> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rubixkube.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Chat Basics: Your First Queries

> Learn the Chat interface with hands-on examples. Start here for your first conversation with RubixKube.

# 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.

<Info>
  **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.
</Info>

***

## 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".

<Frame>
  <img src="https://mintcdn.com/rubixkube/032w9Cp9WbBNyuCD/images/chat/chat-query-health-check.png?fit=max&auto=format&n=032w9Cp9WbBNyuCD&q=85&s=1c45f0eeb4ecc11dfa103597c38d55f2" alt="Chat interface - clean slate" width="3456" height="1774" data-path="images/chat/chat-query-health-check.png" />
</Frame>

### What you see:

| Element           | Purpose                                |
| ----------------- | -------------------------------------- |
| **Message Box**   | Where you type your questions          |
| **SRI Agent**     | The AI assistant you're talking to     |
| **Attachment**    | Upload YAML files for analysis         |
| **Quick Actions** | Pre-built query categories             |
| **Chat History**  | Your 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?
```

<Frame>
  <img src="https://mintcdn.com/rubixkube/C26e5ASLJOMtwsHE/images/chat/chat-query-01-typed.png?fit=max&auto=format&n=C26e5ASLJOMtwsHE&q=85&s=ca3f59b42ea285a50fc8a9de9b64ea0e" alt="First query typed in chat" width="3456" height="1774" data-path="images/chat/chat-query-01-typed.png" />
</Frame>

Press **Enter**  to send.

***

### Watch the Agent Think

The SRI Agent immediately starts working. You'll see its reasoning process:

<Frame>
  <img src="https://mintcdn.com/rubixkube/C26e5ASLJOMtwsHE/images/chat/chat-response-01-failing-pods.png?fit=max&auto=format&n=C26e5ASLJOMtwsHE&q=85&s=937766bed991649a7de3da852cd6821f" alt="Agent analyzing the query" width="3456" height="1774" data-path="images/chat/chat-response-01-failing-pods.png" />
</Frame>

### What's happening:

<Steps>
  <Step title="Thought: Identifying Failing Pods">
    The agent plans its approach - "How do I find failing pods?"
  </Step>

  <Step title="Function Call: fetch_kubernetes_graph_snapshot">
    The agent queries your Kubernetes cluster
  </Step>

  <Step title="Executed Function">
    The query completed - agent got the data
  </Step>

  <Step title="Analysis">
    Agent processes results and formulates a response
  </Step>
</Steps>

<Note>
  **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.
</Note>

***

### The Response

The agent responds with intelligent analysis:

<Frame>
  <img src="https://mintcdn.com/rubixkube/C26e5ASLJOMtwsHE/images/chat/chat-response-02-complete.png?fit=max&auto=format&n=C26e5ASLJOMtwsHE&q=85&s=119cc65f76138ceaf8badf72c1e33f9e" alt="Complete agent response with guidance" width="3456" height="1774" data-path="images/chat/chat-response-02-complete.png" />
</Frame>

### 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?"

<Tip>
  **The agent asks clarifying questions!** It doesn't fail silently - it guides you to provide the information it needs. This is intelligent assistance.
</Tip>

***

## Following Up: Natural Conversation

Chat remembers context. You can respond naturally:

**You type:** `"Check the rubixkube-tutorials namespace"`

### Agent responds:

<Frame>
  <img src="https://mintcdn.com/rubixkube/C26e5ASLJOMtwsHE/images/chat/chat-response-03-namespace-query.png?fit=max&auto=format&n=C26e5ASLJOMtwsHE&q=85&s=cf613a1efdac0be8721e8ad11351c663" alt="Agent investigating specific namespace" width="3456" height="1774" data-path="images/chat/chat-response-03-namespace-query.png" />
</Frame>

> "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:

<CardGroup cols={2}>
  <Card title="What Was Found" icon="magnifying-glass">
    The actual data from your cluster - pods, statuses, errors
  </Card>

  <Card title="Reasoning" icon="brain">
    Why the agent reached this conclusion
  </Card>

  <Card title="Suggestions" icon="lightbulb">
    Next steps or follow-up questions
  </Card>

  <Card title="Context Clues" icon="link">
    References to related incidents or documentation
  </Card>
</CardGroup>

***

## Quick Action Categories

Before typing custom queries, explore the **Quick Actions** :

<Frame>
  <img src="https://mintcdn.com/rubixkube/032w9Cp9WbBNyuCD/images/chat/chat-query-health-check.png?fit=max&auto=format&n=032w9Cp9WbBNyuCD&q=85&s=1c45f0eeb4ecc11dfa103597c38d55f2" alt="Quick action categories visible" width="3456" height="1774" data-path="images/chat/chat-query-health-check.png" />
</Frame>

### Available Categories

<Tabs>
  <Tab title="Infrastructure">
    **Click to get prompts for:** - Managing clusters and nodes

    * Viewing resource usage
    * Listing deployments and services

    **Example:** "Show me node resource usage"
  </Tab>

  <Tab title="Security">
    **Click to get prompts for:** - Security analysis

    * Compliance checks
    * RBAC audits

    **Example:** "Are there any pods running as root?"
  </Tab>

  <Tab title="Monitoring">
    **Click to get prompts for:** - Performance metrics

    * Resource trends
    * Alert status

    **Example:** "What's the CPU usage across all pods?"
  </Tab>

  <Tab title="Troubleshoot">
    **Click to get prompts for:** - Debugging issues

    * Log analysis
    * Error investigation

    **Example:** "Why is my checkout service slow?"
  </Tab>

  <Tab title="Networking">
    **Click to get prompts for:** - Network policies

