AI Model Governance & Configuration

​

Overview

• AI Model Access Control: Restrict which AI models organization members can use
• Organization Credentials: Centrally manage API keys and authentication credentials
• AI Proxy Routing: Route AI provider requests through custom proxy URLs

​

AI Model Access Control

​

Overview

​

Accessing Model Access Control

​

Key Features

• Allow/Deny Lists: Configure either an allow list (only selected models available) or deny list (selected models blocked)
• Provider-Level Control: Select/deselect all models from specific AI providers at once
• Fallback Model: Automatically designate a fallback model when restricted models are requested

​

Supported AI Providers

• OpenAI (GPT models)
• Anthropic (Claude models)
• Google (Gemini models)
• DeepSeek models
• Azure OpenAI
• Grok
• And other integrated providers

​

Configuration Process

​

1. Enable Model Restrictions

• Toggle “restrict models?” switch to enable access control
• When first enabled, all models are selected by default (allow list mode)

​

2. Choose Control Mode

• Best for strict control environments
• Users can only access explicitly permitted models
• Recommended for compliance-heavy organizations

• Best for general restrictions with specific exclusions
• Users can access all models except those explicitly blocked
• Recommended for cost control scenarios

​

3. Select Models by Provider

• Browse models organized by AI provider
• Use provider-level checkboxes to select/deselect all models from a provider
• Individual model selection available within each provider

​

4. Configure Fallback Model

• System automatically selects first available model as fallback
• Used when a restricted model is requested in workflows
• Ensures workflows continue to function even when preferred models are restricted

​

Common Use Cases

• Enable allow list mode
• Only permit models that meet data residency requirements
• Block models that don’t comply with industry regulations

• Allow only fast-response models for real-time workflows
• Block slower but more capable models for batch processing

​

Organization Credentials

​

Overview

​

Accessing Organization Credentials

​

Key Features

• Centralized Management: Single point of control for all organization AI API keys
• Automatic Override: Organization credentials take precedence over both personal and workspace credentials

​

Supported AI Providers

• When an organization credential exists for an AI provider, it will always be used
• Workspace credentials are ignored when organization credentials are present
• Personal credentials are only used when neither organization nor workspace credentials exist
• This ensures consistent billing and access across all organization workflows

​

Configuration Process

​

1. Add New AI Credential

• Click “Add Credential” button
• Select from available AI providers: OpenAI, Anthropic, Perplexity, or XAI
• Choose “API Key” as the credential type

​

2. Configure API Key

• Enter your organization’s API key for the selected provider
• Set any provider-specific configuration options

​

Benefits

• All organization members automatically use organization API keys
• Eliminates need for individual API key management across teams
• Ensures consistent access across all workflows regardless of workspace

• Centralized billing for all AI API usage
• All AI requests use organization’s negotiated rates and limits
• Easier tracking of consumption across teams and projects
• Simplified budget management and forecasting

• Ensure all AI calls use audited and compliant credentials
• Maintain consistent security policies across workflows
• Prevent unauthorized AI usage through personal credentials

​

AI Proxy Routing

​

Overview

​

Key Features

• Custom Proxy URLs: Route requests through organization-controlled proxy servers
• Model Name Mapping: Map Gumloop model names to custom proxy model identifiers
• Provider-Specific Configuration: Configure different proxies for different AI providers

​

Configuration Process

​

1. Access Proxy Configuration

• Navigate to organization credentials page
• Find AI provider credential (OpenAI, Anthropic, etc.)
• Click “Configure Proxy” button (only visible for AI providers)

​

2. Set Proxy URL

• Enter custom proxy base URL (e.g., https://your-proxy.company.com)
• URL will replace the default AI provider endpoint
• Must be accessible from Gumloop infrastructure

​

3. Configure Model Mappings (Optional)

• Map Gumloop model names to proxy-specific model identifiers
• Example: Map gpt-4 to custom-gpt-4-enterprise
• Useful when proxy uses non-standard model names

​

Supported Scenarios

• Route through corporate AI management platforms
• Implement centralized logging, monitoring, and cost tracking
• Apply consistent security policies across all AI usage

• Access organization-fine-tuned models through custom endpoints
• Map standard model names to specialized variants
• Maintain compatibility with existing workflows

• Ensure all AI requests stay within specific geographic regions
• Route through compliant proxy infrastructure
• Maintain audit trails for regulatory requirements

​

Configuration Examples

​

Basic Proxy Setup

Proxy URL: https://ai-gateway.company.com/v1
Model Mappings: (none - use default model names)

​

Custom Model Names

Proxy URL: https://custom-ai.company.com/api
Model Mappings:
  gpt-4 → company-gpt-4-tuned
  gpt-3.5-turbo → company-gpt-35-fast
  claude-3-opus → company-claude-opus

​

Regional Compliance

Proxy URL: https://eu-ai-proxy.company.com/v1
Model Mappings: (none - standard names with EU routing)

​

How These Features Work Together

1. Model Selection: AI Model Access Control determines which models are available
2. Credential Resolution: Organization Credentials provide the authentication
3. Request Routing: AI Proxy Routing determines the endpoint and model mapping
4. Execution: Request is sent through the configured proxy with organization credentials

​

Security and Compliance

​

Access Control

• All features require organization admin privileges

​

Data Security

• Encrypted storage for all credentials and configuration
• Secure transmission of proxy configurations
• Audit logging for all administrative actions

​

See Also

• Custom User Roles
• Usage Data Export
• Organizations and Workspaces