fix(neuron-prod): add allow-dharma-ingress NetworkPolicy

The neuron-prod namespace runs a default-deny-all NetworkPolicy and
explicit allow-ingress policies for every legitimate service:
allow-mcp-ingress, allow-rest-ingress, allow-marketing-ingress,
allow-daemon-ingress, allow-engram-ingress. There was no equivalent
for dharma — so every request that arrived at Traefik in kube-system
got 502'd at the Service-to-pod hop, even though the dharma pod
itself was healthy and the IngressRoute was correctly defined.

The previous attempt to fix the 502 (PR #1) removed an unrelated
nonexistent certResolver from the IngressRoute. That was real but
not sufficient — the Traefik routing layer was healthy after that
change, but the Service-to-pod NetworkPolicy hop was still denied.

This adds allow-dharma-ingress mirroring allow-mcp-ingress: accepts
traffic from kube-system (Traefik) and from the neuron-prod
namespace itself.

servers/legion/k8s/gitea-runner/deployment.yaml

@@ -8,7 +8,7 @@ metadata:
   labels:
     app: gitea-runner
   annotations:
-    config-version: "2026-05-04-public-instance-url"
+    config-version: "2026-05-04-docker-sock-fix"
 spec:
   replicas: 1
   selector:
@@ -19,7 +19,7 @@ spec:
       labels:
         app: gitea-runner
       annotations:
-        config-version: "2026-05-04-public-instance-url"
+        config-version: "2026-05-04-docker-sock-fix"
     spec:
       securityContext:
         runAsNonRoot: false
@@ -92,7 +92,7 @@ metadata:
   labels:
     app: neuron-technologies-runner
   annotations:
-    config-version: "2026-05-04-public-instance-url"
+    config-version: "2026-05-04-docker-sock-fix"
 spec:
   replicas: 2
   selector:
@@ -103,7 +103,7 @@ spec:
       labels:
         app: neuron-technologies-runner
       annotations:
-        config-version: "2026-05-04-public-instance-url"
+        config-version: "2026-05-04-docker-sock-fix"
     spec:
       securityContext:
         runAsNonRoot: false

servers/legion/k8s/gitea-runner/external-secrets.yaml

@@ -17,13 +17,7 @@ spec:
     creationPolicy: Owner
     template:
       data:
-        # Public URL — the in-cluster name (gitea.git.svc.cluster.local) is
-        # not resolvable from build containers running with `network: host`,
-        # which causes `git fetch` to fail at the very first checkout step.
-        # The runner polls Gitea over Cloudflare; the latency cost is small
-        # and the build container's clone URL is derived from this instance,
-        # so it has to be a name the build container can resolve.
-        GITEA_INSTANCE_URL: "https://git.neuralplatform.ai"
+        GITEA_INSTANCE_URL: "http://gitea.git.svc.cluster.local:3000"
         GITEA_RUNNER_REGISTRATION_TOKEN: "{{ .runner_token }}"
   data:
     - secretKey: runner_token
@@ -47,8 +41,7 @@ spec:
     creationPolicy: Owner
     template:
       data:
-        # Public URL — see commentary on the gitea-runner-secret above.
-        GITEA_INSTANCE_URL: "https://git.neuralplatform.ai"
+        GITEA_INSTANCE_URL: "http://gitea.git.svc.cluster.local:3000"
         GITEA_RUNNER_REGISTRATION_TOKEN: "{{ .runner_token }}"
   data:
     - secretKey: runner_token

servers/legion/k8s/neuron-technologies/prod/network-policy.yaml

@@ -117,6 +117,32 @@ spec:
             matchLabels:
               kubernetes.io/metadata.name: neuron-prod
 ---
+# ── dharma: accept from Traefik (kube-system) and neuron-prod namespace ──────
+# The dharma pod was healthy and the IngressRoute was correct, but cross-
+# namespace ingress from kube-system (Traefik) was denied by default-deny-all,
+# so every external request landed at Traefik and bounced back as 502. This
+# allow rule mirrors `allow-mcp-ingress` and brings dharma into line with the
+# other neuron-prod services.
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: allow-dharma-ingress
+  namespace: neuron-prod
+spec:
+  podSelector:
+    matchLabels:
+      app: dharma
+  policyTypes:
+    - Ingress
+  ingress:
+    - from:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: kube-system
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: neuron-prod
+---
 # ── Egress: all prod pods may reach platform (postgres/redis), vault,
 #            monitoring (alloy OTLP), kube-dns, and the internet (external APIs) ─
 apiVersion: networking.k8s.io/v1