Skip to content

Commit 2d4b963

Browse files
gitplayer4gitplayer4
authored
Update agent.md
1 parent 2e35b51 commit 2d4b963

1 file changed

Lines changed: 221 additions & 0 deletions

File tree

‎docs/user_manual/ai/agent.md‎

Lines changed: 221 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1 +1,222 @@
1+
# OpenClaw Installation and Deployment Guide
2+
This document mainly introduces the installation and deployment process of OpenClaw on Linux clients or servers. We recommend using 1Panel, an open-source Linux operation and maintenance management panel, for installation and deployment, which makes the entire process simple and fast. After OpenClaw is installed, the Feishu channel docking configuration can also be quickly completed through the 1Panel operation and maintenance management panel, facilitating the rapid implementation of a Feishu-based personal AI assistant.
13

4+
# 1. Resource and Environment Preparation
5+
The following resource and environment preparations are required for the installation and deployment of OpenClaw:
6+
7+
## Client/Server Preparation
8+
- Operating System: Supports mainstream Linux distributions (based on Debian/RedHat, including domestic operating systems)
9+
- Server Architecture: x86_64, aarch64
10+
- Memory Requirement: It is recommended to have more than 2GB of available memory
11+
- Browser Requirement: Use modern browsers such as Chrome, FireFox, IE10+, Edge, etc.
12+
- Network Requirement: Internet access is available
13+
14+
## Large Language Model API Key
15+
- Public Models: API Keys of large language models such as DeepSeek, Kimi, OpenAI, etc.
16+
- Local Models
17+
18+
## 1Panel v2.1.0 Installation and Deployment
19+
Refer to the official 1Panel documentation for installation and deployment details at the following link: https://1panel.cn/docs/v2/installation/online_installation/
20+
21+
Based on the above environment, the entire installation and deployment process can be completed in four parts: 1Panel installation and deployment, large language model API Key preparation, OpenClaw installation and deployment, and Feishu channel configuration. For detailed instructions, see the following documentation.
22+
23+
# 2. 1Panel Installation and Deployment
24+
To install and deploy 1Panel, first log in to the Linux client or server, then follow the steps below for quick deployment:
25+
26+
## 2.1. Step 1: Obtain root Privileges
27+
After logging in to the server, enter the following command to switch to root privileges first:
28+
```
29+
sudo su -
30+
```
31+
32+
## 2.2. Step 2: Execute the Online Installation Command
33+
Enter the following online installation script command to start the installation:
34+
```
35+
bash -c "$(curl -sSL https://resource.fit2cloud.com/1panel/package/v2/quick_start.sh)"
36+
```
37+
38+
## 2.3. Step 3: Specify the 1Panel Installation Directory
39+
The installation script will first prompt to set a specified installation directory. Press Enter to select the default directory if there are no special requirements:
40+
41+
Set the 1Panel installation directory (default: /opt):
42+
43+
## 2.4. Step 4: Complete Docker Installation
44+
After specifying the directory, the installation script will automatically detect whether Docker is installed on the current server. If not, it will ask whether to install it. Enter "y" here to confirm the installation:
45+
46+
Docker not detected, install it? [y/n]: y
47+
48+
## 2.5. Step 5: Configure Image Accelerator and Set Default Parameters
49+
After Docker is installed, the system will prompt whether to configure the image accelerator and then guide you to set the 1Panel port, secure access entry, panel user password, etc., as shown in the figure below. Configure these settings according to your actual situation, and ensure that the specified port is open.
50+
51+
图片1
52+
53+
## 2.6. Step 6: Obtain 1Panel Login Information
54+
After completing all settings, the system will automatically print out the 1Panel login and access information, as shown in the figure below. Save this information for subsequent use.
55+
56+
图片2
57+
58+
## 2.7. Step 7: Verify the Successful Deployment of 1Panel
59+
Enter the external address obtained in Step 6 into the browser to access the 1Panel login page. Enter the corresponding panel user and password to confirm that 1Panel has been successfully installed and deployed.
60+
图片3
61+
62+
## 2.8. Step 8: Set the 1Panel Access Address
63+
After entering the panel, switch to the "Panel Settings" page and set the default access address to the public network IP for 1Panel access, so that subsequently deployed applications can be accessed by clicking the "Jump" button.
