opscli copilot Command Usage

The opscli copilot command assists in managing your operations with LLM (Large Language Model) integration. Below are the details for the available flags, environment variables, and example commands.

Flags

Usage:
  opscli copilot [flags]

Flags:
  -e, --endpoint string   e.g. https://api.openai.com/v1
  -h, --help              help for copilot
      --history int        (default 5)
  -k, --key string        e.g. sk-xxx
  -m, --model string      e.g. gpt-3.5-turbo
  -s, --silence           Suppress output
  -v, --verbose string    Verbose mode

Environment Variables

copilot will automatically fetch the following environment variables if set. If not, it will use default values:

OPENAI_API_KEY (API key for OpenAI)
OPENAI_API_HOST (Host URL for OpenAI API)
OPENAI_API_BASE (Base URL for OpenAI API)
OPENAI_API_MODEL (Model type for OpenAI API)
OPS_SERVER (Server URL for ops server)
OPS_TOKEN (Token for authentication)

You can set these environment variables as follows:

export OPENAI_API_KEY=sk-xxxx
export OPENAI_API_HOST=https://llmapi.YOUR-OPENAI-SERVER.com/v1
export OPS_SERVER=http://1.1.1.1
export OPS_TOKEN=xxxx

Running Copilot

To run Copilot:

/usr/local/bin/opscli copilot

You will be prompted with:

Welcome to Opscli Copilot. Please type "exit" or "q" to quit.
Opscli>

Available Operations

To check available operations:

Opscli> What operations are available?
Here are the available operations and their descriptions:

1. list-cluster: Query the list of Kubernetes clusters.
2. list-task: Query the list of tasks.
3. list-pipeline: Query the list of pipelines.
4. restart-pod: Restart or delete a Pod. Variables: podname (one or more Pod names).
5. force-restart-pod: Force restart or delete a Pod. Variables: podname (one or more Pod names).
6. get-cluster-ip: Query the IP addresses of clusters. Variables: clusterip (one or more cluster IP addresses).
7. clear-disk: Clear disk space. Variables: nodeName (one or more node names).

Querying Clusters

To list clusters:

Opscli> What clusters are available?
The following clusters are available:
1. ops-system/xx-xx: A cluster deployed on xxx cloud for 88 inference cluster.
2. ops-system/xx-xx: A cluster deployed on xxx for the 119 cluster.
3. ops-system/xx-xx: A cluster deployed on xxx for integrated training and inference.
4. ops-system/xx-xx: A cluster deployed on xxx for NPU training.

Restarting a Pod

To force restart a pod:

Opscli> Force restart the pod ubuntu-8474647969-qszcj in the training-inference cluster

The process will look like this:

Check if the pod exists:
- Output: "Pod ubuntu-8474647969-qszcj found in default namespace."
Delete the pod:
- Output: "Warning: Force delete, no confirmation waiting for the pod to terminate. The resource may run indefinitely on the cluster."
- Output: "Pod 'ubuntu-8474647969-qszcj' has been force deleted."

In your cluster, you can watch the pod status with:

kubectl get pod ubuntu-8474647969-qszcj -w

The output will show the pod terminating and restarting:

NAME                      READY   STATUS    RESTARTS   AGE
ubuntu-8474647969-qszcj   1/1     Running   0          20h
ubuntu-8474647969-qszcj   1/1     Terminating   0          20h
ubuntu-8474647969-qszcj   1/1     Terminating   0          20h

Copilot