    * Connectivity issues
    * Service discovery

    **Example:** "What services are exposed?"
  </Tab>
</Tabs>

<Tip>
  **For beginners:** Start with Quick Actions to see example queries, then customize them for your needs.
</Tip>

***

## Chat Features You Should Know

### Expandable Sections

Click on **"Thought"**  or **"Thought"**  to see details:

<AccordionGroup>
  <Accordion title=" Thought - Agent Planning">
    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.
  </Accordion>

  <Accordion title=" Function Call - What Was Queried">
    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.
  </Accordion>

  <Accordion title="⏱ Timing - How Long It Took">
    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.
  </Accordion>
</AccordionGroup>

***

## Keyboard Shortcuts

| Shortcut                        | Action                    |
| ------------------------------- | ------------------------- |
| `⌘K` (Mac) / `Ctrl+K` (Windows) | Open search → type "chat" |
| `Enter`                         | Send message              |
| `Shift + Enter`                 | New line in message       |
| `↑`                             | Edit your last message    |
| `Esc`                           | Close 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:

<CardGroup cols={2}>
  <Card title=" Good Response" icon="thumbs-up">
    Mark helpful responses - improves the agent
  </Card>

  <Card title=" Bad Response" icon="thumbs-down">
    Flag unhelpful responses for review
  </Card>

  <Card title=" Copy" icon="copy">
    Copy the response to clipboard
  </Card>

  <Card title=" More Options" icon="ellipsis">
    Export, regenerate, or share
  </Card>
</CardGroup>

***

## 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

<Tip>
  <strong>Pro Tip:</strong> Give conversations descriptive names like "OOMKilled Investigation - May 3" for easy reference later.
</Tip>

***

## What Makes Chat Different?

<CardGroup cols={2}>
  <Card title="Cluster-Aware" icon="server">
    Responses use YOUR actual cluster data, not generic Kubernetes knowledge
  </Card>

  <Card title="Context Memory" icon="brain">
    Maintains conversation thread - no need to repeat yourself
  </Card>

  <Card title="Multi-Agent Backend" icon="robot">
    Coordinates with Observer, RCA, and Memory agents
  </Card>

  <Card title="RCA Integration" icon="magnifying-glass-chart">
    Can explain incidents detected by RubixKube
  </Card>

  <Card title="Learning Mode" icon="graduation-cap">
    Teaches Kubernetes concepts with cluster examples
  </Card>

  <Card title="Transparent" icon="eye">
    Shows its reasoning - you see how it thinks
  </Card>
</CardGroup>

***

## Scope of what Chat does

<Info>
  **What Chat does today:**

  * Answers questions about any connected environment.
  * Investigates incidents and generates evidence-linked RCA reports.
  * Explains platform concepts using your own resources as examples.
  * Provides step-by-step troubleshooting guidance.
  * Drafts the exact commands or configuration changes needed to fix an issue.

  **What Chat does not do without approval:**

  * Apply mutating changes directly to your environment.
  * Restart, scale, or delete workloads on its own.

  Every recommended action is gated by Guardian policies and needs explicit human sign-off before it lands. See [Safety and Guardrails](/concepts/guardrails) for the full picture.
</Info>

***

## Tips for Effective Queries

<Accordion title=" DO: Be Specific">
  **Good:** - "Show me OOMKilled pods in production"

  * "Why did payment-service-v2 fail?"

  **Why:** Specific queries = specific answers
</Accordion>

<Accordion title=" DON'T: Be Vague">
  **Avoid:** - "Fix everything"

  * "What's wrong?"

  **Why:** Agent needs context
</Accordion>

<Accordion title=" DO: Use Natural Language">
  **Good:** - "What's broken?"

  * "Why is this pod unhappy?"

  **Why:** Agent understands casual phrasing
</Accordion>

<Accordion title=" DON'T: Use kubectl Syntax">
  **Avoid:** - "kubectl get pods -n prod"

  **Why:** Just ask! "Show me pods in prod"
</Accordion>

***

## What You Learned

<CardGroup cols={2}>
  <Card title="Chat Interface" icon="comment">
    How to open Chat and navigate the UI
  </Card>

  <Card title="First Query" icon="pen">
    How to type and send your first question
  </Card>

  <Card title="Agent Reasoning" icon="brain">
    What Thoughts and Function Calls mean
  </Card>

  <Card title="Context Awareness" icon="link">
    How Chat maintains conversation thread
  </Card>

  <Card title="Quick Actions" icon="bolt">
    Using pre-built query categories
  </Card>

  <Card title="Best Practices" icon="check">
    What makes a good vs. bad query
  </Card>
</CardGroup>

***

## Next: Investigate Real Incidents

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

<Card title="Continue: Troubleshooting with Chat" icon="wrench" href="/tutorials/chat-troubleshooting">
  Learn to investigate ImagePullBackOff, OOMKilled, and CrashLoop errors using real Chat conversations
</Card>

***

## 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

**Quota:** Questions are unmetered. Only full detect-to-resolution cycles (investigations) count against your plan.

***

## Need Help?

<CardGroup cols={2}>
  <Card title="Support" icon="headset">
    Email: [connect@rubixkube.ai](mailto:connect@rubixkube.ai)
  </Card>

  <Card title="Concepts" icon="book">
    [Learn about the SRI Agent](/concepts/agent-mesh)
  </Card>
</CardGroup>