feat(helm): add Helm chart for Presidio text services

docs/samples/deployments/k8s/charts/presidio/.helmignore

@@ -14,8 +14,12 @@
 *.swp
 *.bak
 *.tmp
+*.orig
 *~
 # Various IDEs
 .project
 .idea/
 *.tmproj
+.vscode/
+# CI/test artifacts
+ci/

docs/samples/deployments/k8s/charts/presidio/Chart.yaml

@@ -1,5 +1,20 @@
 apiVersion: v2
-description: A context aware, born to the cloud, customizable data loss prevention service
 name: presidio
-version: 2.0
-appVersion: latest
+description: Presidio services (analyzer + anonymizer, optional image-redactor) — context-aware PII detection, anonymization, and image redaction
+type: application
+home: https://github.com/microsoft/presidio
+sources:
+  - https://github.com/microsoft/presidio
+keywords:
+  - presidio
+  - pii
+  - anonymization
+  - data-loss-prevention
+  - privacy
+maintainers:
+  - name: Presidio Maintainers
+    url: https://github.com/microsoft/presidio
+# Chart version — bump on any chart change (SemVer).
+version: 0.2.0
+# Tracks the Presidio app release this chart targets. Quoted because "latest" is not SemVer.
+appVersion: "latest"

docs/samples/deployments/k8s/charts/presidio/README.md

@@ -0,0 +1,134 @@
+# Presidio Helm chart
+
+Deploys the [Presidio](https://github.com/microsoft/presidio) stack to Kubernetes — the
+`presidio-analyzer` (PII detection) and `presidio-anonymizer` (PII anonymization) text
+services, plus an optional `presidio-image-redactor` (PII redaction in images). It mirrors the
+`docker-compose-text.yml` / `docker-compose-image.yml` topology and is scaffolded from
+`helm create` following community best practices.
+
+> Scope: the analyzer and anonymizer are enabled by default. The image-redactor is **disabled
+> by default** (`imageRedactor.enabled=true` to opt in) — it bundles spaCy NLP models and
+> Tesseract OCR, so it is heavy and slow to start. The transformers NLP engine and Ollama
+> remain out of scope for this chart.
+
+## Prerequisites
+
+- Kubernetes 1.23+
+- Helm 3.8+
+- An Ingress controller (only if `ingress.enabled=true`); the default path rules assume
+  [ingress-nginx](https://kubernetes.github.io/ingress-nginx/)
+
+## Install
+
+```bash
+# From this chart directory: docs/samples/deployments/k8s/charts/presidio
+helm install presidio . --namespace presidio --create-namespace
+
+# Override the image tag (defaults to the chart appVersion)
+helm install presidio . --namespace presidio --create-namespace --set image.tag=latest
+
+# Custom values
+helm install presidio . --namespace presidio --create-namespace -f my-values.yaml
+```
+
+Check rollout (the analyzer loads NLP models on first start and can take a few minutes):
+
+```bash
+kubectl rollout status deployment/presidio-presidio-analyzer -n presidio
+helm test presidio -n presidio
+```
+
+## Uninstall
+
+```bash
+helm uninstall presidio
+```
+
+## What gets deployed
+
+| Resource | analyzer | anonymizer | image-redactor |
+| --- | --- | --- | --- |
+| Enabled by default | ✅ | ✅ | ❌ (`imageRedactor.enabled`) |
+| Deployment | ✅ | ✅ | ✅ |
+| Service (ClusterIP) | ✅ | ✅ | ✅ |
+| HorizontalPodAutoscaler | optional (`analyzer.autoscaling.enabled`) | optional (`anonymizer.autoscaling.enabled`) | optional (`imageRedactor.autoscaling.enabled`) |
+| ServiceAccount (shared) | ✅ (`serviceAccount.create`) | | |
+| Ingress (shared, routes all enabled) | optional (`ingress.enabled`) | | |
+
+Pods run as the non-root UID `1001` baked into the Presidio images, with `ALL` capabilities
+dropped, `allowPrivilegeEscalation: false`, and the `RuntimeDefault` seccomp profile.
+
+Enable the image-redactor with:
+
+```bash
+helm install presidio . --namespace presidio --create-namespace --set imageRedactor.enabled=true
+```
+
+## Configuration
+
+### Global
+
+| Key | Default | Description |
+| --- | --- | --- |
+| `image.registry` | `mcr.microsoft.com` | Registry for both images. Set `""` for bare repos. |
+| `image.pullPolicy` | `IfNotPresent` | Default pull policy. |
+| `image.tag` | `""` | Default tag; empty falls back to chart `appVersion`. |
+| `imagePullSecrets` | `[]` | Private registry pull secrets. |
+| `serviceAccount.create` | `true` | Create a shared ServiceAccount. |
+| `serviceAccount.automount` | `false` | Mount the SA token (Presidio does not call the K8s API). |
+| `podSecurityContext` | non-root 1001 | Pod-level security context for all components. |
+| `securityContext` | drop ALL caps | Container-level security context for all components. |
+| `nodeSelector` / `tolerations` / `affinity` | empty | Default scheduling; per-component overrides supported. |
+| `ingress.enabled` | `false` | Enable the shared Ingress. |
+| `ingress.className` | `nginx` | IngressClass. |
+| `ingress.hosts` | `presidio.local` | Host/path rules; each path maps to a `service` alias (`analyzer`, `anonymizer`, `image-redactor`). |
+
+### Per component (`analyzer.*`, `anonymizer.*`, `imageRedactor.*`)
+
+| Key | analyzer default | anonymizer default | imageRedactor default | Description |
+| --- | --- | --- | --- | --- |
+| `enabled` | `true` | `true` | `false` | Toggle the component. |
+| `image.repository` | `presidio-analyzer` | `presidio-anonymizer` | `presidio-image-redactor` | Image repo (joined with `image.registry`). |
+| `replicaCount` | `1` | `1` | `1` | Replicas (ignored when autoscaling is on). |
+| `containerPort` | `3000` | `3000` | `3000` | Listen port; exported as the `PORT` env var. |
+| `workers` | `1` | `1` | `1` | gunicorn workers; exported as the `WORKERS` env var. |
+| `extraEnv` | `[]` | `[]` | `[]` | Extra environment variables. |
+| `service.type` / `service.port` | `ClusterIP` / `8080` | `ClusterIP` / `8080` | `ClusterIP` / `8080` | Service exposure. |
+| `resources` | 1.5–3Gi / 1.5–2 CPU | 128–512Mi / 0.125–0.5 CPU | 1.5–3Gi / 1.5–2 CPU | Requests/limits. |
+| `startupProbe` / `livenessProbe` / `readinessProbe` | `/health` | `/health` | `/health` | HTTP probes. |
+| `autoscaling.enabled` | `false` | `false` | `false` | Enable the HPA. |
+
+> The image-redactor exposes `POST /redact` (multipart image upload) in addition to `/health`.
+> Note the [`docker-compose-image.yml`](../../../../../../docker-compose-image.yml) reference sets
+> `PORT=5001`, but this chart standardizes every component on container port `3000` (the image's
+> Dockerfile default).
+
+## Scaling caveat
+
+Balancing traffic across **more than one replica** of a service requires a service mesh
+(e.g. Linkerd or Istio) because the clients hold persistent connections — see
+[microsoft/presidio#304](https://github.com/microsoft/presidio/issues/304). Plain
+`replicaCount > 1` or HPA without a mesh may not distribute load evenly.
+
+## Example: enable Ingress
+
+```yaml
+ingress:
+  enabled: true
+  className: nginx
+  annotations:
+    nginx.ingress.kubernetes.io/rewrite-target: /$2
+  hosts:
+    - host: presidio.example.com
+      paths:
+        - path: /analyzer(/|$)(.*)
+          pathType: ImplementationSpecific
+          service: analyzer
+        - path: /anonymizer(/|$)(.*)
+          pathType: ImplementationSpecific
+          service: anonymizer
+        # Only routed when imageRedactor.enabled is true.
+        - path: /image-redactor(/|$)(.*)
+          pathType: ImplementationSpecific
+          service: image-redactor
+```

