AI Gateway Configuration

Configure providers, endpoints, and advanced settings for your MLflow AI Gateway.

Provider Configurations

Configure endpoints for different LLM providers using these YAML examples:

OpenAI

endpoints:
  - name: gpt4-chat
    endpoint_type: llm/v1/chat
    model:
      provider: openai
      name: gpt-4
      config:
        openai_api_key: $OPENAI_API_KEY
        openai_api_base: https://api.openai.com/v1  # Optional
        openai_organization: your_org_id  # Optional

Azure OpenAI

endpoints:
  - name: azure-chat
    endpoint_type: llm/v1/chat
    model:
      provider: azuread
      name: gpt-35-turbo
      config:
        openai_api_key: $AZURE_OPENAI_API_KEY
        openai_api_base: https://your-resource.openai.azure.com/
        openai_api_version: "2023-05-15"
        openai_deployment_name: your-deployment-name

Anthropic

endpoints:
  - name: claude-chat
    endpoint_type: llm/v1/chat
    model:
      provider: anthropic
      name: claude-2
      config:
        anthropic_api_key: $ANTHROPIC_API_KEY

AWS Bedrock

endpoints:
  - name: bedrock-chat
    endpoint_type: llm/v1/chat
    model:
      provider: bedrock
      name: anthropic.claude-instant-v1
      config:
        aws_config:
          aws_access_key_id: $AWS_ACCESS_KEY_ID
          aws_secret_access_key: $AWS_SECRET_ACCESS_KEY
          aws_region: us-east-1

Cohere

endpoints:
  - name: cohere-completions
    endpoint_type: llm/v1/completions
    model:
      provider: cohere
      name: command
      config:
        cohere_api_key: $COHERE_API_KEY

  - name: cohere-embeddings
    endpoint_type: llm/v1/embeddings
    model:
      provider: cohere
      name: embed-english-v2.0
      config:
        cohere_api_key: $COHERE_API_KEY

MLflow Models

endpoints:
  - name: custom-model
    endpoint_type: llm/v1/chat
    model:
      provider: mlflow-model-serving
      name: my-model
      config:
        model_server_url: http://localhost:5001

Environment Variables

Store API keys as environment variables for security:

# OpenAI
export OPENAI_API_KEY=sk-...

# Azure OpenAI
export AZURE_OPENAI_API_KEY=your-azure-key
export AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/

# Anthropic
export ANTHROPIC_API_KEY=sk-ant-...

# AWS Bedrock
export AWS_ACCESS_KEY_ID=AKIA...
export AWS_SECRET_ACCESS_KEY=...
export AWS_REGION=us-east-1

# Cohere
export COHERE_API_KEY=...

Advanced Configuration

Rate Limiting

Configure rate limits per endpoint:

endpoints:
  - name: rate-limited-chat
    endpoint_type: llm/v1/chat
    model:
      provider: openai
      name: gpt-3.5-turbo
      config:
        openai_api_key: $OPENAI_API_KEY
    limit:
      renewal_period: minute
      calls: 100  # max calls per renewal period

Model Parameters

Set default model parameters:

endpoints:
  - name: configured-chat
    endpoint_type: llm/v1/chat
    model:
      provider: openai
      name: gpt-3.5-turbo
      config:
        openai_api_key: $OPENAI_API_KEY
        temperature: 0.7
        max_tokens: 1000
        top_p: 0.9

Multiple Endpoints

Configure multiple endpoints for different use cases:

endpoints:
  # Fast, cost-effective endpoint
  - name: fast-chat
    endpoint_type: llm/v1/chat
    model:
      provider: openai
      name: gpt-3.5-turbo
      config:
        openai_api_key: $OPENAI_API_KEY

  # High-quality endpoint
  - name: quality-chat
    endpoint_type: llm/v1/chat
    model:
      provider: openai
      name: gpt-4
      config:
        openai_api_key: $OPENAI_API_KEY

  # Embeddings endpoint
  - name: embeddings
    endpoint_type: llm/v1/embeddings
    model:
      provider: openai
      name: text-embedding-ada-002
      config:
        openai_api_key: $OPENAI_API_KEY

Dynamic Configuration Updates

The AI Gateway supports hot-reloading of configurations without server restart. Simply update your config.yaml file and changes are detected automatically.

Security Best Practices

API Key Management

1. Never commit API keys to version control
2. Use environment variables for all sensitive credentials
3. Rotate keys regularly and update environment variables
4. Use separate keys for development and production

Network Security

1. Use HTTPS in production with proper TLS certificates
2. Implement authentication and authorization layers
3. Configure firewalls to restrict access to the gateway
4. Monitor and log all gateway requests for audit trails

Configuration Security

# Secure configuration example
endpoints:
  - name: production-chat
    endpoint_type: llm/v1/chat
    model:
      provider: openai
      name: gpt-4
      config:
        openai_api_key: $OPENAI_API_KEY  # From environment
    limit:
      renewal_period: minute
      calls: 1000

Next Steps

Now that your providers are configured, learn how to use and integrate your gateway:

Usage Guide

Query endpoints with Python client and REST APIs

Start using →

Integration Guide

Integrate with applications, frameworks, and production systems

Learn integrations →

Tutorial

Step-by-step walkthrough with examples

Follow tutorial →