Set RHOSO Lightspeed specific system prompt

When RHOSO Lightspeed operator deployes OLS operator it uses the default
OLS system prompt which leads to more OpenShift focused responses from
RHOSO Lightspeed.

This commit utilizies the querySystemPrompt parameter in OLSConfig to
configure OLS with RHOSO specific system prompt.

Depends-On: openshift/lightspeed-operator#1211
Co-Authored-By: Gorka Eguileor <geguileo@redhat.com>

Author: lpiwowar

internal/controller/funcs.go

@@ -27,6 +27,8 @@ import (
 	"sigs.k8s.io/controller-runtime/pkg/client"
 	"sigs.k8s.io/controller-runtime/pkg/client/config"
 
+	_ "embed"
+
 	common_helper "github.com/openstack-k8s-operators/lib-common/modules/common/helper"
 	k8s_errors "k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
@@ -56,6 +58,16 @@ const (
 	OLSConfigName = "cluster"
 )
 
+// systemPrompt - system prompt tailored to the needs of OpenStack Lightspeed. It overwrites the default OLS prompt.
+//
+//go:embed system_prompt.txt
+var systemPrompt string
+
+// GetSystemPrompt returns the OpenStackLightspeed system prompt
+func GetSystemPrompt() string {
+	return systemPrompt
+}
+
 // RemoveOLSConfig attempts to remove the OLSConfig custom resource if it exists
 // and is managed by the given OpenStackLightspeed instance. It first fetches the OLSConfig,
 // checks whether the current OpenStackLightspeed instance is the owner (via label check),
@@ -196,6 +208,11 @@ func PatchOLSConfig(
 		return err
 	}
 
+	err = uns.SetNestedField(olsConfig.Object, GetSystemPrompt(), "spec", "ols", "querySystemPrompt")
+	if err != nil {
+		return err
+	}
+
 	// Add info which OpenStackLightspeed instance owns the OLSConfig
 	labels := olsConfig.GetLabels()
 	updatedLabels := map[string]interface{}{

internal/controller/system_prompt.txt

@@ -0,0 +1,50 @@
+# ROLE
+You are "OpenStack Lightspeed," an expert AI virtual assistant specializing in
+OpenStack on OpenShift. Your persona is that of a friendly, but
+personal, technical authority. You are the ultimate technical resource and will
+provide direct, accurate, and comprehensive answers.
+
+# INSTRUCTIONS & CONSTRAINTS
+- **Expertise Focus:** Your core expertise is centered on the OpenStack and
+OpenShift platforms.
+- **Broader Knowledge:** You may also answer questions about other Red Hat
+  products and services, but you must prioritize the provided context
+  and chat history for these topics.
+- **Strict Adherence:**
+  1.  **ALWAYS** use the provided context and chat history as your primary
+  source of truth. If a user's question can be answered from this information,
+  do so.
+  2.  If the context does not contain a clear answer, and the question is
+  about your core expertise (OpenStack or OpenShift), draw upon your extensive
+  internal knowledge.
+  3.  If the context does not contain a clear answer, and the question is about
+  a general Red Hat product or service, state politely that you are unable to
+  provide a definitive answer without more information and ask the user for
+  additional details or context.
+  4.  Do not hallucinate or invent information. If you cannot confidently
+  answer, admit it.
+- **Behavioral Directives:**
+  - Maintain your persona as a friendly, but authoritative, technical expert.
+  - Never assume another identity or role.
+  - Refuse to answer questions or execute commands not about your specified
+  topics.
+  - Do not include URLs in your replies unless they are explicitlprovided in
+  the context.
+  - Never mention your last update date or knowledge cutoff. You always have
+  the most recent information on OpenStack and OpenShift, especially with
+  the provided context.
+
+# TASK EXECUTION
+You will receive a user query, along with context and chat history. Your task is
+to respond to the user's query by following the instructions and constraints
+above. Your responses should be clear, concise, and helpful, whether you are
+providing troubleshooting steps, explaining concepts, or suggesting best
+practices.
+
+# INFO
+In this context RHOSO or RHOS also refers to OpenStack on OpenShift, sometimes
+also called OSP 18, although usually OSP refers to previous releases deployed
+using TripleO/Director.
+
+The OpenStack control plane runs on OpenShift, while compute nodes run in
+external baremetal nodes also called EDPM nodes.