docs/samples/deployments/k8s/charts/presidio/templates/NOTES.txt

@@ -1,9 +1,42 @@
-presidio is now installed!
+Presidio has been deployed. 🎉
 
-To find out about your newly configured system, run:
+Release:   {{ .Release.Name }}
+Namespace: {{ .Release.Namespace }}
 
-  $ helm status {{ .Release.Name }}
+Services deployed:
+{{- if .Values.analyzer.enabled }}
+  - presidio-analyzer  → {{ include "presidio.componentFullname" (dict "root" $ "name" "analyzer") }}:{{ .Values.analyzer.service.port }}
+{{- end }}
+{{- if .Values.anonymizer.enabled }}
+  - presidio-anonymizer → {{ include "presidio.componentFullname" (dict "root" $ "name" "anonymizer") }}:{{ .Values.anonymizer.service.port }}
+{{- end }}
+{{- if .Values.imageRedactor.enabled }}
+  - presidio-image-redactor → {{ include "presidio.componentFullname" (dict "root" $ "name" "image-redactor") }}:{{ .Values.imageRedactor.service.port }}
+{{- end }}
 
-To test your new Presidio service, forward the ports: the Analyzer publishes a service at port 8080. To use the Analyzer, you first need to forward this port to your local IP address:
+NOTE: the analyzer{{ if .Values.imageRedactor.enabled }} and image-redactor{{ end }} load NLP models on first start and can
+take a few minutes to become Ready. Watch rollout with:
 
-  $ kubectl port-forward -n presidio $(kubectl get pod -n presidio -l app=demo-presidio-analyzer -o jsonpath='{.items[0].metadata.name}') 8080:8080
+  kubectl --namespace {{ .Release.Namespace }} rollout status deployment/{{ include "presidio.componentFullname" (dict "root" $ "name" "analyzer") }}
+
+{{- if .Values.ingress.enabled }}
+
+Reach the services via the configured Ingress:
+{{- range $host := .Values.ingress.hosts }}
+  {{- range $host.paths }}
+  http{{ if $.Values.ingress.tls }}s{{ end }}://{{ $host.host }}{{ .path }}  → {{ .service }}
+  {{- end }}
+{{- end }}
+{{- else }}
+
+No Ingress is enabled. Test a service from your machine with port-forward, e.g. the anonymizer:
+
+  kubectl --namespace {{ .Release.Namespace }} port-forward svc/{{ include "presidio.componentFullname" (dict "root" $ "name" "anonymizer") }} {{ .Values.anonymizer.service.port }}:{{ .Values.anonymizer.service.port }}
+  curl http://localhost:{{ .Values.anonymizer.service.port }}/health
+
+Then anonymize a sample:
+
+  curl -X POST http://localhost:{{ .Values.anonymizer.service.port }}/anonymize \
+    -H "Content-Type: application/json" \
+    -d '{"text":"My name is Bond.","anonymizers":{"DEFAULT":{"type":"replace","new_value":"ANONYMIZED"}},"analyzer_results":[{"start":11,"end":15,"score":0.8,"entity_type":"PERSON"}]}'
+{{- end }}