path/to/audio.mp3"),
    function_call: None,
};

// 音频消息(从URL)
let audio_url_msg = Message {
    role: Role::User,
    content: Content::new_audio(
        Some("https://example.com/audio.mp3".to_string()),
        Some("audio/mpeg".to_string()),
    ),
    function_call: None,
};

错误处理

处理不同类型的错误:

match client.chat_completion(req).await {
    Ok(response) => {
        if let Some(first) = response.choices.first() {
            println!("成功: {}", first.message.content.as_text());
        }
    }
    Err(e) if e.is_retryable() => {
        // 处理可重试错误(网络、速率限制)
        println!("可重试错误: {}", e);
        // 实现重试逻辑
    }
    Err(e) => {
        // 处理永久错误(认证、无效请求)
        println!("永久错误: {}", e);
    }
}

性能优化

连接池配置

use ai_lib::{AiClient, Provider, ConnectionOptions};

let client = AiClient::with_options(
    Provider::Groq,
    ConnectionOptions {
        // 配置连接池大小
        pool_size: Some(16),
        // 设置空闲超时
        idle_timeout: Some(Duration::from_secs(30)),
        ..Default::default()
    }
)?;

并发控制

use tokio::sync::Semaphore;

let semaphore = Arc::new(Semaphore::new(10)); // 限制并发数为10

for request in requests {
    let permit = semaphore.clone().acquire_owned().await?;
    let client = client.clone();
    
    tokio::spawn(async move {
        let _permit = permit;
        let result = client.chat_completion(request).await;
        // 处理结果
    });
}

注意事项

提示:

  • 检查docs.rs是否有任何重命名(例如chat vs chat_completion)。
  • 如果需要最终答案,将流式增量收集到String中。
  • 批处理+流式处理一起使用?启动多个chat_completion_stream任务并聚合。
  • 更多模式:高级示例
Build: b635c6a · 2025-09-17T16:29:53.012Z · v0.21