Digital Employee Settings
Basic settings define a Digital Employee's name, appearance, underlying models, and most critically, its work mode. These settings differ from the three behavior definition documents (AGENTS.md / SOUL.md / IDENTITY.md)—the latter determine "how the Digital Employee speaks", while the former determine "how it runs".
Overview
The settings entry is located under the Settings menu item in the Digital Employee detail page. This page is accessible only to the Digital Employee's creator and administrators.
Basic Information
- Name: The Digital Employee's identifier, displayed externally
- Description: Functional description of the Digital Employee
Avatar
Digital Employees support selecting from preset avatars or uploading a custom one. The avatar appears on the Digital Employee's side of the chat interface.
Underlying Models
A Digital Employee can select one or more underlying large language models. When multiple models are selected, the system automatically routes requests based on quotas and load.
Thinking Level
Thinking Level controls the depth of reasoning when the Digital Employee thinks. Available levels:
| Level | Description |
|---|---|
| minimal | Least thinking, fastest responses |
| low | Less thinking |
| medium | Medium thinking (default) |
| high | Stronger thinking |
| xhigh | Strongest thinking, slowest but highest quality |
The default is Not set, which uses the model's own default thinking level. If the model does not support reasoning (e.g., non-reasoning models), the platform automatically degrades and the setting has no effect.
Work Mode
Work mode is the Digital Employee's most differentiating setting, as it determines the execution environment for tools and skills. The platform supports three work modes:
Assistant Mode
In assistant mode, the Digital Employee's tools and skills execute on the current user's computer. The Digital Employee operates system resources with the user's own identity and permissions.
This mode is suitable for personal use. The Digital Employee functions like a desktop assistant running in the background—it can see your files, execute your commands, and access internal systems with your permissions. Everything happens on your own computer.
From a security perspective, assistant mode naturally inherits the user's permission boundaries: what you can access, the Digital Employee can access; resources you lack permission for, the Digital Employee cannot touch. No additional permission configuration is needed—it works out of the box.
Standalone Mode
In standalone mode, the Digital Employee runs on a designated workstation. A workstation is an independent physical or virtual machine with its own operating system environment and user identity.
This mode is suitable for enterprise deployment scenarios. A typical use case: the IT department prepares a dedicated server as a workstation, installs the Gendial client on it, and registers it with the platform. Afterwards, multiple Digital Employees can share this machine's computing resources and tool environment by selecting the workstation.
When choosing standalone mode, a registered workstation must be specified. If no workstation has been created yet, register one first on the admin panel's Digital Employees → Workstations page.
Dynamic Mode
Dynamic mode is a hybrid mode—the Digital Employee is not pre-bound to a specific execution method but instead determined dynamically by the user at runtime.
This means the same Digital Employee can run in assistant mode on the user's own computer, or switch to standalone mode on a workstation when needed. This flexibility is particularly suitable for scenarios such as:
- Validating on a workstation during development and testing, then running on a local computer for production use
- Switching to a workstation (which has dedicated network permissions) when accessing sensitive internal systems, while performing daily operations locally
Voice Configuration
- Voice Input: When enabled, the Digital Employee's chat interface supports voice input
- Voice Output: When enabled, the Digital Employee's replies can be read aloud via text-to-speech
API Key
The Digital Employee automatically generates an API key for external systems to call it via API. The key can be regenerated, but the old key becomes invalid immediately upon regeneration.