Build an AI Super App with DeepSeek and Tencent Cloud Code Assistant in Minutes

Overview

Tencent Cloud AI Code Assistant can be extended to use DeepSeek large language models, providing intelligent code‑related Q&A, real‑time search and code completion directly inside the IDE.

Method 1: Connect via DeepSeek public API

Register a DeepSeek account at https://platform.deepseek.com (WeChat QR code or phone number).

Create an API key in the DeepSeek console and give it a memorable name.

Store the API key securely – it can be viewed only once.

Configure the key in AI Code Assistant :

Open the technical dialogue panel.

In the model selector (default hunyuan‑turbo), choose DeepSeek R1 or another DeepSeek model.

Click the Configure button next to DeepSeek, paste the API key, select the desired model, and save.

After saving, the assistant will route requests to DeepSeek.

Method 2: Use a locally‑deployed Ollama + DeepSeek R1

Hardware requirements (Mac example) :

CPU: Apple M4 or newer

RAM: 16 GB (more improves performance)

Disk: ≥30 GB for model files

OS: macOS Ventura or later

Step 1 – Install Ollama (download from https://ollama.com).

Step 2 – Pull the desired DeepSeek R1 model . Example for the 7B version: ollama run deepseek-r1:7b Available model sizes: 1.5B, 7B, 8B, 14B, 32B, 70B, 671B.

Step 3 – Start the Ollama service (default port 11434):

# Create log directory
mkdir -p /root/.ollama/logs
# Run service in background
PORT=11434 && nohup ollama serve > /root/.ollama/logs/server.log 2>&1 &

Verify the service is reachable at http://localhost:11434.

Step 4 – Configure the assistant to use the local model :

Open the technical dialogue panel.

Select Ollama as the source.

Set Base URL to http://localhost:11434.

Set Model ID to deepseek-r1 (or deepseek-r1:7b for the 7B variant).

Save the configuration.

After saving, the assistant will query the locally running DeepSeek model.

Additional notes

Local deployment removes API rate limits and data‑privacy concerns. Model size determines resource consumption: larger models (e.g., 32B, 70B, 671B) require more GPU/CPU memory and longer startup times.