From d3e1226d15d07884d5037de34b4343e4e827ba91 Mon Sep 17 00:00:00 2001
From: justin-tahara <justintahara@gmail.com>
Date: Fri, 29 Aug 2025 17:01:14 -0700
Subject: [PATCH 1/2] feat(infra): Adding new KEDA Helm templates

---
 deployment/helm/charts/onyx/Chart.yaml        |  2 +-
 .../keda/api-server-scaledobject.yaml         | 24 +++++++++
 .../celery-worker-common-scaledobject.yaml    | 51 +++++++++++++++++++
 ...elery-worker-docfetching-scaledobject.yaml | 46 +++++++++++++++++
 ...ery-worker-docprocessing-scaledobject.yaml | 46 +++++++++++++++++
 .../keda/model-server-scaledobject.yaml       | 31 +++++++++++
 .../templates/keda/slackbot-scaledobject.yaml | 28 ++++++++++
 .../keda/web-server-scaledobject.yaml         | 24 +++++++++
 8 files changed, 251 insertions(+), 1 deletion(-)
 create mode 100644 deployment/helm/charts/onyx/templates/keda/api-server-scaledobject.yaml
 create mode 100644 deployment/helm/charts/onyx/templates/keda/celery-worker-common-scaledobject.yaml
 create mode 100644 deployment/helm/charts/onyx/templates/keda/celery-worker-docfetching-scaledobject.yaml
 create mode 100644 deployment/helm/charts/onyx/templates/keda/celery-worker-docprocessing-scaledobject.yaml
 create mode 100644 deployment/helm/charts/onyx/templates/keda/model-server-scaledobject.yaml
 create mode 100644 deployment/helm/charts/onyx/templates/keda/slackbot-scaledobject.yaml
 create mode 100644 deployment/helm/charts/onyx/templates/keda/web-server-scaledobject.yaml

diff --git a/deployment/helm/charts/onyx/Chart.yaml b/deployment/helm/charts/onyx/Chart.yaml
index 40cbbbbc9f..a4f2ca07e2 100644
--- a/deployment/helm/charts/onyx/Chart.yaml
+++ b/deployment/helm/charts/onyx/Chart.yaml
@@ -5,7 +5,7 @@ home: https://www.onyx.app/
 sources:
   - "https://github.com/onyx-dot-app/onyx"
 type: application
-version: 0.2.9
+version: 0.2.10
 appVersion: latest
 annotations:
   category: Productivity
diff --git a/deployment/helm/charts/onyx/templates/keda/api-server-scaledobject.yaml b/deployment/helm/charts/onyx/templates/keda/api-server-scaledobject.yaml
new file mode 100644
index 0000000000..58951173c7
--- /dev/null
+++ b/deployment/helm/charts/onyx/templates/keda/api-server-scaledobject.yaml
@@ -0,0 +1,24 @@
+{{- if and .Values.keda.enabled .Values.keda.apiServer .Values.keda.apiServer.enabled }}
+apiVersion: keda.sh/v1alpha1
+kind: ScaledObject
+metadata:
+  name: {{ include "onyx-stack.fullname" . }}-api-server-scaledobject
+  namespace: {{ .Release.Namespace }}
+  labels:
+    {{- include "onyx-stack.labels" . | nindent 4 }}
+    app: api-server
+spec:
+  scaleTargetRef:
+    apiVersion: apps/v1
+    kind: Deployment
+    name: {{ include "onyx-stack.fullname" . }}-api-server
+  pollingInterval: {{ .Values.keda.apiServer.pollingInterval | default 30 }}
+  cooldownPeriod: {{ .Values.keda.apiServer.cooldownPeriod | default 300 }}
+  minReplicaCount: {{ .Values.keda.apiServer.minReplicas | default 1 }}
+  maxReplicaCount: {{ .Values.keda.apiServer.maxReplicas | default 10 }}
+  triggers:
+    - type: cpu
+      metadata:
+        type: Utilization
+        value: {{ .Values.keda.apiServer.cpuThreshold | default "70" | quote }}
+{{- end }}
diff --git a/deployment/helm/charts/onyx/templates/keda/celery-worker-common-scaledobject.yaml b/deployment/helm/charts/onyx/templates/keda/celery-worker-common-scaledobject.yaml
new file mode 100644
index 0000000000..a38cc46b42
--- /dev/null
