𝐏𝐫𝐨𝐱𝐲𝐗𝐀𝐈

Configuration Management

Configuration Management

XAI provides a powerful configuration management system that allows administrators to dynamically update system settings without service restart.

Overview

The configuration system manages:

  • Model mappings and routing
  • Resource access controls
  • Email settings
  • System behaviors

Configuration API

Get Current Configuration

Retrieve all current configuration values.

Endpoint: GET /x-config

Response:

{
  "success": true,
  "configs": {
    "MODEL_MAPPER": "gpt-3.5-turbo=gpt-4,gpt-4-32k=gpt-4",
    "LEVEL_MAPPER": "gpt-4=2,claude-3=2",
    "SWITCH_OVER": "1=2,2=3",
    "RESOURCES": "/v1/chat/completions,/v1/embeddings",
    "XAI_MAIL": "[email protected]",
    "EMAIL_SMTP": "smtp.gmail.com",
    "EMAIL_PORT": "587"
  }
}

Update Configuration

Update one or more configuration values.

Endpoint: POST /x-config

Request Body:

{
  "MODEL_MAPPER": "gpt-3.5-turbo=gpt-4,claude-instant=claude-3",
  "LEVEL_MAPPER": "gpt-4=2,claude-3=3",
  "RESOURCES": "/v1/chat/completions,/v1/embeddings,/v1/images/generations"
}

Response:

{
  "success": true,
  "updated": {
    "MODEL_MAPPER": "gpt-3.5-turbo=gpt-4,claude-instant=claude-3",
    "LEVEL_MAPPER": "gpt-4=2,claude-3=3",
    "RESOURCES": "/v1/chat/completions,/v1/embeddings,/v1/images/generations"
  },
  "message": "Configurations updated successfully"
}

Delete Configuration

Reset configuration to default values.

Endpoint: DELETE /x-config

Request Body:

{
  "keys": ["MODEL_MAPPER", "LEVEL_MAPPER"]
}

Response:

{
  "success": true,
  "deleted": ["MODEL_MAPPER", "LEVEL_MAPPER"],
  "message": "Configurations deleted successfully (restored to default env)"
}

Configuration Parameters

MODEL_MAPPER

Maps incoming model requests to different models.

Format: source_model=target_model

Examples:

gpt-3.5-turbo=gpt-4
gpt-4-32k=gpt-4
claude-instant=claude-3-sonnet

Use Cases:

  • Upgrade requests from older models
  • Route to more cost-effective alternatives
  • A/B testing different models

LEVEL_MAPPER

Maps models to specific key levels for routing.

Format: model=level

Examples:

gpt-4=2
claude-3-opus=3
gpt-3.5-turbo=1

Purpose:

  • Route expensive models to premium keys
  • Distribute load across key levels
  • Control access to specific providers

SWITCH_OVER

Defines automatic level switching on failures.

Format: from_level=to_level

Examples:

1=2    # If level 1 fails, try level 2
2=3    # If level 2 fails, try level 3
3=1    # If level 3 fails, try level 1

Benefits:

  • Automatic failover
  • High availability
  • Load distribution

RESOURCES

Defines allowed API endpoints globally.

Format: Comma-separated endpoint paths

Examples:

/v1/chat/completions,/v1/embeddings
/v1/chat/completions,/v1/embeddings,/v1/images/generations,/v1/audio/transcriptions

Security:

  • Restrict access to specific endpoints
  • Prevent unauthorized API usage
  • Compliance with usage policies

Email Configuration

Configure email settings for notifications.

Parameters:

  • XAI_MAIL - System notification email
  • EMAIL_SMTP - SMTP server address
  • EMAIL_PORT - SMTP port (25, 587, 465)
  • EMAIL_AUTH - Authentication email
  • EMAIL_PASS - Authentication password
  • EMAIL_SENDER - Sender name/email
  • EMAIL_TLS - Enable TLS (true/false)

Example:

{
  "EMAIL_SMTP": "smtp.gmail.com",
  "EMAIL_PORT": "587",
  "EMAIL_AUTH": "[email protected]",
  "EMAIL_PASS": "app-specific-password",
  "EMAIL_SENDER": "XAI System <[email protected]>",
  "EMAIL_TLS": "true"
}

Advanced Configuration

System-Wide Settings

Some configurations require root access:

# Enable logging
curl -X POST https://api.proxyxai.com/x-conf \
  -H "Authorization: Bearer $ROOT_KEY" \
  -H "Content-Type: application/json" \
  -d '{"LogEnable": true}'

# Force key reload
curl -X POST https://api.proxyxai.com/x-conf \
  -H "Authorization: Bearer $ROOT_KEY" \
  -H "Content-Type: application/json" \
  -d '{"LoadKeys": true}'

# Update stats interval
curl -X POST https://api.proxyxai.com/x-conf \
  -H "Authorization: Bearer $ROOT_KEY" \
  -H "Content-Type: application/json" \
  -d '{"StatsInterval": 100}'

Model Limits Configuration

Set global model-specific limits:

{
  "ModelLimits": {
    "gpt-4": {
      "rpm": 100,
      "tpm": 150000
    },
    "claude-3-opus": {
      "rpm": 50,
      "tpm": 100000
    }
  }
}

Dynamic Model Mapping

Use wildcards for flexible routing:

gpt-4*=gpt-4
claude-*=claude-3-sonnet
*-vision=gpt-4-vision-preview

Configuration Best Practices

1. Model Mapping Strategy

Development Environment:

gpt-4=gpt-3.5-turbo  # Use cheaper models
claude-3-opus=claude-3-sonnet

Production Environment:

gpt-3.5-turbo=gpt-4  # Upgrade for better quality
claude-instant=claude-3-sonnet

2. Level Distribution

Typical Setup:

# Level 1: High-volume, low-cost models
gpt-3.5-turbo=1
claude-instant=1

# Level 2: Standard models
gpt-4=2
claude-3-sonnet=2

# Level 3: Premium models
gpt-4-32k=3
claude-3-opus=3

3. Failover Configuration

High Availability:

# Circular failover
1=2
2=3
3=1

# Or hierarchical failover
1=2
2=3
# No 3=X means level 3 doesn't fail over

4. Resource Security

Minimal Access:

/v1/chat/completions  # Only chat API

Full Access:

/v1/chat/completions,/v1/embeddings,/v1/images/generations,/v1/audio/transcriptions,/v1/moderations

Configuration Persistence

  • Configurations are stored in Redis
  • Changes are applied immediately without restart
  • Configurations persist across service restarts
  • Default values from environment variables are used as fallback

Monitoring Configuration Changes

Configuration changes are logged and can be monitored:

  1. Through Logs:

    • All configuration updates are logged
    • Failed updates include error details

  2. Through API:

    • Query current configuration state
    • Compare with expected values

  3. Through Webhooks:

    • Set up notifications for configuration changes
    • Monitor critical setting modifications

Troubleshooting

Common Issues

  1. Configuration Not Taking Effect:

    • Verify the configuration was saved successfully
    • Check for syntax errors in mappings
    • Ensure proper permissions for updates

  2. Model Routing Issues:

    • Verify MODEL_MAPPER syntax
    • Check LEVEL_MAPPER assignments
    • Confirm target models exist

  3. Email Not Working:

    • Test SMTP credentials
    • Verify port and TLS settings
    • Check firewall rules

Configuration Validation

Before applying configurations:

  1. Test model mappings with sample requests
  2. Verify level assignments match available keys
  3. Confirm resource paths are valid
  4. Test email settings with test message
Configuration changes may affect all users immediately. Test changes in a development environment first. Keep backups of working configurations.
← XAI Reference
Features →