Replace hand-applied llm-serving-default-deny NetworkPolicy that used
llm-client=true pod label selector (any pod in any namespace could
self-grant access) with a proper namespace-scoped policy.
Ingress now restricted to:
- api namespace (gateway) on ports 8080/80/8000/11434
- monitoring namespace (Prometheus) on ports 8080/9000
- intra-namespace (pod-to-pod within llm-serving)
Tested live:
- Gateway -> reasoning/ornith/embeddings/reranker: 200 OK
- default namespace -> llm-serving: timeout (blocked)
- portfolio namespace -> llm-serving: timeout (blocked)
Closes#13
reasoning keeps its 2 GPUs untouched. verifier's freed GPU goes to a second ornith replica instead of a standalone qwen-only pod -- both replicas load ornith:35b + qwen2.5:3b-instruct, k8s Service load-balances across them, so 2 concurrent implementer-style calls get independent instances.
Frees verifier's GPU from an underused vLLM PRM deployment. qwen2.5:3b-instruct moves off ornith-predictor's shared pod onto its own dedicated GPU (grm.yaml), so verification/judge traffic stops contending with ornith:35b's agent traffic. /v1/qwen/chat/completions now points at grm-predictor; path unchanged.
rock
Story Crater Bot