# Responses API Request Validation
# Requires `--features openai-responses` because these filters are opt-in.
#
# Validates Responses API JSON and enriches request metadata.
#
# Build:
#   cargo build -p praxis-ai-proxy --features openai-responses
#
# openai_responses_request recognizes the Responses create operation from
# the request head, then deserializes the body exactly once. That single
# parse produces every downstream fact: it buffers the body, classifies the
# request as Responses API, promotes model/stream/store/background to
# openai_responses_format.* metadata, extracts conversation.id when present,
# generates response and conversation IDs, and writes responses.* metadata
# for downstream pipeline filters.
#
# Valid supported JSON requests are forwarded to the upstream backend with
# the original body unchanged. Invalid JSON and unsupported background-mode
# requests receive a 400 error response.
#
# Example valid request:
#
#   curl -X POST http://localhost:8080/v1/responses \
#     -H "Content-Type: application/json" \
#     -d '{"model":"gpt-4.1","input":"Hello, world!"}'
#
# Example unsupported background request (rejected before the backend):
#
#   curl -X POST http://localhost:8080/v1/responses \
#     -H "Content-Type: application/json" \
#     -d '{"model":"gpt-4.1","input":"test","stream":true,"background":true}'
#
listeners:
  - name: responses-gateway
    address: "127.0.0.1:8080"
    filter_chains:
      - validate-and-route

filter_chains:
  - name: validate-and-route
    filters:
      - filter: openai_responses_request
        on_invalid: reject
      - filter: router
        routes:
          - path: "/v1/responses"
            cluster: inference
      - filter: load_balancer
        clusters:
          - name: inference
            endpoints:
              - "127.0.0.1:8000"

insecure_options:
  allow_private_endpoints: true # example proxies to local backends