+++ b/deployment/helm/charts/onyx/templates/keda/celery-worker-common-scaledobject.yaml
@@ -0,0 +1,51 @@
+{{- if and .Values.keda.enabled .Values.keda.celeryWorkers.enabled }}
+{{- range $workerType, $workerConfig := .Values.keda.celeryWorkers }}
+{{- if and (ne $workerType "enabled") $workerConfig.enabled (ne $workerType "docprocessing") (ne $workerType "docfetching") }}
+---
+apiVersion: keda.sh/v1alpha1
+kind: ScaledObject
+metadata:
+  name: {{ include "onyx-stack.fullname" $ }}-celery-worker-{{ $workerType }}-scaledobject
+  namespace: {{ $.Release.Namespace }}
+  labels:
+    {{- include "onyx-stack.labels" $ | nindent 4 }}
+    app: celery-worker-{{ $workerType }}
+spec:
+  scaleTargetRef:
+    apiVersion: apps/v1
+    kind: Deployment
+    name: {{ include "onyx-stack.fullname" $ }}-celery-worker-{{ $workerType }}
+  pollingInterval: {{ $workerConfig.pollingInterval | default 30 }}
+  cooldownPeriod: {{ $workerConfig.cooldownPeriod | default 300 }}
+  minReplicaCount: {{ $workerConfig.minReplicas | default 1 }}
+  maxReplicaCount: {{ $workerConfig.maxReplicas | default 10 }}
+  triggers:
+    {{- if $workerConfig.triggers }}
+    # Default Prometheus-based trigger for Redis queue depth if none specified
+    # Scaling Logic:
+    # - When queue depth > 5: Scale up by factor of 2 (moderate scaling)
+    # - When queue depth <= 5: Scale down by factor of 0.5 (conservative scaling)
+    # - Threshold of 1 ensures scaling triggers when metric value > 1
+    - type: prometheus
+      metadata:
+        serverAddress: "http://prometheus-redis.monitoring.svc.cluster.local:9090"
+        metricName: "redis_key_size_sum"
+        metricType: "Value"
+        threshold: "1"
+        query: |
+          # Simplified scaling logic for generic celery workers
+          # Returns 2 when queue depth > 5, 0.5 when <= 5
+          # This creates a clear scaling decision boundary
+          (
+            (sum(redis_key_size{key=~"connector_{{ $workerType }}.*"}) > 5)
+              * 2
+          )
+          +
+          (
+            (sum(redis_key_size{key=~"connector_{{ $workerType }}.*"}) <= 5)
+              * 0.5
+          )
+    {{- end }}
+{{- end }}
+{{- end }}
+{{- end }}
diff --git a/deployment/helm/charts/onyx/templates/keda/celery-worker-docfetching-scaledobject.yaml b/deployment/helm/charts/onyx/templates/keda/celery-worker-docfetching-scaledobject.yaml
new file mode 100644
index 0000000000..f8d52a2c59
--- /dev/null
+++ b/deployment/helm/charts/onyx/templates/keda/celery-worker-docfetching-scaledobject.yaml
@@ -0,0 +1,46 @@
+{{- if and .Values.keda.enabled .Values.keda.celeryWorkers.docfetching .Values.keda.celeryWorkers.docfetching.enabled }}
+apiVersion: keda.sh/v1alpha1
+kind: ScaledObject
+metadata:
+  name: {{ include "onyx-stack.fullname" . }}-celery-worker-docfetching-scaledobject
+  namespace: {{ .Release.Namespace }}
+  labels:
+    {{- include "onyx-stack.labels" . | nindent 4 }}
+    app: celery-worker-docfetching
+spec:
+  scaleTargetRef:
+    apiVersion: apps/v1
+    kind: Deployment
+    name: {{ include "onyx-stack.fullname" . }}-celery-worker-docfetching
+  pollingInterval: {{ .Values.keda.celeryWorkers.docfetching.pollingInterval | default 30 }}
+  cooldownPeriod: {{ .Values.keda.celeryWorkers.docfetching.cooldownPeriod | default 300 }}
+  minReplicaCount: {{ .Values.keda.celeryWorkers.docfetching.minReplicas | default 1 }}
