Make sandbox runner pluggable with Kubernetes driver support

.env.example

@@ -96,6 +96,12 @@ CONTAINER_LLM_MODEL=
 # Increase if you see truncated specs/PRDs (default 4096 is often too small)
 LLM_MAX_TOKENS=16384
 
+# Optional provider-neutral per-stage policy (JSON). Credentials remain in the
+# provider-native variables above; credential_ref is an identifier, never a key.
+# MODEL_CONNECTIONS={"vertex-prod":{"backend":"vertex-ai","credential_ref":"gcp-adc","project":"your-gcp-project-id","location":"global","allowed_models":["gemini-3.5-pro","gemini-3.5-flash"]}}
+# MODEL_DEFAULT={"connection":"vertex-prod","model":"gemini-3.5-flash"}
+# MODEL_POLICY={"generate_prd":{"connection":"vertex-prod","model":"gemini-3.5-pro"}}
+
 # =============================================================================
 # MCP Servers Configuration
 # =============================================================================
@@ -217,6 +223,20 @@ CONTAINER_LANGCHAIN_VERBOSE=false
 # inspect commands (podman logs, podman export). Remember to clean up with podman rm.
 FORGE_CONTAINER_KEEP=false
 
+# Kubernetes sandbox driver (used when the Forge worker runs in Kubernetes).
+# The worker and sandbox Jobs must mount the same workspace PVC.
+# FORGE_SANDBOX_DRIVER=kubernetes
+# WORKSPACE_BASE_DIR=/mnt/workspaces
+# K8S_NAMESPACE=forge
+# K8S_WORKSPACE_PVC=forge-workspaces
+# K8S_WORKSPACE_BASE_PATH=/mnt/workspaces
+# K8S_IMAGE_PULL_SECRETS=
+# K8S_SERVICE_ACCOUNT=forge-runner
+# Optional Google ADC Secret mounted by the Kubernetes driver into sandbox Jobs:
+# K8S_GOOGLE_CREDENTIALS_SECRET=google-adc
+# K8S_GOOGLE_CREDENTIALS_KEY=forge-gcp-credentials.json
+# K8S_GOOGLE_CREDENTIALS_MOUNT_PATH=/etc/forge-gcp-credentials.json
+
 # =============================================================================
 # Git Configuration (for commits made by Forge)
 # =============================================================================

Dockerfile

@@ -16,7 +16,7 @@ COPY pyproject.toml README.md ./
 COPY src/ ./src/
 
 RUN pip install --no-cache-dir build && \
-    pip wheel --no-cache-dir --wheel-dir /wheels -e .
+    pip wheel --no-cache-dir --wheel-dir /wheels -e ".[k8s]"
 
 # Production stage
 FROM python:3.11-slim

charts/forge/Chart.yaml

@@ -0,0 +1,6 @@
+apiVersion: v2
+name: forge
+description: Forge API and Kubernetes sandbox worker
+type: application
+version: 0.1.0
+appVersion: "0.1.0"

charts/forge/README.md

@@ -0,0 +1,39 @@
+# Forge Helm chart
+
+This chart deploys the Forge API, worker, Redis Stack, worker RBAC, and the
+workspace PVC used by Kubernetes sandbox Jobs. It targets OpenShift by default.
+
+Create the namespace and application secret before installing. Keep secrets out
+of Helm values and source control. Values consumed as JSON, such as
+`MODEL_CONNECTIONS`, must contain raw JSON in the env file; do not surround
+them with shell quotes because `oc --from-env-file` preserves those quotes.
+
+```bash
+oc new-project forge
+oc create secret generic forge-env --from-env-file=.env -n forge
+helm upgrade --install forge charts/forge -n forge
+```
+
+For Vertex AI, create a Secret from the same service-account credential used by
+the existing Forge deployment and enable the optional mount. The chart mounts
+the credential into both the worker and every sandbox Job:
+
+```bash
+oc create secret generic google-adc \
+  --from-file=forge-gcp-credentials.json=/path/to/forge-gcp-credentials.json \
+  -n forge
+helm upgrade --install forge charts/forge -n forge \
+  --set googleCredentials.enabled=true
+```
+
+Build and push two images before installing:
+
+* `image.repository`: the Forge API/worker image built from `Dockerfile`
+* `sandboxImage.repository`: the task image built from `containers/Containerfile`
+
+Override repositories, tags, storage class, or Route settings in a local values
+file. The worker uses in-cluster service-account authentication; do not mount a
+kubeconfig into it.
+
+The workspace PVC is retained when the Helm release is removed. Delete it
+explicitly when its workspaces are no longer needed.

