02-ai-gateway.yml

# 02-ai-gateway.yml - AI gateway with mock OpenAI-compatible backend
#
# This example demonstrates the ai_proxy action, which provides an OpenAI-compatible
# API gateway. It supports provider routing, model selection, and fallback chains.
#
# The test backend at test.sbproxy.dev/v1 is a mock OpenAI server that accepts
# standard chat completion requests and returns synthetic responses.
#
# Start:  sbproxy serve -f examples/02-ai-gateway.yml
#
# List models:
#   curl -H "Host: ai.example.com" http://localhost:8080/v1/models
#
# Chat completion:
#   curl -H "Host: ai.example.com" \
#        -H "Content-Type: application/json" \
#        -X POST http://localhost:8080/v1/chat/completions \
#        -d '{"model":"gpt-4o-mini","messages":[{"role":"user","content":"Hello"}]}'
#
# With API key auth (see the authentication block below):
#   curl -H "Host: ai.example.com" \
#        -H "X-API-Key: dR7tN3mK9pL2vX5" \
#        -H "Content-Type: application/json" \
#        -X POST http://localhost:8080/v1/chat/completions \
#        -d '{"model":"gpt-4o-mini","messages":[{"role":"user","content":"Hello"}]}'

proxy:
  http_bind_port: 8080

origins:
  "ai.example.com":
    action:
      type: ai_proxy # AI gateway action - OpenAI-compatible API

      # Providers are the upstream AI backends. sbproxy routes requests to
      # the appropriate provider based on the requested model name.
      providers:
        - name: openai-mock         # Logical name for this provider
          api_key: dR7tN3mK9pL2         # API key sent to the backend (use ${OPENAI_API_KEY} for real keys)
          base_url: https://test.sbproxy.dev/v1 # Override endpoint - points to mock server
          models:
            - gpt-4o                # Models this provider handles
            - gpt-4o-mini

      default_model: gpt-4o-mini   # Fallback model when none is specified in the request

      routing:
        strategy: fallback_chain   # Try providers in order; fall back to next on failure

    # Protect the AI gateway with an API key
    authentication:
      type: api_key                # Require X-API-Key header
      api_keys:
        - dR7tN3mK9pL2vX5             # Accepted key values
        - another-valid-key

    # Add rate limiting to control costs
    policies:
      - type: rate_limiting
        requests_per_minute: 60    # Max 60 requests per minute per client IP
        algorithm: sliding_window  # More accurate than fixed window

    # Inject headers so clients can see which provider was used
    response_modifiers:
      - headers:
          set:
            X-AI-Gateway: sbproxy    # Identify this proxy in responses