64+
图片4
65+
66+
## 2.9. Step 9: Upgrade to the Latest Version V2.1.0 (Optional)
67+
For an existing 1Panel installation, log in to the panel and click the update button in the lower right corner to upgrade the current 1Panel to version v2.1.0.
68+
图片5
69+
70+
# 3. Large Language Model API Application
71+
You can check the list of models supported on the OpenClaw official website and obtain the API Key of the large language model respectively. This document takes DeepSeek as an example. For other public models and local models, prepare the relevant API Keys by referring to the corresponding documents on your own.
72+
73+
## 3.1. Step 1: Register on the DeepSeek Developer Platform
74+
Access the DeepSeek Developer Platform via the link: https://platform.deepseek.com/, first complete personal real-name authentication and registration, and then make a recharge, as shown in the figure below.
75+
图片6
76+
77+
## 3.2. Step 2: Obtain the DeepSeek API Key
78+
As shown in the figure below, enter the API keys page, click "Create API key", and keep the generated API Key in a safe place for subsequent use after creation.
79+
图片7
80+
81+
# 4. OpenClaw Installation and Deployment
82+
The installation of OpenClaw is deployed based on the agent management of 1Panel, and the DeepSeek large model API key obtained in the previous step needs to be used during the deployment process. In 1Panel, the deployment of OpenClaw and the management of large model API key accounts are divided into two separate parts and decoupled from each other, mainly to facilitate the adjustment of model configurations for your OpenClaw personal assistant.
83+
84+
## 4.1. Step 1: Add a Model Account
85+
Enter the "Agent" management menu under "AI" management, click to enter and switch to "Model Account" management first, then click "Add model account". Select the model provider and enter the model account information as required to complete the model account creation.
86+
图片8
87+
88+
## 4.2. Step 2: Create an Agent
89+
After preparing the model account, switch to the "Agent" page and click "Create Agent", then enter the relevant parameters as required, as shown in the figure below：
90+
图片9
91+
92+
The parameter definitions during agent creation are detailed as follows:
93+
94+
- Name: The default is openclaw, which can be customized as the container name.
95+
- Application Version: Refers to the version of OpenClaw, e.g., the latest version 2026.2.9 as shown in the above figure by default.
96+
- WebUI Port: The default port of OpenClaw is 18789, which can be customized and must be open.
97+
- Bridge Port: The default port of OpenClaw is 18790, which can be customized and must be open.
98+
- Model Provider: The currently supported models are shown in the figure. Select the DeepSeek model provider added just now. If the model account for the corresponding model provider has been added, the number of model accounts will be prompted directly.
99+
100+
图片10
101+
102+
- Other Model Parameters: After selecting the model provider, the system will automatically retrieve the maintained model accounts, as shown in the figure below. You can also check "Manual input" to enter the model information manually. If no model account has been created, click "Create model account" to complete the creation.
103+
104+
图片11
105+
106+
- Token: A token for the OpenClaw Web UI access address, which is automatically generated by the system for direct access in the follow-up.
107+
- Other Parameters: Keep the default configurations for all other parameters.
108+
109+
After configuring all the above parameters, click "Confirm" directly to start the installation of OpenClaw. The installation is completed when the prompt as shown in the figure below appears.
110+
图片12
111+
112+
## 4.3. Step 3: Verify the Successful Deployment of OpenClaw
113+
After the installation and deployment of OpenClaw are completed, enter the agent list page and click "WebUI" to jump directly to the OpenClaw page, as shown in the figure below.
114+
图片13
115+
116+
After entering the OpenClaw page, send a message and check if the AI assistant responds. A normal response, as shown in the figure below, indicates that OpenClaw has been successfully deployed.
117+
图片14
118+
119+
# 5. Feishu Channel Configuration
120+
Up to this point, OpenClaw has been fully deployed. Next, we will configure the Feishu channel. To configure the Feishu channel, we first need to create an available robot in Feishu. Follow the steps below to complete the configuration step by step.
