๐—๐€๐ˆ ๐‘๐จ๐ฎ๐ญ๐ž๐ซ

Admin Console

XAI Admin is a graphical configuration management console for owners, used to manage provider keys, system configuration, news broadcasts, and operation logs.

Feature Overview

The XAI Admin console includes the following core functional modules:

  1. AI API Documentation - View supported AI service providers and SDK examples
  2. Model Services (Provider Management) - Manage upstream AI provider keys
  3. System Configuration - Configure model mapping, level mapping, resource allowlists, etc.
  4. Broadcast Notifications - Publish system-level or targeted user notifications
  5. Operation Logs - View all account operation records

Access Control

  • Owner Only: Only XAI Control owner accounts (isOwner=true) can log in to the Admin console
  • Owner Definition: An owner is a first-level account created by self-signup on xaicontrol.com. Owners can configure and manage their own third-party AI API Keys in a.xaicontrol.com and distribute resources to sub-accounts
  • Sub-account and product entry points: Sub-accounts use m.xaicontrol.com (Manage) for daily operations; m.xairouter.com is the login entry for XAI Router users and is not an XAI Control Admin entry
  • Local Key Storage: API Key is only stored in browser local storage, never uploaded to server
  • Session Management: Login state is stored in browser, requires re-login after closing page

Common Task Guide

  • Add upstream provider โ†’ open "Model Services" โ†’ submit form
  • View or filter providers โ†’ use the "Model Services" list
  • Configure model mapping, level mapping, and governance boundaries โ†’ use "System Configuration"
  • Publish system or targeted news โ†’ use "Broadcast Notifications"
  • View operation logs โ†’ use "Operation Logs"

Feature and Permission Overview

This page focuses on the console interface and usage flow, not the underlying endpoint inventory.

  • Provider management: owner only
  • Owner configuration: owner only
  • Broadcast notifications: owner only for system-wide notices; targeted actions require the appropriate ancestor scope
  • Operation logs: owner only

Terminology: uses Model Mapper, Level Mapper, Resources (allowlist), Model Limits, Sleep, Switch Over โ€” see "Terminology".

Feature Details

1. AI API Tab

Displays platform-supported AI service providers and SDK usage examples.

Features:

  • Model List Link - Dynamic display of /v1/models endpoint link
  • Supported AI Service Providers - Shows 9 major AI services (OpenAI, Claude, Perplexity, Mistral, Grok, Gemini, Cohere, DeepSeek, AI21)
  • SDK Example Code - Complete examples for OpenAI SDK and Claude SDK
  • One-Click Copy - Example code supports one-click copy

Note: This page only explains what is visible in the console and how it is used.

2. Model Services (Provider Management)

Manage upstream AI provider key pools with support for multiple provider types.

Features:

  • Statistics Cards - Display total levels, total keys, active keys
  • Add Provider - Support three configuration types:
    • Standard configuration (OpenAI, Claude, etc.)
    • Azure OpenAI configuration
    • Google Vertex AI configuration
  • Provider List - Display all providers grouped by level
  • Edit/Delete Provider - Support modifying provider configuration or deletion
  • Status Management - Enable/disable providers

Add Provider Form Fields:

  • Name - Provider name (optional, for identification)
  • Level - Load pool level (required, for routing)
  • Provider - Upstream API URL (required)
  • SecretKey - API key (required)
  • Standard Config - No additional fields
  • Azure Config - Model Mapping, API Version mapping
  • Vertex Config - Base URL, Project ID, Client Email, Private Key

Note: Provider management is backed by server-side capabilities, but low-level interface details are intentionally omitted here.

3. System Configuration Tab

Manage owner-level system configuration parameters.

Configuration Items:

Owner Pricing Overrides:

  • A PRICING editor is available at the top of the tab (provide only deltas from system defaults).
  • Actions:
    • โ€œLoad Currentโ€ reads configs.PRICING from backend
    • โ€œSave Pricing (Apply Now)โ€ writes PRICING via PUT /x-config
    • โ€œClear (Restore Defaults)โ€ removes PRICING via DELETE /x-config
  • Constraints: JSON size โ‰ค 128k, total entries โ‰ค 1024; unknown fields are rejected; all numbers must be finite and nonโ€‘negative.

Basic Configuration:

  • XAI_MAIL - System notification email
  • RESOURCES - Allowed API resource paths (comma-separated)
  • MODEL_LIMITS - Model-level rate limits (JSON format; RPD/TPD are natural-day quotas, while RPM/RPH/TPM/TPH are short-window throttles)
  • MODEL_LIMITS accounting is merged into the effective model, so aliases, mappings, or upstream snapshot names do not create duplicate counters

Email Configuration:

  • EMAIL_SMTP - SMTP server address
  • EMAIL_PORT - SMTP port
  • EMAIL_AUTH - Authentication email
  • EMAIL_PASS - Email password

Mapping Configuration:

  • MODEL_MAPPER - System model mapping (format: source=target)
    • Example: gpt-3.5*=gpt-5-nano,gpt-4=gpt-5
    • Purpose: Automatically convert requested models to other models
  • LEVEL_MAPPER - Model group mapping (format: model=level)
    • Example: gpt*=1,claude*=2,deepseek*=3
    • Purpose: Specify which level's provider pool handles which models
  • SWITCH_OVER - Model group primary/backup switching (format: primary=backup)
    • Example: 1=10,2=20
    • Purpose: When primary level has >5 errors in 1 minute, auto-switch to backup level

Note: System configuration is backed by server-side capabilities, but low-level interface details are intentionally omitted here.

