推断

🌐 Inference

通过 Mastra 的模型路由访问 9 个推断模型。身份验证会自动使用 INFERENCE_API_KEY 环境变量处理。

🌐 Access 9 Inference models through Mastra's model router. Authentication is handled automatically using the INFERENCE_API_KEY environment variable.

在推断文档中了解更多信息。

🌐 Learn more in the Inference documentation.

.env

INFERENCE_API_KEY=your-api-key

src/mastra/agents/my-agent.ts

import { Agent } from "@mastra/core/agent";

const agent = new Agent({
  id: "my-agent",
  name: "My Agent",
  instructions: "You are a helpful assistant",
  model: "inference/google/gemma-3"
});

// Generate a response
const response = await agent.generate("Hello!");

// Stream a response
const stream = await agent.stream("Tell me a story");
for await (const chunk of stream) {
  console.log(chunk);
}

info

Mastra 使用与 OpenAI 兼容的 /chat/completions 端点。某些特定于提供商的功能可能无法使用。详见 推断文档。

🌐 Mastra uses the OpenAI-compatible /chat/completions endpoint. Some provider-specific features may not be available. Check the Inference documentation for details.

模型

🌐 Models

9 available models

Model	Context	Tools	Reasoning	Image	Audio	Video	Input $/1M	Output $/1M
inference/google/gemma-3	125K						$0.15	$0.30
inference/meta/llama-3.1-8b-instruct	16K						$0.03	$0.03
inference/meta/llama-3.2-11b-vision-instruct	16K						$0.06	$0.06
inference/meta/llama-3.2-1b-instruct	16K						$0.01	$0.01
inference/meta/llama-3.2-3b-instruct	16K						$0.02	$0.02
inference/mistral/mistral-nemo-12b-instruct	16K						$0.04	$0.10
inference/osmosis/osmosis-structure-0.6b	4K						$0.10	$0.50
inference/qwen/qwen-2.5-7b-vision-instruct	125K						$0.20	$0.20
inference/qwen/qwen3-embedding-4b	32K						$0.01	—

高级配置

🌐 Advanced Configuration

自定义头

🌐 Custom Headers

src/mastra/agents/my-agent.ts

const agent = new Agent({
  id: "custom-agent",
  name: "custom-agent",
  model: {
    url: "https://inference.net/v1",
    id: "inference/google/gemma-3",
    apiKey: process.env.INFERENCE_API_KEY,
    headers: {
      "X-Custom-Header": "value"
    }
  }
});

动态模型选择

🌐 Dynamic Model Selection

src/mastra/agents/my-agent.ts

const agent = new Agent({
  id: "dynamic-agent",
  name: "Dynamic Agent",
  model: ({ requestContext }) => {
    const useAdvanced = requestContext.task === "complex";
    return useAdvanced
      ? "inference/qwen/qwen3-embedding-4b"
      : "inference/google/gemma-3";
  }
});