Build DeepSeek‑Powered AI Apps on Tencent Cloud in Minutes

Introduction

On February 8, Tencent Cloud released API endpoints for DeepSeek‑R1 and DeepSeek‑V3 models, allowing users to access these large language models through stable public‑cloud services.

Knowledge Engine × DeepSeek

The Tencent Cloud Large‑Model Knowledge Engine is an enterprise‑focused platform that combines proprietary data with large‑model capabilities to quickly build agents, Retrieval‑Augmented Generation (RAG) workflows, and other AI applications. The engine ships with DeepSeek‑R1 and DeepSeek‑V3 pre‑installed, enabling minute‑level deployment of connected AI services via API.

Creating a DeepSeek Application

DeepSeek application creation guide: https://cloud.tencent.com/document/product/1759/116006

In the application configuration page, select Model Settings > Generate Model and choose DeepSeek‑R1/V3 as the model.

After configuration, test the dialogue in the built‑in conversation window to verify the connected mode.

Large‑Model API Integration

The Knowledge Engine’s atomic capability for model dialogue follows the OpenAI API specification, so existing OpenAI SDKs can call the endpoint directly.

Replace the base_url and api_key in your SDK configuration with the values below; no further code changes are required.

API endpoint: https://api.lkeap.cloud.tencent.com/v1

API KEY Management

In the console, go to Immediate Access > Management and click Create API KEY . After creation, you can add, view, or delete keys in the API KEY Management section.

Webman AI Deep Integration with DeepSeek

Access the Admin Backend

http://example.com/app/admin

Add DeepSeek Model to AI Settings

Name: Tencent Cloud × DeepSeek

Type: deepseek

Model handler: plugin\ai\app\handler\Gpt (default)

Supported models:

deepseek-r1
deepseek-v3

DeepSeek‑V3 (model parameter deepseek‑v3) – a 671B MoE model excelling in encyclopedia knowledge and mathematical reasoning, supporting 64K context and up to 8K output.

DeepSeek‑R1 (model parameter deepseek‑r1) – a 671B model trained with reinforcement learning, capable of chain‑of‑thought reasoning up to tens of thousands of tokens, also supporting 64K context and 8K output.

API: https://api.lkeap.cloud.tencent.com API key: sk-xxxx (obtain from the API key management page)

Configure the External Model List

Add DeepSeek models

{
  "deepseek-r1": "deepseek-r1",
  "deepseek-v3": "deepseek-v3"
}

Add a Role

Role name: Tencent Cloud × DeepSeek

Dialogue Debugging

Test address: http://example.com/app/ai