+  maxReplicaCount: {{ .Values.keda.celeryWorkers.docfetching.maxReplicas | default 10 }}
+  triggers:
+    {{- if .Values.keda.celeryWorkers.docfetching.triggers }}
+    # Default Prometheus-based trigger for Redis queue depth if none specified
+    # Scaling Logic: 
+    # - When queue depth > 5: Scale up by factor of 2 (aggressive scaling)
+    # - When queue depth <= 5: Scale down by factor of 0.5 (conservative scaling)
+    # - Threshold of 1 ensures scaling triggers when metric value > 1
+    - type: prometheus
+      metadata:
+        serverAddress: "http://prometheus-redis.monitoring.svc.cluster.local:9090"
+        metricName: "redis_key_size_sum"
+        metricType: "Value"
+        threshold: "1"
+        query: |
+          # Simplified scaling logic for docfetching workers
+          # Returns 2 when queue depth > 5, 0.5 when <= 5
+          # This creates a clear scaling decision boundary
+          (
+            (sum(redis_key_size{key=~"connector_docfetching.*"}) > 5)
+              * 2
+          )
+          +
+          (
+            (sum(redis_key_size{key=~"connector_docfetching.*"}) <= 5)
+              * 0.5
+          )
+    {{- end }}
+{{- end }}
diff --git a/deployment/helm/charts/onyx/templates/keda/celery-worker-docprocessing-scaledobject.yaml b/deployment/helm/charts/onyx/templates/keda/celery-worker-docprocessing-scaledobject.yaml
new file mode 100644
index 0000000000..2bd2e1a4b8
--- /dev/null
+++ b/deployment/helm/charts/onyx/templates/keda/celery-worker-docprocessing-scaledobject.yaml
@@ -0,0 +1,46 @@
+{{- if and .Values.keda.enabled .Values.keda.celeryWorkers.docprocessing .Values.keda.celeryWorkers.docprocessing.enabled }}
+apiVersion: keda.sh/v1alpha1
+kind: ScaledObject
+metadata:
+  name: {{ include "onyx-stack.fullname" . }}-celery-worker-docprocessing-scaledobject
+  namespace: {{ .Release.Namespace }}
+  labels:
+    {{- include "onyx-stack.labels" . | nindent 4 }}
+    app: celery-worker-docprocessing
+spec:
+  scaleTargetRef:
+    apiVersion: apps/v1
+    kind: Deployment
+    name: {{ include "onyx-stack.fullname" . }}-celery-worker-docprocessing
+  pollingInterval: {{ .Values.keda.celeryWorkers.docprocessing.pollingInterval | default 30 }}
+  cooldownPeriod: {{ .Values.keda.celeryWorkers.docprocessing.cooldownPeriod | default 300 }}
+  minReplicaCount: {{ .Values.keda.celeryWorkers.docprocessing.minReplicas | default 1 }}
+  maxReplicaCount: {{ .Values.keda.celeryWorkers.docprocessing.maxReplicas | default 50 }}
+  triggers:
+    {{- if .Values.keda.celeryWorkers.docprocessing.triggers }}
+    # Default Prometheus-based trigger for Redis queue depth if none specified
+    # Scaling Logic:
+    # - When queue depth > 20: Scale up by factor of 4 (very aggressive scaling)
+    # - When queue depth <= 20: Scale down by factor of 0.25 (very conservative scaling)
+    # - Threshold of 1 ensures scaling triggers when metric value > 1
+    - type: prometheus
+      metadata:
+        serverAddress: "http://prometheus-redis.monitoring.svc.cluster.local:9090"
+        metricName: "redis_key_size_sum"
+        metricType: "Value"
+        threshold: "1"
+        query: |
+          # Simplified scaling logic for docprocessing workers
+          # Returns 4 when queue depth > 20, 0.25 when <= 20
+          # This creates a clear scaling decision boundary for high-volume processing
+          (
+            (sum(redis_key_size{key=~"connector_docprocessing.*"}) > 20)
