How to use Config Analysis & Explain in rConfig V8

Config Analysis & Explain puts an AI assistant next to every device configuration in rConfig. Inherited a firewall nobody documented? Ask it to walk you through the access lists. Reviewing a switch before an audit? Ask for a security review of the management plane. The assistant reads the whole config, so you can ask about it the way you’d ask a colleague who has it open in front of them.

Prerequisites

• The AI Center set up with an active provider (see Getting Started).
• Config Analysis & Explain turned on in AI Configuration.
• You’ve accepted the AI Center terms, and you have permission to view the device’s configs.

Three ways to ask

Method	Best for	Answer appears in
Ask AI toolbar button	Questions about the whole config, and follow-up conversation	The AI Assistant side panel
Lightbulb menu on a line or selection	One-click explain, issues, improvements, or security review of a block	The AI Assistant side panel
Inline prompt (Ctrl/Cmd+K)	A quick question about specific lines without leaving the editor	A box inside the editor, under the lines

Ask AI: chat about the whole config

1. Open a device, then open its configuration in the config viewer.
2. Click the Ask AI icon in the toolbar.

The AI Assistant panel opens on the right. The header shows the config ID and how many lines were loaded, and the status line shows AI Online with the provider and model in use. The assistant greets you with where the config lives and its size, then waits for your question.

Lightbulb menu: act on a block of config

Hover over a line, or select several lines, and a 💡 appears beside them. Click it and choose:

Action	What the assistant is asked
Explain configuration	“Please explain what this configuration does and how it works”
Find issues	“Please analyze this configuration for potential issues, misconfigurations, or problems”
Suggest improvements	“Please suggest improvements and best practices for this configuration”
Security analysis	“Please perform a security analysis of this configuration and identify any vulnerabilities”
Ask custom question…	Opens the inline prompt so you write your own

The first four open the AI Assistant panel and send the request with your selected lines straight away. The assistant still has the full config for context, so it can tell you that an ACL you selected is applied to a particular interface elsewhere in the file.

Inline prompt: a quick question in place

Select some lines and press Ctrl+K (Cmd+K on macOS), or choose Ask custom question… from the lightbulb. A box titled Ask AI about lines N-M opens under your selection with a starter question you can edit. Press Enter to send or Esc to cancel. The answer appears in an AI Response box in the editor.

Inline prompt in action with selected lines and AI response

Worked examples

These examples show the kind of question that gets a useful answer. The screenshots are real responses; your wording and model will produce different text.

Example 1: Understand an unfamiliar config

Scenario: you’ve been handed a branch router you’ve never seen.

Ask: “Give me a summary of this router: its role, interfaces in use, routing protocols, and anything unusual.”

A structured summary of an unfamiliar router

Detailed analysis of the selected router lines

Good follow-ups: “Which interfaces are shut down?”, “What would break if I removed ACL 110?”

Example 2: Security review before an audit

Scenario: an auditor wants evidence that management access is locked down.

Ask: select the line vty, snmp-server, and aaa sections, then choose Security analysis from the lightbulb.

Security analysis of the management plane

Suggested configuration change for the VTY section

Good follow-ups: “Rewrite the VTY section to allow SSH only from 10.0.0.0/8.” The assistant can draft config, but it never pushes anything: you copy what you want into your own change process.

Example 3: Explain one block

Scenario: a BGP config with several match and set clauses is behaving unexpectedly.

Ask: select the BGP config, then choose Explain configuration.

Clause-by-clause explanation of a BGP config

Assistant’s reasoning for the BGP config explanation

Getting better answers

• Be specific. “Is BGP to 192.0.2.1 using authentication?” gets a better answer than “Check BGP.”
• Ask for the format you want: a table, five bullets, or “explain it for a junior engineer”.
• Select the lines that matter for long configs. The assistant sees the whole config, but pointing at the relevant block focuses the answer.
• One topic per conversation. Click New chat (tooltip Start a new conversation) when you change subject. Long mixed conversations dilute answers.
• Ask it to show its evidence: “Quote the lines that support that.” This makes answers easy to check.

Check before you act

AI answers can be wrong, especially on vendor-specific syntax or very long configs. Treat suggestions as a starting point and verify them against vendor documentation before making changes.

Conversation history

Conversations are saved per user and per config. Reopening Ask AI on the same config resumes your latest conversation. Click the history icon (Past conversations) to open an earlier one, or New chat to start fresh.

If the config has changed since a conversation started, a yellow banner appears above the input: The configuration has changed since this chat started. Replies may reference the older version. Click Start a fresh conversation to work from the current config.

Past conversations and the config-changed warning

What gets sent to your provider

On the first message of a conversation, rConfig sends your question, the device config, and some config metadata. Configs up to about 100,000 characters are sent whole; larger ones are trimmed to the most relevant sections. Follow-up messages send only the new question; the config isn’t sent again. If configs contain secrets you don’t want leaving your network, use a local provider such as Ollama. See Governance, audit & privacy for every capability.

Limits

Limit	Value
Length of one question	8,000 characters
Config sent whole	Up to about 100,000 characters; larger configs are trimmed by section
Config sent with the inline prompt	Up to 50,000 characters; above that, only your selected lines are sent

Common gotchas

Common issues

• Ask AI button missing: check that Config Analysis & Explain is on in AI Configuration, that you’ve accepted the terms, and reload the page.
• Panel says “No AI provider activated — messages cannot be sent”: no provider is active. An admin needs to activate one in AI Center → AI Settings.
• “Config not loaded” in the header: the assistant couldn’t read the config, so it can only answer general questions. Close the panel and reopen it, or check the config file is present on the server.
• Answers about a huge config miss details: select the section you care about, or ask about one area at a time.

For any other message, see Troubleshooting.

Better answers

Getting better answers from config analysis shows what Iris reads and how to ask questions that get specific, checkable answers.

What’s next

• Narrate a config change with Drift Detection
• Teach the AI your conventions
• Review what data each capability sends to your provider