Rust开源AI工作流

Quick Start

Add to your Cargo.toml:

[dependencies]
chat-rs = { version = "0.2.0", features = ["openai"] }
tokio = { version = "1", features = ["macros", "rt-multi-thread"] }

use chat_rs::{ChatBuilder, openai::OpenAIBuilder, types::messages};

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    let client = OpenAIBuilder::new().with_model("gpt-4o-mini").build();
    let mut chat = ChatBuilder::new().with_model(client).build();

    let mut messages = messages::from_user(vec!["Hey there!"]);
    let res = chat.complete(&mut messages).await?;
    println!("{:?}", res.content);

    Ok(())
}

Set your API key via environment variable (OPENAI_API_KEY, GEMINI_API_KEY, or CLAUDE_API_KEY), or pass it explicitly with .with_api_key().

Examples

Run examples with the appropriate feature flags:

```bash

OpenAI-Compatible Endpoints

For any server speaking the OpenAI Chat Completions wire spec (vLLM, llama.cpp's llama-server, LiteLLM, etc.), use ChatCompletionsBuilder directly:

use chat_rs::completions::ChatCompletionsBuilder;

let client = ChatCompletionsBuilder::new()
    .with_base_url("http://localhost:8000/v1")
    .with_model("my-model")
    .with_api_key("sk-...")  // optional — omit for servers that don't require auth
    .build();

Dedicated wrappers preset URL/env-var/auth for popular targets:

• Ollama — OllamaBuilder defaults to http://localhost:11434/v1, honors OLLAMA_HOST, supports .pull() to fetch a model via Ollama's native API.
• Hugging Face Router — HuggingFaceBuilder defaults to https://router.huggingface.co/v1, reads HF_TOKEN.
• Cerebras — CerebrasBuilder defaults to https://api.cerebras.ai/v1, reads CEREBRAS_API_KEY.

OpenAIBuilder::with_custom_url() also exists for endpoints implementing the OpenAI Responses API (POST /responses), which is a different wire format from Chat Completions.