+              * 4
+          )
+          +
+          (
+            (sum(redis_key_size{key=~"connector_docprocessing.*"}) <= 20)
+              * 0.25
+          )
+    {{- end }}
+{{- end }}
diff --git a/deployment/helm/charts/onyx/templates/keda/model-server-scaledobject.yaml b/deployment/helm/charts/onyx/templates/keda/model-server-scaledobject.yaml
new file mode 100644
index 0000000000..f54ab3086c
--- /dev/null
+++ b/deployment/helm/charts/onyx/templates/keda/model-server-scaledobject.yaml
@@ -0,0 +1,31 @@
+{{- if and .Values.keda.enabled .Values.keda.modelServers.enabled }}
+{{- range $serverType, $serverConfig := .Values.keda.modelServers }}
+{{- if and (ne $serverType "enabled") $serverConfig.enabled }}
+---
+apiVersion: keda.sh/v1alpha1
+kind: ScaledObject
+metadata:
+  name: {{ include "onyx-stack.fullname" $ }}-{{ $serverType }}-model-server-scaledobject
+  namespace: {{ $.Release.Namespace }}
+  labels:
+    {{- include "onyx-stack.labels" $ | nindent 4 }}
+    app: {{ $serverType }}-model-server
+spec:
+  scaleTargetRef:
+    apiVersion: apps/v1
+    kind: Deployment
+    name: {{ include "onyx-stack.fullname" $ }}-{{ $serverType }}-model-server
+  pollingInterval: {{ $serverConfig.pollingInterval | default 30 }}
+  cooldownPeriod: {{ $serverConfig.cooldownPeriod | default 300 }}
+  minReplicaCount: {{ $serverConfig.minReplicas | default 1 }}
+  maxReplicaCount: {{ $serverConfig.maxReplicas | default 5 }}
+  triggers:
+    {{- if $serverConfig.triggers }}
+    - type: cpu
+      metadata:
+        type: Utilization
+        value: {{ $serverConfig.cpuThreshold | default "70" | quote }}
+    {{- end }}
+{{- end }}
+{{- end }}
+{{- end }}
diff --git a/deployment/helm/charts/onyx/templates/keda/slackbot-scaledobject.yaml b/deployment/helm/charts/onyx/templates/keda/slackbot-scaledobject.yaml
new file mode 100644
index 0000000000..bfd7dc0104
--- /dev/null
+++ b/deployment/helm/charts/onyx/templates/keda/slackbot-scaledobject.yaml
@@ -0,0 +1,28 @@
+{{- if and .Values.keda.enabled .Values.keda.slackbot .Values.keda.slackbot.enabled }}
+# Note: This KEDA ScaledObject will conflict with HPA if slackbot.autoscaling.enabled is also true
+# Only one autoscaler should be active per deployment
+apiVersion: keda.sh/v1alpha1
+kind: ScaledObject
+metadata:
+  name: {{ include "onyx-stack.fullname" . }}-slackbot-scaledobject
+  namespace: {{ .Release.Namespace }}
+  labels:
+    {{- include "onyx-stack.labels" . | nindent 4 }}
+    app: slackbot
+spec:
+  scaleTargetRef:
+    apiVersion: apps/v1
+    kind: Deployment
+    name: {{ include "onyx-stack.fullname" . }}-slackbot
+  pollingInterval: {{ .Values.keda.slackbot.pollingInterval | default 30 }}
+  cooldownPeriod: {{ .Values.keda.slackbot.cooldownPeriod | default 300 }}
+  minReplicaCount: {{ .Values.keda.slackbot.minReplicas | default 1 }}
+  maxReplicaCount: {{ .Values.keda.slackbot.maxReplicas | default 3 }}
+  triggers:
+    {{- if .Values.keda.slackbot.triggers }}
+    - type: cpu
+      metadata:
+        type: Utilization
+        value: {{ .Values.keda.slackbot.cpuThreshold | default "70" | quote }}
+    {{- end }}
+{{- end }}
diff --git a/deployment/helm/charts/onyx/templates/keda/web-server-scaledobject.yaml b/deployment/helm/charts/onyx/templates/keda/web-server-scaledobject.yaml
new file mode 100644
index 0000000000..317ea9821f
--- /dev/null
+++ b/deployment/helm/charts/onyx/templates/keda/web-server-scaledobject.yaml
