ava knowledge add -s git -r https://github.com/MyPrivateOrg/my-private-repository.git -t "ghp_dflkjcIO..."

Use the flag --enable-executors=true

Example :

go run main.go chat -m "Pod web-server-5b866987d8-4nhsg in namespace default Crashlooping." --enable-executors=true

You are free to enable or disable executors in the body of the requests.

Example :

curl -X POST https://your-url/chat \
-H "Content-Type: application/json" \
-d '{
  "backend": "openai",
  "enableExecutors": true,
  "language": "en",
  "message": "Pod web-server-5b866987d8-sxmtj in namespace default Crashlooping."
}'

For reacting to AlertManager webhooks, you simply need to enable executors in the configuration file.

Example :

# ava config
executors:
    enabled: true
    ...

Read-only

• getClusterRole: Retrieve details of a specific ClusterRole
• getCronJob: Retrieve details of a specific CronJob
• getConfigMap: Retrieve details of a specific ConfigMap
• getCrd: Retrieve details of a CustomResourceDefinition (CRD)
• getDaemonSet: Retrieve details of a specific DaemonSet
• getDeployment: Retrieve details of a specific Deployment
• getEndpointSlices: Retrieve details of a specific EndpointSlice
• getHPA: Retrieve details of a specific HorizontalPodAutoscaler
• getIngress: Retrieve details of a specific Ingress
• getLimitRange: Retrieve details of a specific LimitRange
• getJob: Retrieve details of a specific Job
• getNode: Retrieve details of a specific Node
• getPod: Retrieve details of a specific Pod
• getPdb: Retrieve details of a specific PodDisruptionBudget
• getPersistentVolume: Retrieve details of a specific PersistentVolume
• getPersistentVolumeClaim: Retrieve details of a specific PersistentVolumeClaim
• getRole: Retrieve details of a specific Role
• getRoleBinding: Retrieve details of a specific RoleBinding
• getServiceAccount: Retrieve details of a specific ServiceAccount
• getSecret: Retrieve details of a specific Secret
• getStorageClass: Retrieve details of a specific StorageClass
• getStatefulSet: Retrieve details of a specific StatefulSet
• listClusterRoles: List all ClusterRoles in the cluster
• listCrds: List all CustomResourceDefinitions (CRDs) in the cluster
• listCronJobs: List all CronJobs in a namespace
• listConfigMaps: List all ConfigMaps in a namespace
• listDaemonSets: List all DaemonSets in a namespace
• listDeployments: List all Deployments in a namespace
• listEndpointSlices: List all EndpointSlices in a namespace
• listIngresses: List all Ingresses in a namespace
• listJobs: List all Jobs in a namespace
• listLimitRanges: List all LimitRanges in a namespace
• listNamespaces: List all Namespaces in the cluster
• listServicesAccounts: List all ServiceAccounts in a namespace
• listSecrets: List all Secrets in a namespace
• listStorageClasses: List all StorageClasses in the cluster
• listStatefulSets: List all StatefulSets in a namespace
• listPods: List all Pods in a namespace
• listPersistentVolumes: List all PersistentVolumes in the cluster
• listPersistentVolumeClaims: List all PersistentVolumeClaims in a namespace
• listPdbs: List all PodDisruptionBudgets in a namespace
• listRoles: List all Roles in a namespace
• listRoleBindings: List all RoleBindings in a namespace
• podLogs: Retrieve logs of a specific Pod
• topPods: Show resource usage for Pods in a namespace

Write

By default, these executors can impact environments and are not enabled. To activate them, simply enable them in the configuration.

• deletePod: Delete a pod.
• rolloutDeployment: Perform a rollout restart for a deployment.

• wait: Waits before the next action.

• getUrl: Makes a GET request to a URL and returns status and timing.

Before starting the server mode, you need a PostgreSQL database up and running. You can use Docker or a cloud solution like CloudSQL.

1. Rename .envrc-sample to .envrc.
2. Update the environment variable export DATABASE_URL="CHANGE_ME" with a valid connection string.
3. Export the variables: direnv allow.
4. Initialize the PostgreSQL schema:

   go run github.com/steebchen/prisma-client-go db push
   

5. Generate the Swagger documentation:

   make swagger
   

6. Start the server mode:

   go run main.go serve
   

Once launched, you can access the Swagger documentation at the following URL: http://localhost:8080/swagger/index.html#.

Installation

Before connecting Slack, you must deploy Ava. To connect Ava to Slack, follow the steps below:

1. Create a new Slack App named Ava Bot. The name is crucial and must not be changed!
2. Copy the Verification Token from the Basic Information page and the Bot User OAuth Token from the OAuth & Permissions page of your Slack application, then paste them into Ava's configuration file (e.g., ava config edit or ./charts/ava/values.yaml if you uses Kubernetes):

   # ava config
   events:
       type: slack
       slack:
           validationToken: ${SLACK_VALIDATION_TOKEN}
           botToken: ${SLACK_BOT_TOKEN}
   

   • If you use Kubernetes dont forget to create the K8s secret for Slack kubectl create secret generic slack-secret --from-literal=validation-token=$(echo $SLACK_VALIDATION_TOKEN) --from-literal=bot-token=$(echo $SLACK_BOT_TOKEN)
3. Start Ava in server mode with a publicly accessible URL so Slack can send events.
4. On the Event Subscriptions page, enable events and add your URL in the Request URL section ($MY_URL/event/slack).
5. Further down on the same page, open the Subscribe to bot events section and add the permissions message:channels and message:groups.
6. On the OAuth & Permissions page, add the scopes chat:write, users.profile:read, and users:read.
7. Install your app into the desired channel(s).

You can test the setup by sending a direct message to your bot: @Ava Bot Hello how are you today?. You should see Ava react to the event in its logs and send a Slack message containing 👀.

Interacting with Ava

To start interacting with Ava, you can:

• Begin your message by mentioning it: @Ava Bot my pod example is crashlooping. Do we have any runbook to understand & fix the problem?
• Configure AlertManager to send a Slack message. Ava will respond not only when mentioned but also to messages from other bots.

This section shows how to set up a local environment to demonstrate Ava with AlertManager webhooks. It installs:

• AlertManager
• Prometheus
• Ava (server mode)
• Example webserver-chaos

1. Install Prometheus

Run these commands:

helm repo add prometheus-community https://prometheus-community.github.io/helm-charts
helm repo update
helm install prometheus prometheus-community/prometheus --namespace monitoring --values ./docs/examples/custom-values.yaml --create-namespace
helm install prometheus-operator-crds prometheus-community/prometheus-operator-crds --namespace monitoring

2. Create a secret with your OpenAI API Key

Run this command:

kubectl create secret generic ai-backend-secret --from-literal=openai-api-key=$(echo $OPENAI_API_KEY) -n monitoring

3. Deploy Ava using Helm

Run this command:

cd charts/ava/
helm install ava . -n monitoring

4. Deploy demo web-server

Run this command:

kubectl apply -f ./docs/examples/crashloop.yaml -n default

The demo webserver handles two routes:

1. /: Returns Hello from Ava :).
2. /chaos: Runs sys.exit(1).

When everything is installed, run this command:

curl http://$(kubectl get svc web-server-service -o jsonpath='{.status.loadBalancer.ingress[*].ip}' -n default)/chaos

This will trigger chaos in the webserver. Ava should detect the issue and fix it. After a few minutes, your pod should be healthy.