Skip to main content

​
Chat Basics: Your First Queries

Welcome to RubixKube’s conversational interface! This tutorial walks you through your first Chat queries with real screenshots and responses from the SRI Agent.
Hands-on approach: We’ll type real queries, show you actual responses, and explain what’s happening at each step. By the end, you’ll be comfortable using Chat for daily tasks.

​
What is the Chat Interface?

The Chat is your natural language gateway to infrastructure intelligence. Instead of:
  • Remembering complex kubectl commands
  • Navigating multiple dashboards
  • Searching through logs manually
You simply ask questions like you would to a colleague: 
"What pods are failing?"
"Why is memory-hog-demo crashing?"
"Show me resource usage"
The SRI Agent responds with data from YOUR cluster, not generic answers.

​
Opening Chat for the First Time

​
Step 1: Navigate to Chat

Click Chat in the left navigation or press ⌘K and type “chat”.
Chat interface - clean slate
What you see:
ElementPurpose
💬 Message BoxWhere you type your questions
🤖 SRI AgentThe AI assistant you’re talking to
📎 AttachmentUpload YAML files for analysis
🚀 Quick ActionsPre-built query categories
📝 Chat HistoryYour past conversations (left sidebar)

​
Your First Query: “What pods are failing?”

Let’s start with a common question every engineer asks.

​
Type the Query

In the message box, type: 
What pods are failing in my cluster?
First query typed in chat
Press Enter to send.

​
Watch the Agent Think

The SRI Agent immediately starts working. You’ll see its reasoning process:
Agent analyzing the query
What’s happening:
1

Thought: Identifying Failing Pods

The agent plans its approach - “How do I find failing pods?”
2

Function Call: fetch_kubernetes_graph_snapshot

The agent queries your Kubernetes cluster
3

Executed Function

The query completed - agent got the data
4

Analysis

Agent processes results and formulates a response
Complete transparency! You can see EXACTLY what the agent is doing - every function call, every thought. This builds trust and helps you understand how it works.

​
The Response

The agent responds with intelligent analysis:
Complete agent response with guidance
Example response:
“I’ve checked for pods and deployments in the default namespace and found nothing. This suggests one of the following:
  • There are no workloads in the default namespace
  • Your applications are in other namespaces
  • The cluster is new or empty
To help you, could you tell me:
  • Which namespace are your applications in?”
The agent asks clarifying questions! It doesn’t fail silently - it guides you to provide the information it needs. This is intelligent assistance.

​
Following Up: Natural Conversation

Chat remembers context. You can respond naturally: You type: "Check the rubixkube-tutorials namespace" Agent responds:
Agent investigating specific namespace
“Of course, Connect. I’ll check the rubixkube-tutorials namespace for you.”
Notice:
  • ✅ You didn’t repeat “show me failing pods”
  • ✅ Agent understood the context
  • ✅ Conversation flows naturally

​
Understanding Agent Responses

Every response includes:

What Was Found

The actual data from your cluster - pods, statuses, errors

Reasoning

Why the agent reached this conclusion

Suggestions

Next steps or follow-up questions

Context Clues

References to related incidents or documentation

​
Quick Action Categories

Before typing custom queries, explore the Quick Actions:
Quick action categories visible

​
Available Categories

  • Infrastructure
  • Security
  • Monitoring
  • Troubleshoot
  • Networking
Click to get prompts for:
  • Managing clusters and nodes
  • Viewing resource usage
  • Listing deployments and services
Example: “Show me node resource usage”
For beginners: Start with Quick Actions to see example queries, then customize them for your needs.

​
Chat Features You Should Know

​
Expandable Sections

Click on “Thought” or “Function Call” to see details:
Shows the agent’s reasoning process:
  • “Identifying Failing Pods”
  • “Evaluating the Strategy”
  • “Analyzing Pod’s Status”
Why this matters: You understand HOW the agent arrived at its answer.
Shows what APIs were called:
  • fetch_kubernetes_graph_snapshot
  • fetch_kubernetes_logs
  • analyze_kubernetes_resources
Why this matters: Complete transparency - you see exactly what data the agent accessed.
Bottom of response shows:
  • “3 tools called, took 28.53 seconds”
Why this matters: You know if the agent is slow due to cluster size or complexity.