@@ -0,0 +1,24 @@
+{{- if and .Values.keda.enabled .Values.keda.webServer .Values.keda.webServer.enabled }}
+apiVersion: keda.sh/v1alpha1
+kind: ScaledObject
+metadata:
+  name: {{ include "onyx-stack.fullname" . }}-web-server-scaledobject
+  namespace: {{ .Release.Namespace }}
+  labels:
+    {{- include "onyx-stack.labels" . | nindent 4 }}
+    app: web-server
+spec:
+  scaleTargetRef:
+    apiVersion: apps/v1
+    kind: Deployment
+    name: {{ include "onyx-stack.fullname" . }}-web-server
+  pollingInterval: {{ .Values.keda.webServer.pollingInterval | default 30 }}
+  cooldownPeriod: {{ .Values.keda.webServer.cooldownPeriod | default 300 }}
+  minReplicaCount: {{ .Values.keda.webServer.minReplicas | default 1 }}
+  maxReplicaCount: {{ .Values.keda.webServer.maxReplicas | default 5 }}
+  triggers:
+    - type: cpu
+      metadata:
+        type: Utilization
+        value: {{ .Values.keda.webServer.cpuThreshold | default "70" | quote }}
+{{- end }}

From 78aee1946189cb2ad86ffd2e389f78a79c4c4ce4 Mon Sep 17 00:00:00 2001
From: justin-tahara <justintahara@gmail.com>
Date: Mon, 1 Sep 2025 11:31:19 -0700
Subject: [PATCH 2/2] commit changes

---
 .../keda/api-server-scaledobject.yaml         | 10 ++-
 .../celery-worker-common-scaledobject.yaml    |  4 +-
 ...elery-worker-docfetching-scaledobject.yaml |  4 +-
 ...ery-worker-docprocessing-scaledobject.yaml |  4 +-
 .../keda/model-server-scaledobject.yaml       |  4 +-
 .../templates/keda/slackbot-scaledobject.yaml | 12 ++-
 .../keda/web-server-scaledobject.yaml         | 10 ++-
 deployment/helm/charts/onyx/values.yaml       | 84 +++++++++++++++++++
 8 files changed, 115 insertions(+), 17 deletions(-)

diff --git a/deployment/helm/charts/onyx/templates/keda/api-server-scaledobject.yaml b/deployment/helm/charts/onyx/templates/keda/api-server-scaledobject.yaml
index 58951173c7..0df982997e 100644
--- a/deployment/helm/charts/onyx/templates/keda/api-server-scaledobject.yaml
+++ b/deployment/helm/charts/onyx/templates/keda/api-server-scaledobject.yaml
@@ -11,14 +11,22 @@ spec:
   scaleTargetRef:
     apiVersion: apps/v1
     kind: Deployment
-    name: {{ include "onyx-stack.fullname" . }}-api-server
+    name: {{ include "onyx-stack.fullname" . }}
   pollingInterval: {{ .Values.keda.apiServer.pollingInterval | default 30 }}
   cooldownPeriod: {{ .Values.keda.apiServer.cooldownPeriod | default 300 }}
   minReplicaCount: {{ .Values.keda.apiServer.minReplicas | default 1 }}
   maxReplicaCount: {{ .Values.keda.apiServer.maxReplicas | default 10 }}
+  # Use HPA mode to generate an HPA that works alongside existing HPA infrastructure
+  hpaName: {{ include "onyx-stack.fullname" . }}-api-server-keda-hpa
   triggers:
     - type: cpu
       metadata:
         type: Utilization
         value: {{ .Values.keda.apiServer.cpuThreshold | default "70" | quote }}
+    {{- if .Values.keda.apiServer.memoryThreshold }}
+    - type: memory
+      metadata:
+        type: Utilization
+        value: {{ .Values.keda.apiServer.memoryThreshold | quote }}
+    {{- end }}
 {{- end }}
diff --git a/deployment/helm/charts/onyx/templates/keda/celery-worker-common-scaledobject.yaml b/deployment/helm/charts/onyx/templates/keda/celery-worker-common-scaledobject.yaml
index a38cc46b42..d85b6a3bf4 100644
--- a/deployment/helm/charts/onyx/templates/keda/celery-worker-common-scaledobject.yaml
