Riverbed IQ Ops Chat

Chat is where you hold a working conversation with Riverbed Q (Q), the default Riverbed IQ Ops assistant, or with a published catalog agent you choose for the session. If you have never used an AI chatbot at work, start with How AI chat works below, then return here for the Chat page controls.

For the bigger picture (Skills, Agents, Feed), see AI-Assisted Operations in Riverbed IQ Ops.

How AI chat works

An AI chatbot is software that reads what you type (and sometimes files you attach), applies a large language model, and writes a reply. It does not run your network by itself. It suggests explanations, steps, and drafts you can verify against Riverbed IQ Ops data, runbooks, and your own judgment.

In Riverbed IQ Ops, Chat is built for operations work:

You ask questions in everyday language (for example, how to investigate latency or how to structure a runbook branch).
Q may call tools behind the scenes (shown in the thread as tool activity) to fetch summaries or run allowed actions defined by Riverbed and your catalog.
You can invoke published Skills and Commands Berkeley Packet Filter. A syntax for selecting which network packets are captured or harvested based on protocol, address, port, and other packet attributes. from the Skill Catalog using the Commands picker or quick-start suggestions.
Each back-and-forth happens in a chat session you can rename, search, export, or discard.

Important: Treat every answer as assistance, not authority. Confirm critical decisions against dashboards, runbook output, and change policies before you act in production.

What Q can and cannot do today

What Q can help with

Explaining concepts and suggesting troubleshooting or automation approaches.
Drafting or refining text for skills, runbooks, and operational summaries.
Running catalog-backed commands and skills you or your organization published.
Continuing a thread across multiple turns within the same chat session.

Limitations you should plan for

Not context-aware yet: Q does not automatically know what is on your screen—open incident A collection of one or more related triggers. Relationships that cause triggers to be combined into incidents include application, location, operating system, or a trigger by itself., selected chart, or runbook node. Say what you are looking at (names, IDs, time range, symptoms) in your message. Do not assume Q sees the same UI you do.
May be wrong or incomplete: Language models can hallucinate details. Verify facts before you change production systems or notify customers.
Bounded by tools and catalog: Q can only use integrations and catalog entries your tenant exposes. If a skill is draft or unpublished, it will not behave like a production command.
Separate from legacy Data Forensics chat: Older AI chat panels in Data Forensics are a different feature. Click Q in the left menu to open Chat with Q. See Chat, Skills, and Riverbed IQ Ops Assist.

Models, data, and who pays for compute

Riverbed IQ Ops does not show the name of the underlying language model in the Chat UI. Riverbed operates the inference services behind Q as part of your entitled platform offering, using large language models provided through Riverbed subprocessors (similar to Riverbed IQ Ops Assist). The exact models and versions may change over time as Riverbed improves the service.

Who pays: Inference for entitled Chat usage is provided through your Riverbed IQ Ops subscription and Riverbed's hosted services—not as a separate meter you configure in the Chat window. Your organization does not bring its own API key in the Chat UI. For data handling, residency, and AI-in-products policies, see the Riverbed Trust Center and, where applicable, the data-region guidance on Riverbed IQ Ops Assist Configuration (Assist and Chat may share similar subprocessors even though they are different features).

Information you send in Chat (messages, attachments, and metadata you trigger through commands) is processed to generate responses subject to your chat terms and organizational policies.

Open Chat and accept terms

To open Chat:

Click Q in the left menu.

The page title begins with Chat Session with. When you chat with the default assistant, the full title is Chat Session with Riverbed Q.

Use the agent chooser in the page header to start a session with a catalog agent instead of Riverbed Q. For example, choose an agent authored to troubleshoot a specific application. The title then shows Chat Session with followed by that agent's name. See Agents.

Before your first message, accept the terms in the Orion chat terms dialog. If terms change, you may need to accept again. Review acceptance under User Preferences (chat terms section).

Chat sessions

Each conversation is a chat session. From the session list you can:

Start a New Chat.
Search conversations with Search conversations….
Show or hide the session list with Show All Chats and the sessions sidebar controls.
Rename, download a transcript, or discard a session from the row menu.

Composer and commands

Type in Ask anything… to send a message. Be explicit about context (what broke, for whom, since when, which app or site) because Q is not context-aware yet.

Optional capabilities include:

Voice Input for speech-to-text where your browser supports it.
Commands (slash commands) to invoke published catalog Skills and Commands from the picker.
File attachments (subject to size and type limits shown in the UI).

Quick starts on the composer help you begin common analysis and troubleshooting scenarios before you send your first message. See Chat quick starts.

Responses, feedback, and long conversations

Q responds with answers that may include tool activity summaries, suggested next steps, and partial responses while generation is in progress. Use Mark helpful or Mark not helpful on a message to send feedback. Use Stop generating to cancel a long response.

When older messages are summarized

Very long chat sessions can exceed what the model can hold in memory at once. When that happens, the Riverbed-hosted chat service compacts the conversation: earlier turns are summarized so recent messages and findings stay available. The UI shows a notice such as: This conversation was long, so earlier messages were summarized to stay within context. Your recent messages and findings stay fully available. You can dismiss the notice.

There is no fixed message count or length published in the product UI. Compaction depends on how much content is in the session (messages, tool results, attachments). If you need full fidelity for an audit or handoff, download the transcript before the session grows extremely long, or start a New Chat for a new topic.

Embedded Chat

You can also use Q without leaving another page. See Embedded AI Assistant. The same context limitations apply: describe what you are working on in your prompt.

AI-Generation disclaimer

When using Generative AI technologies to create and generate content, the target audience must be aware that the content is AI generated. We recommend adding a disclaimer to the output of the content generated using the GenAI node. For example, appending "AI-Generated content".