fix: improve default_prompt

[camigira]

📑 Description

(closes: #417)

Attempt to improve the default_prompt aiming to:

• Prevent Prompt Injection by specifying variable text win triple dashes -.
• Keep output concise but insightful.
• Clearer output Error vs Solution.

Original Prompt

New Prompt

Note: After a few tests, the text Provide the most possible solution seemed to be better than Provide the most likely solution as it explored multiple options as opposed to focusing the answer on a single option.

✅ Checks

• My pull request adheres to the code style of this project
• My code requires changes to the documentation
• I have updated the documentation as required
• All the tests have passed

ℹ Additional Information

[matthisholleville]

I wonder how the prompt behaves in case there are several errors on an object? Could you please provide an example with this case?

[camigira]

Example for a broken pod and deployment with multiple issues (hope this is what you had in mind with your question)

broken-pod.yml

apiVersion: v1
kind: Pod
metadata:
  name: broken-pod
spec:
  containers:
    - name: broken-pod
      image: nginx:1.a.b.c
      livenessProbe:
        httpGet:
          path: /
          port: 90
        initialDelaySeconds: 3
        periodSeconds: 3

broken-deployment.yml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: broken-deployment
spec:
  replicas: 3
  selector:
    matchLabels:
      app: myapp
  template:
    metadata:
      labels:
        app: myapp
    spec:
      containers:
        - name: nginx-container
          image: nginx:latest
          ports:
            - containerPort: 80
          livenessProbe:
            httpGet:
              path: /healthz #doesn't exist in nginx
              port: 80
            initialDelaySeconds: 30
            periodSeconds: 10
          readinessProbe:
            httpGet:
              path: /readiness #doesn't exist in nginx
              port: 80
            initialDelaySeconds: 20
            periodSeconds: 5

[matthisholleville]

You still have only one error per object.

Please try with this object:

apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: fake-ingress
spec:
  rules:
  - host: fake-ingress.example.com
    http:
      paths:
      - backend:
          service:
            name: fake-service
            port:
              number: 80
        path: /
        pathType: Prefix
  tls:
  - hosts:
    - fake-ingress.example.com
    secretName: fake-secret

You should have this result :

0 k8sgpt/fake-ingress(fake-ingress)
- Error: Ingress k8sgpt/fake-ingress does not specify an Ingress class.
- Error: Ingress uses the service k8sgpt/fake-service which does not exist.
- Error: Ingress uses the secret k8sgpt/fake-secret as a TLS certificate which does not exist.

[camigira]

Thanks @matthisholleville for providing the example, here the results:

Current Default Prompt

0 default/fake-ingress(fake-ingress)
- Error: Ingress default/fake-ingress does not specify an Ingress class.
- Error: Ingress uses the service default/fake-service which does not exist.
- Error: Ingress uses the secret default/fake-secret as a TLS certificate which does not exist.
The Ingress resource is trying to use a service and a TLS certificate that do not exist, and it also does not specify which Ingress class to use.

To fix this error, create the necessary service and secret resources, and specify the Ingress class to use in the Ingress resource.

Proposed Default Prompt

0 default/fake-ingress(fake-ingress)
- Error: Ingress default/fake-ingress does not specify an Ingress class.
- Error: Ingress uses the service default/fake-service which does not exist.
- Error: Ingress uses the secret default/fake-secret as a TLS certificate which does not exist.
Error: Ingress default/fake-ingress needs an Ingress class, service default/fake-service doesn't exist, and secret default/fake-secret is a non-existent TLS cert.

Solution: 
1. Create a service named default/fake-service.
2. Create a secret named default/fake-secret with your TLS cert.
3. Add an ingress class to your Ingress resource, like nginx or traefik. 
4. Update your Ingress resource to use the service and secret created in step 1 and 2.

[matthisholleville]

I really like the solution part @camigira but the error part in the prompt seems redondant.

What do you think @AlexsJones ?

[AlexsJones]

I really like the solution part @camigira but the error part in the prompt seems redondant.

What do you think @AlexsJones ?

I think as long as it yields the outputs that are clear and understandable to the user its good.
I don't know whether the error being specified in the block improves the prompt at all - we need to try with additional backends to be sure.

[camigira]

👋 @AlexsJones @matthisholleville

I saw this PR was approved but there are still unresolved conversations. Should I mark as completed and merge?

Thanks!