charts/forge/templates/_helpers.tpl

@@ -0,0 +1,13 @@
+{{- define "forge.name" -}}forge{{- end }}
+{{- define "forge.fullname" -}}{{ .Release.Name }}{{- end }}
+{{- define "forge.labels" -}}
+app.kubernetes.io/name: {{ include "forge.name" . }}
+app.kubernetes.io/instance: {{ .Release.Name }}
+app.kubernetes.io/managed-by: {{ .Release.Service }}
+{{- end }}
+{{- define "forge.image" -}}
+{{ printf "%s:%s" .Values.image.repository .Values.image.tag }}
+{{- end }}
+{{- define "forge.sandboxImage" -}}
+{{ printf "%s:%s" .Values.sandboxImage.repository .Values.sandboxImage.tag }}
+{{- end }}

charts/forge/templates/api.yaml

@@ -0,0 +1,80 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: {{ include "forge.fullname" . }}-api
+  labels:
+    {{- include "forge.labels" . | nindent 4 }}
+    app.kubernetes.io/component: api
+spec:
+  replicas: {{ .Values.api.replicas }}
+  selector:
+    matchLabels:
+      app.kubernetes.io/instance: {{ .Release.Name }}
+      app.kubernetes.io/component: api
+  template:
+    metadata:
+      annotations:
+        {{- toYaml .Values.podAnnotations | nindent 8 }}
+      labels:
+        app.kubernetes.io/instance: {{ .Release.Name }}
+        app.kubernetes.io/component: api
+    spec:
+      securityContext:
+        {{- toYaml .Values.podSecurityContext | nindent 8 }}
+      {{- with .Values.imagePullSecrets }}
+      imagePullSecrets:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
+      containers:
+        - name: api
+          image: {{ include "forge.image" . }}
+          imagePullPolicy: {{ .Values.image.pullPolicy }}
+          securityContext:
+            {{- toYaml .Values.containerSecurityContext | nindent 12 }}
+          envFrom:
+            - secretRef:
+                name: {{ .Values.existingSecret }}
+          env:
+            - name: REDIS_URL
+              value: redis://{{ include "forge.fullname" . }}-redis:6379/0
+          ports:
+            - name: http
+              containerPort: 8000
+          readinessProbe:
+            httpGet: {path: /api/v1/ready, port: http}
+          livenessProbe:
+            httpGet: {path: /api/v1/live, port: http}
+          resources:
+            {{- toYaml .Values.resources.api | nindent 12 }}
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: {{ include "forge.fullname" . }}-api
+spec:
+  selector:
+    app.kubernetes.io/instance: {{ .Release.Name }}
+    app.kubernetes.io/component: api
+  ports:
+    - name: http
+      port: {{ .Values.api.service.port }}
+      targetPort: http
+{{- if .Values.api.route.enabled }}
+---
+apiVersion: route.openshift.io/v1
+kind: Route
+metadata:
+  name: {{ include "forge.fullname" . }}
+spec:
+  {{- with .Values.api.route.host }}
+  host: {{ . }}
+  {{- end }}
+  to:
+    kind: Service
+    name: {{ include "forge.fullname" . }}-api
+  port:
+    targetPort: http
+  tls:
+    termination: {{ .Values.api.route.tlsTermination }}
+    insecureEdgeTerminationPolicy: Redirect
+{{- end }}

charts/forge/templates/rbac.yaml

@@ -0,0 +1,26 @@
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+  name: {{ include "forge.fullname" . }}-worker
+rules:
+  - apiGroups: ["batch"]
+    resources: ["jobs"]
+    verbs: ["create", "get", "list", "watch", "delete"]
+  - apiGroups: [""]
+    resources: ["pods"]
+    verbs: ["get", "list", "watch"]
+  - apiGroups: [""]
+    resources: ["pods/log"]
+    verbs: ["get"]
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+  name: {{ include "forge.fullname" . }}-worker
+subjects:
+  - kind: ServiceAccount
+    name: {{ .Values.worker.serviceAccountName }}
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: Role
+  name: {{ include "forge.fullname" . }}-worker

charts/forge/templates/redis.yaml

@@ -0,0 +1,65 @@
+{{- if .Values.redis.enabled }}
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: {{ include "forge.fullname" . }}-redis
+  labels:
+    {{- include "forge.labels" . | nindent 4 }}
+    app.kubernetes.io/component: redis
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app.kubernetes.io/instance: {{ .Release.Name }}
+      app.kubernetes.io/component: redis
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/instance: {{ .Release.Name }}
+        app.kubernetes.io/component: redis
+    spec:
+      containers:
+        - name: redis
+          image: {{ .Values.redis.image | quote }}
+          ports:
+            - name: redis
+              containerPort: 6379
+          volumeMounts:
+            - name: data
+              mountPath: /data
+      volumes:
+        - name: data
+          {{- if .Values.redis.persistence.enabled }}
+          persistentVolumeClaim:
+            claimName: {{ include "forge.fullname" . }}-redis
+          {{- else }}
+          emptyDir: {}
+          {{- end }}
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: {{ include "forge.fullname" . }}-redis
+spec:
+  selector:
+    app.kubernetes.io/instance: {{ .Release.Name }}
+    app.kubernetes.io/component: redis
+  ports:
+    - name: redis
+      port: 6379
+{{- if .Values.redis.persistence.enabled }}
+---
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: {{ include "forge.fullname" . }}-redis
+spec:
+  accessModes: ["ReadWriteOnce"]
+  {{- with .Values.redis.persistence.storageClassName }}
+  storageClassName: {{ . }}
+  {{- end }}
+  resources:
+    requests:
+      storage: {{ .Values.redis.persistence.size }}
+{{- end }}
+{{- end }}

charts/forge/templates/serviceaccounts.yaml

@@ -0,0 +1,13 @@
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: {{ .Values.worker.serviceAccountName }}
+  labels:
+    {{- include "forge.labels" . | nindent 4 }}
+---
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: {{ .Values.sandbox.jobServiceAccountName }}
+  labels:
+    {{- include "forge.labels" . | nindent 4 }}