Skip to content

fix(envoy_ai_gateway): publish inference.<domain> in keycloak mode only #12

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Mushtaq-BGA wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fix(envoy_ai_gateway): publish inference.<domain> in keycloak mode only #12

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
33 changes: 29 additions & 4 deletions roles/envoy_ai_gateway/tasks/install.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -279,9 +279,16 @@
when: ai_gateway_create | bool

# ── External ingress: eg-gateway → ai-gateway HTTPRoute ──────────────────────
# Single HTTPRoute on eg-gateway that forwards all inference traffic to
# ai-gateway. Per-model AIGatewayRoutes/HTTPRoutes on ai-gateway handle
# model dispatch. Hostname: inference.<base_domain_name>.
# Single HTTPRoute on eg-gateway that forwards all inference traffic straight to
# ai-gateway. Per-model AIGatewayRoutes/HTTPRoutes on ai-gateway handle model
# dispatch. Hostname: inference.<base_domain_name>.
#
# Keycloak mode only. This route is the direct-to-ai-gateway path the gateway
# JWT policy guards. In litellm mode all inference must go through LiteLLM
# (litellm.<domain>) so it can validate the virtual key and apply budgets, rate
# limits and Langfuse logging; LiteLLM reaches ai-gateway over its own internal
# ClusterIP service, not this route. Publishing inference.<domain> in litellm
# mode would expose an unauthenticated bypass around LiteLLM.

- name: "envoy_ai_gateway | Create external inference HTTPRoute (eg-gateway → ai-gateway)"
kubernetes.core.k8s:
Expand Down Expand Up @@ -310,4 +317,22 @@
- name: "{{ ai_gateway_service_name }}"
namespace: "{{ envoy_gateway_namespace | default('envoy-gateway-system') }}"
port: 80
when: ai_gateway_create | bool
when:
- ai_gateway_create | bool
- auth_provider | default('keycloak') == 'keycloak'

# Remove the external route if the cluster was reconfigured from keycloak to
# litellm/none: leaving inference.<domain> published would keep an unauthenticated
# path to ai-gateway open once the gateway JWT policy is gone.
- name: "envoy_ai_gateway | Remove external inference HTTPRoute when not keycloak"
kubernetes.core.k8s:
state: absent
api_version: gateway.networking.k8s.io/v1
kind: HTTPRoute
name: inference-external
namespace: "{{ envoy_gateway_namespace | default('envoy-gateway-system') }}"
# Absent on any install that was never in keycloak mode — not an error.
failed_when: false
when:
- ai_gateway_create | bool
- auth_provider | default('keycloak') != 'keycloak'
Loading