Based on Docker and Pagoda Panels: A Detailed Guide to Private Deployment of Coze Studio Open Source Edition

Recently, AI application development platform Coze announced the official open-sourcing of its Studio version, a major milestone for the developer community. This initiative means that developers can privately deploy Coze's Visual Intelligentsia (Agent) Development Kit to build and manage AI applications with data security and version stability.

For enterprises and developers, localized deployments address a core pain point that has long relied on cloud services. Unexpected outages that may be triggered by cloud platform version iterations or plugin tweaks are significant risks to delivered projects. Open source versions, on the other hand, give development teams complete control and ensure stable workflows over time.

This tutorial will detail how to accomplish this on a cloud server Coze Studio of localized deployment. The entire process follows a step-by-step illustration designed to be successfully practiced by developers of all experience levels.

Estimated completion time: approximately 25 minutes
Technical prerequisites: Familiarity with basic Linux command line operations and understanding of Docker concepts.

Deployment Process Overview:

1. Preparing Cloud Servers
2. Configuring the server environment
3. deployments Coze Studio
4. Getting and Configuring the Model API Key
5. Launch and access Coze Studio

I. Preparing Cloud Servers

Deploying Coze Studio requires at least one 2-core 4G Configured cloud servers.

1. Choose and sign up for a cloud service provider. In the case of Tencent Cloud's lightweight application servers, for example, new users are usually entitled to a free trial credit.
2. Go to the Service Provider Console and select Purchase Server. In the configuration options, the image (Image) is recommended to choose a server with pre-installed Pagoda Linux Panel 8.1.0 This simplifies subsequent configuration of the environment. The region can be selected based on the access speed.

3. After creating the instance, find the "Login" or "Terminal" function on the server management page to enter the command line interface of the server.

4. After logging in, execute the following command to change the default port of the Pagoda Panel (from 8888 to 8887) to avoid conflict with the default port of Coze.

   echo "8887" > /www/server/panel/data/port.pl
   sudo /etc/init.d/bt default
   

   After execution, the terminal will output the new panel access address, username and password, be sure to copy and save this information properly.

   

5. Next, you need to configure the server's firewall rules to open the ports required by Coze Studio. Go to the server's firewall or security group settings page, add the rule, and release it 8887 cap (a poem) 9999 Two ports.

II. Configuring the server environment

1. Open the Pagoda Panel in your browser using the "Extranet Panel Address" you saved earlier. If the address starts with login ending cannot be opened, please try to change the URL suffix to tencentcloudThe

2. Enter your saved username and password to log in. The first time you log in you need to bind an official Pagoda account.

3. After logging in, a pop-up window with recommended packages will appear, so just check the box and install the first one. Nginx, and the rest unchecked.

4. Find and click "Software Store" in the left menu of Pagoda Panel, and select "App Categories". Docker Manager and install it. The installation process may take some time.

5. After the installation is complete, refresh the page. Open the "Terminal" tool on the left side, and execute the following command to verify Docker cap (a poem) Docker Compose Whether the installation was successful or not. If you can see the version number output, the environment is ready.

   docker -v
   docker-compose -v
   

   

Deploying Coze Studio

1. In the Pagoda panel, go to the File Manager and locate the /home directory, create a new directory named Coze of the folder.

2. Go to the newly created Coze folder, click the "Terminal" button at the top to open a new terminal window. Execute the following git Command Clone Coze Studio source code repository.

   git clone https://github.com/coze-dev/coze-studio.git
   

   

   Alternative Programs: If a network problem causes git clone Slow or failing, you can visit directly GitHub repository address Download the ZIP archive and upload it to the server's /home/Coze directory and unzip it.

   

   

Getting and Configuring the Model API Key

Coze Studio A large language model needs to be connected in order to run. The tutorial uses the Volcano Engine's Ark model service as an example to demonstrate how to get the API Key and model access points Endpoint IDThe

4.1 Obtaining Volcano Engine Model Credentials

1. Access and Login volcano engine Console, go to the Ark Modeling Platform.

2. Select "Online Reasoning" in the left menu and click "Create Reasoning Access Point".

3. Name the access point (e.g. DeepSeek-R1) and select a base model such as the deepseek-v2If you are prompted that the model is not open, please follow the instructions to complete the process. If you are prompted that the model is not open, follow the guidelines to complete the opening (free of charge).

4. After successful creation, find the access point you just created in the list."Access Point ID" (Endpoint ID) which is required for subsequent configuration of the model. Please copy and save it.

5. Click on the "API Call" button to the right of the access point and select or create a API KeyThe

6. make a copy of Access Key ID cap (a poem) Secret Access KeyTheSecret Access Key It will only appear once, so be sure to save it immediately. Here's Secret Access Key which is required for subsequent configurations of the api_keyThe

4.2 Configuring Coze Studio Model Files

1. Go back to the Pagoda panel and enter /home/Coze/coze-studio directory, click Terminal at the top and execute the following command to make a copy of the model profile template.

   cp backend/conf/model/template/model_template_ark_doubao-seed-1.6.yaml backend/conf/model/ark_doubao-seed-1.6.yaml
   

   

2. Using Pagoda's file editor, open the just created ark_doubao-seed-1.6.yaml Documentation.

   Path:/home/Coze/coze-studio/backend/conf/model/ark_doubao-seed-1.6.yaml

   

3. Modify three fields in the file:
   • id: A unique identifier for the model, which must be a non-zero integer. For example, a value set to 888. Please do not modify it after it goes live.
   • meta.conn_config.api_key: Paste the Volcano Engine saved earlier Secret Access KeyThe
   • meta.conn_config.model: Paste the Volcano Engine saved earlier Access Point ID (Endpoint ID).

   
   

4. After confirming that the changes are correct, save the file.

V. Launching and accessing Coze Studio

1. return to /home/Coze/coze-studio directory and open the terminal again. Execute the following commands in order to enter the docker directory, create environment variable files, and start all services.

   cd docker
   cp .env.example .env
   docker compose --profile '*' up -d
   

   

2. The first startup will take a while to pull and build multiple Docker images, so please be patient. When the terminal outputs Container coze-server Started or similar success message and no more new logs are rolled over, the service has started successfully.
3. The deployed Coze Studio. The access address is:
   http://<你的服务器公网IP>:9999

   commander-in-chief (military) <你的服务器公网IP> Replace with the IP address displayed in the upper left corner of the Pagoda panel.

   

   Once the page opens, follow the prompts to register your first administrator account to start exploring and building your AI intelligences in your local environment.