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

# Backline Assistant

> Your in-product AI companion for security investigation and insights

## Overview

Backline Assistant is an in-product, context-aware AI companion designed to help you understand and investigate security insights directly inside Backline.

It allows you to ask questions in natural language, explore vulnerabilities and remediations, and receive explainable answers based on:

* Backline's security knowledge base
* Your organization's environment and findings
* Integrated external sources (when relevant)

<Note>
  The assistant is designed to support decision-making and investigation—not to automatically change data or perform destructive actions.
</Note>

## Typical Use Cases

The Assistant is accessible from the application header and is intended for open-ended questions and multi-topic investigations:

* "Show me the highest-risk remediations this week"
* "Which vulnerabilities are about to breach my SLA?"
* "Show me all my remediations that are waiting for approval"

This mode is ideal when you want to explore across multiple vulnerabilities, remediations, or domains.

## Access & Layout

<Steps>
  <Step title="Open the Assistant">
    Click the assistant button in the application header.
  </Step>

  <Step title="Floating Mode">
    When a side panel is already open (Vulnerability or Remediation), the assistant appears as a floating button above it.
  </Step>

  <Step title="Expand View">
    Use the **Expand** option to switch to a full-page assistant view for deeper investigations.
  </Step>
</Steps>

## Conversations & History

### New Chat

Selecting **New Chat** starts a fresh conversation. The previous conversation is automatically saved to history.

### Chat History

* All completed chats are stored in your **Chat History** panel
* You can reopen previous conversations at any time
* Chats persist across user sessions

## Asking Questions

### Input Behavior

| Action          | Result                  |
| --------------- | ----------------------- |
| `Enter`         | Sends the message       |
| `Shift + Enter` | Creates a new line      |
| Empty field     | Send button is disabled |

<Tip>
  The input field supports multi-line messages for complex questions.
</Tip>

### Suggested Questions

When opening a new chat, the assistant may display starter suggestions:

* What are the top 10 critical vulnerabilities in my environment right now?
* Which remediations are about to breach SLA or already overdue?
* Which remediations are waiting for pending approval?

## Context Chips (@ Mention)

Typing `@` in the input box opens a selector that allows you to attach context:

| Context Type    | Description                                             |
| --------------- | ------------------------------------------------------- |
| Vulnerabilities | Attach specific vulnerabilities for focused analysis    |
| Remediations    | Attach remediations to discuss upgrade paths or impacts |

<Info>
  **Rules:**

  * Maximum **5 chips** per chat
  * Searchable by title
  * Context chips help the assistant provide more precise and relevant answers
</Info>

## Recommended Next Steps

Each assistant response ends with **1–3 recommended step chips**. These help you continue your investigation efficiently.

### Investigation Steps

Examples of recommended actions:

* "See related remediations"
* "Show dependency path"
* "Explain this CVE"

<Tip>
  Clicking a step automatically moves it into the input box for quick follow-up.
</Tip>

## What You Can Ask

The assistant is optimized for security analysis, investigation, and triage:

<CardGroup cols={2}>
  <Card title="Vulnerability Investigation" icon="magnifying-glass">
    * Exploitability explanations and evidence
    * Severity reasoning
    * Impact assessment
  </Card>

  <Card title="Remediation Guidance" icon="screwdriver-wrench">
    * Breaking changes summary
    * Safe upgrade paths
    * Dependency analysis
  </Card>

  <Card title="Risk & Prioritization" icon="chart-line">
    * SLA risks
    * High-impact remediations
    * Priority recommendations
  </Card>

  <Card title="Environment Analysis" icon="network-wired">
    * Cross-repository insights
    * Trend analysis
    * Compliance status
  </Card>
</CardGroup>

## Limitations

<Warning>
  The assistant is **view-only and advisory**. It cannot perform actions that modify your data.
</Warning>

The assistant **cannot**:

| Restriction         | Description                                             |
| ------------------- | ------------------------------------------------------- |
| Modify data         | Cannot modify or delete vulnerabilities or remediations |
| Change settings     | Cannot change SLA values or configurations              |
| Cross-tenant access | Cannot perform cross-tenant analysis                    |
| Data boundaries     | Cannot access data outside your tenant                  |
| Execute code        | Cannot execute code or infrastructure changes           |

## Best Practices

<AccordionGroup>
  <Accordion title="Use Context Chips for Precision">
    When asking about specific vulnerabilities or remediations, use `@` mentions to attach them. This helps the assistant provide more accurate and relevant responses.
  </Accordion>

  <Accordion title="Follow Recommended Steps">
    Take advantage of the recommended next steps at the end of each response. They're designed to guide you through a complete investigation workflow.
  </Accordion>

  <Accordion title="Leverage Chat History">
    Don't start from scratch every time. Reopen previous conversations to continue investigations or reference past findings.
  </Accordion>

  <Accordion title="Be Specific in Your Questions">
    Instead of "Show me vulnerabilities," try "Show me critical vulnerabilities in production repositories that are about to breach SLA."
  </Accordion>
</AccordionGroup>