DeepSeek

This guide provides instructions for configuring and using DeepSeek models within the AIGNE Framework via the @aigne/deepseek package. It covers API key setup, model instantiation, and examples for both standard and streaming responses.

The @aigne/deepseek package provides a direct integration with the DeepSeek API, leveraging its powerful language models. It is designed to be compatible with the AIGNE Framework's ChatModel interface, ensuring a consistent development experience.

Installation#

To begin, install the necessary packages using your preferred package manager. The @aigne/core package is a required peer dependency.

npm install @aigne/deepseek @aigne/core

yarn add @aigne/deepseek @aigne/core

pnpm add @aigne/deepseek @aigne/core

Configuration#

The DeepSeekChatModel class is the primary interface for interacting with DeepSeek models. It extends the OpenAIChatModel and is configured to use DeepSeek's specific API endpoints and authentication methods.

Your DeepSeek API key is required for authentication. You can provide it in two ways:

1. Directly in the constructor: Pass the key via the apiKey property.
2. Environment variable: Set the DEEPSEEK_API_KEY environment variable. The model will automatically use it if the apiKey property is not provided.

Parameters#

Usage#

Once configured, the model can be used to generate text completions or stream responses.

Basic Invocation#

To generate a standard response, use the invoke method. Provide a list of messages, and the method will return a promise that resolves with the complete response from the model.

import { DeepSeekChatModel } from "@aigne/deepseek";

const model = new DeepSeekChatModel({
  // Provide API key directly or use environment variable DEEPSEEK_API_KEY
  apiKey: "your-api-key", // Optional if set in env variables
  // Specify model version (defaults to 'deepseek-chat')
  model: "deepseek-chat",
  modelOptions: {
    temperature: 0.7,
  },
});

const result = await model.invoke({
  messages: [{ role: "user", content: "Introduce yourself" }],
});

console.log(result);

The result object contains the generated text and metadata about the model's usage.

Example Response

{
  "text": "Hello! I'm an AI assistant powered by DeepSeek's language model.",
  "model": "deepseek-chat",
  "usage": {
    "inputTokens": 7,
    "outputTokens": 12
  }
}

Streaming Responses#

For real-time applications, you can stream the response from the model. Set the streaming option to true in the invoke method's second argument. This returns an async iterator that yields chunks of the response as they become available.

import { isAgentResponseDelta } from "@aigne/core";
import { DeepSeekChatModel } from "@aigne/deepseek";

const model = new DeepSeekChatModel({
  apiKey: "your-api-key",
  model: "deepseek-chat",
});

const stream = await model.invoke(
  {
    messages: [{ role: "user", content: "Introduce yourself" }],
  },
  { streaming: true },
);

let fullText = "";
const json = {};

for await (const chunk of stream) {
  if (isAgentResponseDelta(chunk)) {
    const text = chunk.delta.text?.text;
    if (text) fullText += text;
    if (chunk.delta.json) Object.assign(json, chunk.delta.json);
  }
}

In this example, the code iterates through the stream, accumulating the text delta from each chunk to build the complete response. Final metadata, such as token usage, is provided in the last chunk.

Summary#

This guide has covered the essential steps to install, configure, and use DeepSeek models with the AIGNE Framework. By following these instructions, you can integrate DeepSeek's chat capabilities into your applications for both single-turn and streaming use cases. For more advanced configurations and features, refer to the API reference and other sections in the documentation.