+++ b/deployment/helm/charts/onyx/templates/keda/celery-worker-common-scaledobject.yaml
@@ -20,8 +20,7 @@ spec:
   minReplicaCount: {{ $workerConfig.minReplicas | default 1 }}
   maxReplicaCount: {{ $workerConfig.maxReplicas | default 10 }}
   triggers:
-    {{- if $workerConfig.triggers }}
-    # Default Prometheus-based trigger for Redis queue depth if none specified
+    # Default Prometheus-based trigger for Redis queue depth
     # Scaling Logic:
     # - When queue depth > 5: Scale up by factor of 2 (moderate scaling)
     # - When queue depth <= 5: Scale down by factor of 0.5 (conservative scaling)
@@ -45,7 +44,6 @@ spec:
             (sum(redis_key_size{key=~"connector_{{ $workerType }}.*"}) <= 5)
               * 0.5
           )
-    {{- end }}
 {{- end }}
 {{- end }}
 {{- end }}
diff --git a/deployment/helm/charts/onyx/templates/keda/celery-worker-docfetching-scaledobject.yaml b/deployment/helm/charts/onyx/templates/keda/celery-worker-docfetching-scaledobject.yaml
index f8d52a2c59..ce8c321840 100644
--- a/deployment/helm/charts/onyx/templates/keda/celery-worker-docfetching-scaledobject.yaml
+++ b/deployment/helm/charts/onyx/templates/keda/celery-worker-docfetching-scaledobject.yaml
@@ -17,8 +17,7 @@ spec:
   minReplicaCount: {{ .Values.keda.celeryWorkers.docfetching.minReplicas | default 1 }}
   maxReplicaCount: {{ .Values.keda.celeryWorkers.docfetching.maxReplicas | default 10 }}
   triggers:
-    {{- if .Values.keda.celeryWorkers.docfetching.triggers }}
-    # Default Prometheus-based trigger for Redis queue depth if none specified
+    # Default Prometheus-based trigger for Redis queue depth
     # Scaling Logic: 
     # - When queue depth > 5: Scale up by factor of 2 (aggressive scaling)
     # - When queue depth <= 5: Scale down by factor of 0.5 (conservative scaling)
@@ -42,5 +41,4 @@ spec:
             (sum(redis_key_size{key=~"connector_docfetching.*"}) <= 5)
               * 0.5
           )
-    {{- end }}
 {{- end }}
diff --git a/deployment/helm/charts/onyx/templates/keda/celery-worker-docprocessing-scaledobject.yaml b/deployment/helm/charts/onyx/templates/keda/celery-worker-docprocessing-scaledobject.yaml
index 2bd2e1a4b8..57555d049f 100644
--- a/deployment/helm/charts/onyx/templates/keda/celery-worker-docprocessing-scaledobject.yaml
+++ b/deployment/helm/charts/onyx/templates/keda/celery-worker-docprocessing-scaledobject.yaml
@@ -17,8 +17,7 @@ spec:
   minReplicaCount: {{ .Values.keda.celeryWorkers.docprocessing.minReplicas | default 1 }}
   maxReplicaCount: {{ .Values.keda.celeryWorkers.docprocessing.maxReplicas | default 50 }}
   triggers:
-    {{- if .Values.keda.celeryWorkers.docprocessing.triggers }}
-    # Default Prometheus-based trigger for Redis queue depth if none specified
+    # Default Prometheus-based trigger for Redis queue depth
     # Scaling Logic:
     # - When queue depth > 20: Scale up by factor of 4 (very aggressive scaling)
     # - When queue depth <= 20: Scale down by factor of 0.25 (very conservative scaling)
@@ -42,5 +41,4 @@ spec:
             (sum(redis_key_size{key=~"connector_docprocessing.*"}) <= 20)
               * 0.25
           )
-    {{- end }}
 {{- end }}
diff --git a/deployment/helm/charts/onyx/templates/keda/model-server-scaledobject.yaml b/deployment/helm/charts/onyx/templates/keda/model-server-scaledobject.yaml
index f54ab3086c..f0599a1dd7 100644
--- a/deployment/helm/charts/onyx/templates/keda/model-server-scaledobject.yaml
