🚀 TSDIAPI DeepSeek Plugin

A TSDIAPI plugin that provides seamless integration with DeepSeek's AI models, enabling structured JSON output using TypeBox schemas and flexible chat interactions.

📦 Installation

npm install --save @tsdiapi/deepseek @sinclair/typebox

Then register the plugin in your TSDIAPI app:

import { createApp } from "@tsdiapi/server";
import createPlugin from "@tsdiapi/deepseek";

createApp({
  plugins: [
    createPlugin({
      apiKey: "your-deepseek-api-key", // Required
      model: "deepseek-chat",          // Optional (default)
    }),
  ],
});

🚀 Features

• 📘 Structured JSON output with TypeBox
• 🧠 Model-agnostic: Supports various DeepSeek models
• ✨ Prompt-to-structured-object in one call
• ⚡ Simple chat and typed response support
• 🧩 Composable schemas and type inference

🔧 Configuration Options

| Option | Type | Default | ENV Variable | Description | |----------|----------|-----------------|-------------------|-----------------------------| | apiKey | string | "" | DEEPSEEK_API_KEY| DeepSeek API Key (Required) | | model | string | "deepseek-chat"| DEEPSEEK_MODEL_ID | Default DeepSeek model |

📌 How to Use

✅ Structured JSON Output (via TypeBox)

import { Type } from "@sinclair/typebox";
import { useDeepSeekProvider } from "@tsdiapi/deepseek";

const UserSchema = Type.Object({
  name: Type.String(),
  email: Type.String({ format: "email" }),
});

async function run() {
  const deepseek = useDeepSeekProvider();
  const response = await deepseek.jsonDTO("Generate a user", UserSchema);
  console.log(response?.result);
}

🧠 You get a strongly typed result conforming to UserSchema, with automatic type casting (no validation errors thrown).

💬 Basic Chat Completion

import { useDeepSeekProvider } from "@tsdiapi/deepseek";

async function run() {
  const deepseek = useDeepSeekProvider();
  const response = await deepseek.chat("Tell me a dad joke.");
  console.log(response?.result);
}

🧾 Raw JSON Schema

You can also pass a plain JSON Schema manually:

const schema = {
  type: "object",
  properties: {
    title: { type: "string" },
    description: { type: "string" },
  },
  required: ["title", "description"],
};

const deepseek = useDeepSeekProvider();
const response = await deepseek.JsonString("Describe a new product", schema);
console.log(response);

📦 Example Response

{
  "message": {
    "role": "assistant",
    "content": "{ \"name\": \"Jane\", \"email\": \"[email protected]\" }"
  },
  "usage": {
    "prompt_tokens": 20,
    "completion_tokens": 15,
    "total_tokens": 35
  },
  "result": {
    "name": "Jane",
    "email": "[email protected]"
  }
}

🔌 Related Plugins

Check out more plugins in the TSDIAPI ecosystem:

👉 Explore @tsdiapi plugins on npm

👨‍💻 Contributing

Have ideas? Want to add more features? PRs and issues are welcome!

Author: unbywyd
📧 Contact: [email protected]

🚀 Happy coding with TSDIAPI DeepSeek Plugin!