Prompt-based PR Automation

Use a light weight coding agent to automate routine GitOps changes

Deterministic PR automations have a win in speed and reliability, but the machinery does require setup and configuration. You can simplify this greatly with our basic AI integration, where Plural will use a fast, sandboxed coding agent to make the needed code changes for your desired PR. The process is basically:

Provide a templated prompt that will define the change needed in the repo.
On PR Automation invocation, the agent will take over and make the needed modification.

Info:

These are especially useful for unstructured code changes. For instance, preserving documentation in YAML files, or modifying terraform code without complex regex logic.

Basic Example

A simple example of one would be:

yaml

apiVersion: deployments.plural.sh/v1alpha1
kind: PrAutomation
metadata:
  name: cluster-updater
spec:
  name: cluster-updater
  documentation: |
    Sets up a PR to provision a new workspace for a team.  This is fairly rudimentary at the moment for demonstration purposes
  ai:
    enabled: true # use this to toggle ai on existing PR automations
    prompt: |
      Update the kubernetes version configuration for the {{ context.name }} cluster.  It will be found in a file under clusters/{{ context.tier }} and should be a simple one-line YAML change.
  scmConnectionRef:
    name: github
  title: 'Updating the {{ context.tier}} cluster {{ context.name }}'
  message: "Updating the {{ context.tier }} cluster {{ context.name }}"
  identifier: pluralsh/pr-automation-demo # id slug for the repo this automation will be applied to
  configuration:
    - name: name
      type: STRING
      documentation: The name of this workspace
    - name: tier
      type: ENUM
      values:
      - dev
      - prod

The automation will allow you to call it with a {"name": string, "tier": string} context block at any time, and the code change will be delegated to the AI according to the guidance given. The prompt is also templatable, in case it's useful to provide customization based on user input.