+++ b/deployment/helm/charts/onyx/templates/keda/model-server-scaledobject.yaml
@@ -14,18 +14,16 @@ spec:
   scaleTargetRef:
     apiVersion: apps/v1
     kind: Deployment
-    name: {{ include "onyx-stack.fullname" $ }}-{{ $serverType }}-model-server
+    name: {{ include "onyx-stack.fullname" $ }}-{{ $serverType }}-model
   pollingInterval: {{ $serverConfig.pollingInterval | default 30 }}
   cooldownPeriod: {{ $serverConfig.cooldownPeriod | default 300 }}
   minReplicaCount: {{ $serverConfig.minReplicas | default 1 }}
   maxReplicaCount: {{ $serverConfig.maxReplicas | default 5 }}
   triggers:
-    {{- if $serverConfig.triggers }}
     - type: cpu
       metadata:
         type: Utilization
         value: {{ $serverConfig.cpuThreshold | default "70" | quote }}
-    {{- end }}
 {{- end }}
 {{- end }}
 {{- end }}
diff --git a/deployment/helm/charts/onyx/templates/keda/slackbot-scaledobject.yaml b/deployment/helm/charts/onyx/templates/keda/slackbot-scaledobject.yaml
index bfd7dc0104..5b1f181ef8 100644
--- a/deployment/helm/charts/onyx/templates/keda/slackbot-scaledobject.yaml
+++ b/deployment/helm/charts/onyx/templates/keda/slackbot-scaledobject.yaml
@@ -1,6 +1,6 @@
 {{- if and .Values.keda.enabled .Values.keda.slackbot .Values.keda.slackbot.enabled }}
-# Note: This KEDA ScaledObject will conflict with HPA if slackbot.autoscaling.enabled is also true
-# Only one autoscaler should be active per deployment
+# Note: This KEDA ScaledObject works alongside existing HPA using KEDA's HPA mode
+# KEDA generates an HPA that can coexist with traditional HPA infrastructure
 apiVersion: keda.sh/v1alpha1
 kind: ScaledObject
 metadata:
@@ -18,11 +18,17 @@ spec:
   cooldownPeriod: {{ .Values.keda.slackbot.cooldownPeriod | default 300 }}
   minReplicaCount: {{ .Values.keda.slackbot.minReplicas | default 1 }}
   maxReplicaCount: {{ .Values.keda.slackbot.maxReplicas | default 3 }}
+  # Use HPA mode to generate an HPA that works alongside existing HPA infrastructure
+  hpaName: {{ include "onyx-stack.fullname" . }}-slackbot-keda-hpa
   triggers:
-    {{- if .Values.keda.slackbot.triggers }}
     - type: cpu
       metadata:
         type: Utilization
         value: {{ .Values.keda.slackbot.cpuThreshold | default "70" | quote }}
+    {{- if .Values.keda.slackbot.memoryThreshold }}
+    - type: memory
+      metadata:
+        type: Utilization
+        value: {{ .Values.keda.slackbot.memoryThreshold | quote }}
     {{- end }}
 {{- end }}
diff --git a/deployment/helm/charts/onyx/templates/keda/web-server-scaledobject.yaml b/deployment/helm/charts/onyx/templates/keda/web-server-scaledobject.yaml
index 317ea9821f..2e54f3a130 100644
--- a/deployment/helm/charts/onyx/templates/keda/web-server-scaledobject.yaml
+++ b/deployment/helm/charts/onyx/templates/keda/web-server-scaledobject.yaml
@@ -11,14 +11,22 @@ spec:
   scaleTargetRef:
     apiVersion: apps/v1
     kind: Deployment
-    name: {{ include "onyx-stack.fullname" . }}-web-server
+    name: {{ include "onyx-stack.fullname" . }}
   pollingInterval: {{ .Values.keda.webServer.pollingInterval | default 30 }}
   cooldownPeriod: {{ .Values.keda.webServer.cooldownPeriod | default 300 }}
   minReplicaCount: {{ .Values.keda.webServer.minReplicas | default 1 }}
   maxReplicaCount: {{ .Values.keda.webServer.maxReplicas | default 5 }}
+  # Use HPA mode to generate an HPA that works alongside existing HPA infrastructure