4. Broadcast Notifications Tab

Publish system-level or targeted user notifications.

Features:

  • Create Notification - Fill in title and content to publish notification
  • Notification List - View historical notification records
  • Delete Notification - Delete expired or incorrect notifications

Field Description:

  • title - Notification title (required, max 100 characters)
  • content - Notification content (required, max 1000 characters)

Note: Broadcast notifications are backed by server-side capabilities, but low-level interface details are intentionally omitted here.

5. Operation Logs Tab

View all account operation records with pagination support.

Features:

  • Log Table - Display time, action, target account, details, IP address
  • Auto Pagination - Default 24 records per page
  • Real-time Update - Auto refresh when entering tab

Note: Operation logs are backed by server-side capabilities, but low-level interface details are intentionally omitted here.

Use Cases

Use Case 1: Add Standard Provider

  1. Go to "Model Services" tab
  2. Fill in "Add Provider" form:
    • Name: OpenAI Production
    • Level: 1
    • Provider: https://api.openai.com
    • SecretKey: sk-...
  3. Click "Add Provider"
  4. System automatically adds gpt*=1 mapping to LEVEL_MAPPER

Use Case 2: Configure Azure OpenAI

  1. Go to "Model Services" tab
  2. Select configuration type: "Azure OpenAI Configuration"
  3. Fill in basic info and Azure-specific config:
    • Provider: https://your-resource.openai.azure.com
    • Model โ†’ Deployment mapping: {"gpt-5": "gpt-4-deployment"}
  4. Click add
  5. System automatically handles Azure-specific request routing

Use Case 3: Set Model Mapping

  1. Go to "System Configuration" tab
  2. In "System Model Mapping" enter:
    gpt-3.5*=gpt-5-nano,o*=gpt-5
  3. In "Model Group Mapping" enter:
    gpt*=1,claude*=2,deepseek*=3
  4. Click "Save Configuration"
  5. System automatically applies new mapping rules

Use Case 4: Publish System Notification

  1. Go to "Broadcast Notifications" tab
  2. Fill in notification title and content:
    • Title: System Maintenance Notice
    • Content: System will undergo maintenance tomorrow 2:00-4:00 AM...
  3. Click "Send Notification"
  4. All users will see this notification in Manage console

Best Practices

Provider Management

  1. Group by Service - Put same service provider's keys in same level
  2. Set Backup Switching - Use SWITCH_OVER to configure backups for critical levels
  3. Regular Check Sleeping Keys - View sleeping status in provider list
  4. Reasonable Level Setting - Divide different levels by cost, speed, reliability

Configuration Management

  1. Model Mapping Purpose - Use for smooth migration or cost optimization
  2. Careful RESOURCES Modification - Wrong config may cause API access issues
  3. MODEL_LIMITS Restrictions - Set strict rate limits for expensive models
  4. Regular Config Backup - Use /x-config endpoint to export configuration

Notification Management

  1. Important Notifications First - Use system-level notifications for critical information
  2. Concise Content - Avoid overly long notification content
  3. Timely Cleanup - Delete expired or invalid notifications

FAQ

Q: Why can't I log in to Admin console?

A: The Admin console is only available to XAI Control owner accounts. Use the correct entry point by product and account type:

  • m.xairouter.com: Login entry for XAI Router (xairouter.com) users (Manage)
  • a.xaicontrol.com: Owner-level configuration entry for XAI Control (xaicontrol.com) (Admin)
  • m.xaicontrol.com: Entry for regular and sub-accounts on XAI Control (Manage)

Q: Models still can't be used after adding Provider?

A: Check the following:

  1. Does LEVEL_MAPPER include mapping for this model
  2. Is Provider status enabled
  3. Is SecretKey valid

Q: How to view system auto-added LEVEL_MAPPER?

A: You can view and modify in "System Configuration" tab's "Model Group Mapping" field.

Q: Is Azure config's API Version mapping required?

A: No, it's optional. If left empty, system will use default versions configured in environment variables.

Q: How to backup my configuration?

A: Use the following command to export configuration:

# SaaS
curl -H "Authorization: Bearer $API_KEY" \
  https://api.xairouter.com/x-config > config-backup.json

# BYOK
curl -H "Authorization: Bearer $API_KEY" \
  https://api.xaicontrol.com/x-config > config-backup.json
โ† Access Control (ACL)
XAI Control โ†’