121+
122+
Note: A personal Feishu account is used in this guide. For enterprise accounts, version release and permission authorization require administrator approval, while other operation steps remain the same.
123+
124+
## 5.1. Step 1: Create a Custom Enterprise App
125+
First, log in to Feishu and enter the Feishu Open Platform (link: https://open.feishu.cn/app), then access the "Developer Console" and select "Custom Apps", click "Create Custom App", as shown in the figure below.
126+
图片15
127+
128+
As shown in the figure below, enter the relevant app name and basic information as required and click "Create".
129+
图片16
130+
131+
## 5.2. Step 2: Create a Bot
132+
As shown in the figure below, click to create a bot to complete the bot creation process.
133+
图片17
134+
135+
After entering the bot page, click the edit button after bot configuration to define the bot name, as shown in the figure below：
136+
图片18
137+
138+
## 5.3. Step 3: Permission Configuration
139+
After creating the bot, click to enter "Permissions & Scopes" and then click "Batch import/export scopes".
140+
图片19
141+
142+
Click "Batch import/export scopes", clear the default permission configuration information, copy and paste the permission authorization script as shown below, and click "Save".
143+
144+
```json
145+
{
146+
"scopes": {
147+
"tenant": [
148+
"aily:file:read",
149+
"aily:file:write",
150+
"application:application.app_message_stats.overview:readonly",
151+
"application:application.self_manage",
152+
"application:bot.menu:write",
153+
"cardkit:card:write",
154+
"contact:contact.base:readonly",
155+
"contact:user.employee_id:readonly",
156+
"corehr:file:download",
157+
"docs:document.content:read",
158+
"event:ip_list",
159+
"im:chat",
160+
"im:chat.access_event.bot_p2p_chat:read",
161+
"im:chat.members:bot_access",
162+
"im:message",
163+
"im:message.group_at_msg:readonly",
164+
"im:message.group_msg",
165+
"im:message.p2p_msg:readonly",
166+
"im:message:readonly",
167+
"im:message:send_as_bot",
168+
"im:resource",
169+
"sheets:spreadsheet",
170+
"wiki:wiki:readonly"
171+
],
172+
"user": [
173+
"aily:file:read",
174+
"aily:file:write",
175+
"contact:contact.base:readonly",
176+
"im:chat.access_event.bot_p2p_chat:read"
177+
]
178+
}
179+
}
180+
```
181+
182+
The effect after pasting is shown in the figure below：
183+
图片20
184+
185+
Click "Next, Review New Scopes" and finally ensure that all permissions are enabled. For personal accounts, confirm the permissions by yourself; for enterprise accounts, administrator review is required. Ensure all permissions are enabled as shown in the figure below：
186+
图片21
187+
188+
## 5.4. Step 4: Obtain Credentials and Configure in 1Panel
189+
Enter the Feishu platform and obtain the app credentials in "Credentials & Basic Info", as shown in the figure below：
190+
图片22
191+
192+
After obtaining the credentials, enter the "Configuration" page of "Agent" in 1Panel, complete the Feishu chat channel configuration, and click "Save", as shown in the figure below：
193+
图片23
194+
195+
## 5.5. Step 5: Create Events and Callbacks
196+
As shown in the figure below, enter the "Events & Callbacks" menu and complete the subscription method setting and event addition respectively.
197+
图片24
198+
199+
### Subscription Method Setting
200+
Select the persistent connection subscription method as shown in the figure below：
201+
图片25
202+
203+
### Add Events:
204+
Enter "im.message.receive_v1" to search, check "Receive messages" based on "App-based Subscription", and finally confirm the addition.
205+
图片26
206+
207+
## 5.6. Step 6: Create and Release a Version
208+
After confirmation, click "Create Version", then enter the relevant version information as required and release it. No approval is required for personal accounts, while enterprise accounts need enterprise approval.
209+
图片27
210+
211+
## 5.7. Step 7: Verify the Feishu Channel Configuration
212+
After completing all the above configurations, open the app in the Feishu client as shown in the figure below：
213+
图片28
214+
215+
Finally, perform a simple test. The prompt as shown in the figure below indicates that the configuration is successful.
216+
图片29
217+
218+
# 6. Model Configuration Modification
219+
If you need to switch the model when using the OpenClaw personal AI assistant, also enter the "Agent" list, click "Configuration", enter the model switching menu, complete the model configuration modification and click "Save", as shown in the figure below：
220+
图片30
221+
222+
图片31

0 commit comments

Comments
 (0)