+  hpaName: {{ include "onyx-stack.fullname" . }}-web-server-keda-hpa
   triggers:
     - type: cpu
       metadata:
         type: Utilization
         value: {{ .Values.keda.webServer.cpuThreshold | default "70" | quote }}
+    {{- if .Values.keda.webServer.memoryThreshold }}
+    - type: memory
+      metadata:
+        type: Utilization
+        value: {{ .Values.keda.webServer.memoryThreshold | quote }}
+    {{- end }}
 {{- end }}
diff --git a/deployment/helm/charts/onyx/values.yaml b/deployment/helm/charts/onyx/values.yaml
index 04cee79642..b7b3ceede4 100644
--- a/deployment/helm/charts/onyx/values.yaml
+++ b/deployment/helm/charts/onyx/values.yaml
@@ -8,6 +8,90 @@ global:
   # Global pull policy for all Onyx component images
   pullPolicy: "IfNotPresent"
 
+keda:
+  # Master switch for all KEDA functionality - disabled by default
+  # KEDA works alongside existing HPA infrastructure using HPA mode
+  # This provides advanced scaling triggers while maintaining HPA stability and coexistence
+  enabled: false
+  
+  # API Server autoscaling configuration
+  # KEDA generates an HPA named: {release}-api-server-keda-hpa
+  # Can coexist with api.autoscaling.* configuration
+  apiServer:
+    enabled: false
+    pollingInterval: 30
+    cooldownPeriod: 300
+    minReplicas: 1
+    maxReplicas: 10
+    cpuThreshold: "70"
+    memoryThreshold: "80"  # Optional: enable memory-based scaling
+  
+  # Web Server autoscaling configuration
+  # KEDA generates an HPA named: {release}-web-server-keda-hpa
+  # Can coexist with webserver.autoscaling.* configuration
+  webServer:
+    enabled: false
+    pollingInterval: 30
+    cooldownPeriod: 300
+    minReplicas: 1
+    maxReplicas: 5
+    cpuThreshold: "70"
+    memoryThreshold: "80"  # Optional: enable memory-based scaling
+  
+  # Slackbot autoscaling configuration
+  # KEDA generates an HPA named: {release}-slackbot-keda-hpa
+  # Can coexist with slackbot.autoscaling.* configuration
+  slackbot:
+    enabled: false
+    pollingInterval: 30
+    cooldownPeriod: 300
+    minReplicas: 1
+    maxReplicas: 3
+    cpuThreshold: "70"
+    memoryThreshold: "80"  # Optional: enable memory-based scaling
+  
+  # Model Servers autoscaling configuration
+  # KEDA generates HPAs for each enabled model server
+  # Can coexist with existing model server autoscaling
+  modelServers:
+    enabled: false
+    # Individual model server configurations can be added here
+    # Example:
+    # inference:
+    #   enabled: false
+    #   pollingInterval: 30
+    #   cooldownPeriod: 300
+    #   minReplicas: 1
+    #   maxReplicas: 5
+    #   cpuThreshold: "70"
+    #   memoryThreshold: "80"
+  
+  # Celery Workers autoscaling configuration
+  # KEDA provides Redis queue-based scaling for high-performance worker management
+  # Can coexist with existing celery worker autoscaling
+  celeryWorkers:
+    enabled: false
+    # Individual worker configurations can be added here
+    # Example:
+    # docprocessing:
+    #   enabled: false
+    #   pollingInterval: 30
+    #   cooldownPeriod: 300
+    #   minReplicas: 1
+    #   maxReplicas: 50
+    # light:
+    #   enabled: false
+    #   pollingInterval: 30
+    #   cooldownPeriod: 300
+    #   minReplicas: 1
+    #   maxReplicas: 10
+    # primary:
+    #   enabled: false
+    #   pollingInterval: 30
+    #   cooldownPeriod: 300
+    #   minReplicas: 1
+    #   maxReplicas: 10
+
 postgresql:
   primary:
     persistence: