> ## Documentation Index
> Fetch the complete documentation index at: https://docs-classic.cognigy.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Install the Extension Runtime

> This guide describes how to install and activate the Extension runtime for on-premises Cognigy installations.

The Extension runtime, previously called Cognigy Serverless, is the execution environment for Extensions. It replaces the legacy `service-execution` service and runs Extensions in isolated, on-demand environments. For more information, refer to [Extension runtime](https://docs-classic.cognigy.com/ai/for-developers/extensions.md#extension-runtime).

> [!WARNING]
>
> Starting with Cognigy 2026.21.0, the Extension runtime is required. The legacy runtime, including `service-execution` and Trusted Extensions, is removed. Install the Extension runtime before upgrading to 2026.21.0. Otherwise, all Extensions stop working after the upgrade.

For this installation guide, consider the namespace as `cognigy-ai` and the Helm chart file name as `cognigy-ai-values.yaml`. Adjust the following instructions to match your namespace and Helm chart file name.

## Prerequisites

* [Migrate Extensions tenant storage](https://docs-classic.cognigy.com/ai/administer/installation/migration/extensions-tenant-storage-migration).
* Cognigy 2026.14.2, 2026.15.1 or later.

## Configure Rollout for the Extension Runtime

Add the following rollout configuration to `cognigy-ai-values.yaml`. This configuration defines the initial rollout policy with the Extension runtime disabled for all organizations, Projects, and Extensions:

```yaml theme={null}
commonConfigs:
  cognigyRollout:
    serverless.yaml: |
      organisations:
        enabled: []
        disabled: []
      projects:
        enabled: []
        disabled: []
      extensions:
        enabled: []
        disabled: []
```

## Install the Extension Runtime in the Cluster

Add the following Extension runtime configuration to your `cognigy-ai-values.yaml`:

```yaml theme={null}
serverless:
  enabled: true

  router:
    # Optional: enable HPA-based autoscaler for router pods
    autoscaling:
      enabled: true

  # Optional: deploy executor pods to a dedicated node pool
  global:
    nodeSelector: "node-role=serverless-executor"
    tolerations: "node-role.cognigy.ai=serverless-executor"
```

## Apply Configuration

Deploy the updated Cognigy Helm chart:

```bash theme={null}
helm registry login cognigy.azurecr.io \
    --username <your-username> \
    --password <your-password>

helm upgrade --install --namespace cognigy-ai cognigy-ai oci://cognigy.azurecr.io/helm/cognigy.ai --version <version> --values cognigy-ai-values.yaml
```

## Verify Serverless Router Pods

Make sure all `serverless-router` pods are running and healthy. Execute the following command to check the pods' status:

```bash theme={null}
kubectl get pods -n cognigy-ai --watch | grep serverless-router
```

All `serverless-router` pods must display the `Running` status and all containers must be ready, for example, displaying `1/1`.

## Activate the Extension Runtime

1. Once all `serverless-router` pods are running and healthy, update the rollout configuration in `cognigy-ai-values.yaml` to activate the Extension runtime for all organizations, Projects, and Extensions:

   ```yaml theme={null}
   commonConfigs:
     cognigyRollout:
       serverless.yaml: |
         organisations:
           enabled: ["*"]
           disabled: []
         projects:
           enabled: ["*"]
           disabled: []
         extensions:
           enabled: ["*"]
           disabled: []
   ```

2. Deploy the updated Cognigy Helm chart:

   ```bash theme={null}
   helm registry login cognigy.azurecr.io \
       --username <your-username> \
       --password <your-password>

   helm upgrade --install --namespace cognigy-ai cognigy-ai oci://cognigy.azurecr.io/helm/cognigy.ai --version <version> --values cognigy-ai-values.yaml
   ```


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.