​
Keyboard Shortcuts

ShortcutAction
⌘K (Mac) / Ctrl+K (Windows)Open search → type “chat”
EnterSend message
Shift + EnterNew line in message
↑Edit your last message
EscClose chat panel

​
Common First Questions

Here are queries perfect for getting started:

​
Basic Cluster Info

"What's in my cluster?"
"Show me all namespaces"
"How many pods are running?"
"What's my cluster health?"
Expected: Overview of resources, counts, health status

​
Pod Status Queries

"List all pods"
"Show me failing pods"
"What pods are in CrashLoopBackOff?"
"Any pods with high restart counts?"
Expected: Pod names, statuses, namespaces, restart counts

​
Resource Usage

"What's the memory usage?"
"Show me CPU usage across pods"
"Which pods use the most resources?"
"Any pods hitting resource limits?"
Expected: Current metrics, comparisons, limit violations

​
Learning Queries

"What is a pod?"
"Explain CrashLoopBackOff"
"What does ImagePullBackOff mean?"
"How do resource limits work?"
Expected: Educational explanations with examples from YOUR cluster

​
Response Feedback

After each response, you can:

👍 Good Response

Mark helpful responses - improves the agent

👎 Bad Response

Flag unhelpful responses for review

📋 Copy

Copy the response to clipboard

⚙️ More Options

Export, regenerate, or share

​
Chat History

All your conversations are saved in the left sidebar: Features:
  • 🔍 Search past conversations
  • 📝 Rename conversations
  • 🗑️ Delete conversations
  • 📤 Export to Markdown or JSON
  • ➕ Start new chat anytime
Pro Tip: Give conversations descriptive names like “OOMKilled Investigation - May 3” for easy reference later.

​
What Makes Chat Different?

Cluster-Aware

Responses use YOUR actual cluster data, not generic Kubernetes knowledge

Context Memory

Maintains conversation thread - no need to repeat yourself

Multi-Agent Backend

Coordinates with Observer, RCA, and Memory agents

RCA Integration

Can explain incidents detected by RubixKube

Learning Mode

Teaches Kubernetes concepts with cluster examples

Transparent

Shows its reasoning - you see how it thinks

​
Beta Limitations

What Chat DOES (Beta):
  • ✅ Answer questions about your cluster
  • ✅ Analyze incidents and provide RCA
  • ✅ Explain Kubernetes concepts
  • ✅ Provide troubleshooting guidance
  • ✅ Show you kubectl commands to run
What Chat CANNOT do (Yet):
  • ❌ Execute kubectl commands automatically
  • ❌ Apply configuration changes
  • ❌ Restart pods or deployments
Coming Q1 2026: Automated execution with approval gatesFor now, Chat is advisory - it tells you WHAT and HOW, you execute.

​
Tips for Effective Queries

Good:
  • “Show me OOMKilled pods in production”
  • “Why did payment-service-v2 fail?”
Why: Specific queries = specific answers
Avoid:
  • “Fix everything”
  • “What’s wrong?”
Why: Agent needs context
Good:
  • “What’s broken?”
  • “Why is this pod unhappy?”
Why: Agent understands casual phrasing
Avoid:
  • “kubectl get pods -n prod”
Why: Just ask! “Show me pods in prod”

​
What You Learned

Chat Interface

How to open Chat and navigate the UI

First Query

How to type and send your first question

Agent Reasoning

What Thoughts and Function Calls mean

Context Awareness

How Chat maintains conversation thread

Quick Actions

Using pre-built query categories

Best Practices

What makes a good vs. bad query

​
Next: Investigate Real Incidents

Now that you know the basics, let’s use Chat to troubleshoot actual failing pods:

Continue: Troubleshooting with Chat

Learn to investigate ImagePullBackOff, OOMKilled, and CrashLoop errors using real Chat conversations

​
Quick Reference

Most Common First Queries:
  1. "What's my cluster health?"
  2. "Show me failing pods"
  3. "Any recent incidents?"
  4. "What changed recently?"
  5. "Help me get started"
Response Time: Usually 5-30 seconds depending on query complexity Cost: No limits in Beta - ask as many questions as you want!

​
Need Help?

Support

Email: connect@rubixkube.ai
⌘I