Deploy OpenManus Locally and Let It Generate a Complete WeChat Mini‑Program

1. Local Deployment

Step 1: Set up a Python 3.12 environment.

# 1、构建开发环境
conda create -n open_manus python=3.12
conda activate open_manus

Step 2: Clone the OpenManus repository and install its Python dependencies.

# 拉取项目代码
git clone https://github.com/mannaandpoem/OpenManus.git
cd OpenManus
# 安装依赖
pip install -r requirements.txt

Step 3: Copy the example configuration file and edit the LLM settings (replace the API key with your own DeepSeek key).

cp config/config.example.toml config/config.toml

# Global LLM configuration
[llm]
model = "deepseek-chat"
base_url = "https://api.deepseek.com/v1"
api_key = "sk-xxx"
max_tokens = 4096
temperature = 0.0

# Optional vision model configuration
[llm.vision]
model = "deepseek-chat"
base_url = "https://api.deepseek.com/v1"
api_key = "sk-xxx"

Step 4: Start the service. python main.py When the console displays the success message (see image), the OpenManus server is running.

2. Using OpenManus to Build a WeChat Mini‑Program

Provide a prompt such as:

请帮我写一个微信小程序的代码实现计算器功能完整的小程序代码请放到当前目录下

OpenManus processes the request, plans the tasks, and generates the complete mini‑program code. The generated project runs correctly, and the total token cost was about ¥1.97, which is acceptable.

3. Summary

The experience highlighted OpenManus’s strong task planning, tool orchestration, and helpful visual cues. It successfully automated the creation of a simple WeChat mini‑program, demonstrating its automation capabilities. The main drawback is latency—running the full workflow took nearly an hour, likely due to cloud LLM inference speed.