Administration & Operations Overview

Applicable roles: Administrator (Super Administrator / Administrator)

Prerequisites: Signed in with an administrator account; see Quick Start

This page is the main entry point for administrators. Read "1. Where to Start" first to determine where to begin, then jump to the corresponding guide as needed.

---

1. Where to Start

The platform has a clear configuration dependency chain. Getting the order wrong causes downstream features to become unavailable or invisible to users.

New deployment: follow this order once

#	Step	Guide	Why it must be at this position
1	Upload the organization logo and enter the organization short name	Organization Settings	Login page branding, purely cosmetic; it can be done first or last
2	Connect a model provider and set the default model	Model Provider Access	⚠️ The most critical step. Without a default language model, no user is able to converse
3	Adjust the AI agent's prompt and tool permissions	AI Agent Settings	The main AI agent is built in and works out of the box; adjust it as needed
4	Configure dynamic scenes	Dynamic Scenes	Depends on groups, so the groups must be created first (step 8)
5	Register and enable plugins	Plugin Management	Optional
6	Connect MCP services	MCP Configuration	Optional. Context7 and WebSearch are already built in
7	Register Git repositories and configure credentials	Integration Settings	Required for team collaboration scenarios
8	Create groups, add users, and assign nodes	Users & Permissions	⚠️ Users cannot use the platform at all until a node is assigned to them
9	Authorize Skills and key documents	Assets	Depends on repositories and groups, so it must come after steps 7 and 8
10	Register extension applications	Extension Settings	Optional, used to connect existing enterprise business systems

For a platform that is already running: routine inspections

Frequency	What to check	Guide
Daily	Usage overview, node health status	Operations & Logs · Nodes & Upgrades
Weekly	Containers exiting abnormally, whether newly added users have been assigned a node	User AI Agent Management
Monthly	Whether the backend and image versions are lagging behind	Nodes & Upgrades

---

2. Permission Matrix

By role

Feature	Super Administrator	Administrator	Group user
Workspace, conversations, tasks, workspace files	✅	✅	✅
Account settings, My MCP, My scenes	✅	✅	✅
Extension applications (usage)	✅	✅	Depends on enabled state
Assets (Skills, key documents)	✅	✅	❌
User management, group management	✅	✅	❌
User AI agent management, node management	✅	✅	❌
Model provider access, system configuration, data dictionary, organization settings	✅	✅	❌
Git repository management, Git credentials	✅	✅	❌
Plugins, MCP, AI agent, and scene configuration	✅	✅	❌
Usage overview, system logs, extension settings	✅	✅	❌

Note · The built-in "Super Administrator" and "Administrator" groups have exactly the same permissions, both with full access. The difference lies in the scope of visible data: the group membership of the Super Administrator cannot be modified, which preserves a highest-privilege entry point that cannot be accidentally downgraded.

Group-driven resource visibility

All resource authorization within the system is done on a per-group basis. This is the most common pitfall during troubleshooting — "the feature is missing" usually does not mean it is broken, it means it has not been authorized.

Resource	How it is authorized	Behavior when unauthorized
Git repository	Visible to specified groups; selecting no group = visible to administrators only	Users cannot see the repository in their workspace
Skill directory	Three states: explicitly authorized groups / follow repository authorization / public	The Skill does not appear in the Skill selector in the input area
Key documents	Same as above	The directory is not visible during import
Dynamic scenes	Owning group	The scene does not appear in the scene selector
User AI agent tier	Set directly by the administrator	—

---

3. Guide Map

Guide	Contents	Number of images	When to read it
Model Provider Access	Built-in / custom providers, API protocols, default model, optional model allowlist, connection test	9	When connecting the first model; when switching models
Capability Configuration	AI agents, dynamic scenes, plugin management, MCP configuration	22	To define what the AI can do
Integration Settings	Git repository management, Git credentials (SSH / HTTP)	11	When the AI needs to operate on code repositories
Foundation Settings	Organization settings, system configuration, data dictionary	12	Changing branding; tuning system parameters; setting up multiple languages
Users & Permissions	User management, group management, node assignment, user AI agent tiers, container management	18	Creating accounts, managing permissions, adjusting resource tiers
Nodes & Upgrades	Node list, health checks, image loading strategy, backend and image upgrades	14	Scaling out; upgrading; troubleshooting
Operations & Logs	Usage overview, system logs, extension settings	9	Reviewing usage, checking operation records, attaching sub-applications

---

4. Three Troubleshooting Tracks

When something goes wrong, first determine which category it belongs to, then open the corresponding guide.

Symptom type	Typical behavior	Look first
Configuration	Feature not found, empty list, missing from the dropdown	"Permission Matrix" on this page → check group authorization
Runtime	The AI does not respond, a container fails to start, a node is offline	Nodes & Upgrades
Operations	Who did what, and when	System Logs

---

5. Related Pages

Page	Contents
Product User Guide Overview	The entry point and feature overview for all users
Installation · Multi-Node Deployment	How nodes join the network and share the database and keys
Installation · Upgrade Server Deployment	Version distribution in intranet / offline environments