Where to Find It
Ask Cielara appears in two forms. The floating panel is available on every page of the dashboard as a collapsible right-side panel, a draggable floating window, or a full-width expanded view; toggle between these with the view picker in the header. The floating window is draggable and resizable, and its size and position are saved between sessions. The workspace page gives you a dedicated three-column layout. On mobile, the panel opens as a full-screen overlay.Workspace Layout
The workspace has three columns. The left column holds your chat sessions grouped by time (Pinned, Today, This Week, Earlier), with controls to create, pin, search, and delete them. The center column is the chat interface with the active conversation. The right column is a canvas pane for Mermaid diagrams and investigation reports the assistant produces. The canvas opens on its own as soon as the assistant produces something to show, and stays visible as you keep chatting, so you can reference a diagram while asking follow-ups. Mermaid diagrams support pan, zoom, and SVG download.Chatting
Type a question in the input field at the bottom. The assistant streams its response as it works, so you can see the answer take shape rather than waiting on a blank screen. Example questions:- “What services depend on the payments API?”
- “Show me the blast radius if Redis goes down”
- “What is our incident escalation policy?”
Message Rendering
An answer is built from three kinds of blocks. Text blocks render as Markdown, with tables, headings, lists, and code blocks you can copy. Tool call blocks show the steps the assistant took to reach its answer, such as querying the World Model or searching your memories; expand one to see what it looked at and what came back. Embed blocks pull other pages inline, so a topology answer can show you the relevant part of the World Model without leaving the conversation. Below each response, layer chips show which parts of the model the assistant drew on, and suggestion chips offer follow-up questions you can click to ask.Starter Suggestions
When you open a new session, the assistant shows generated starter suggestions so you don’t have to think of a first question. These are categorized by source type.Pinning Context with @-Mentions
Type @ in the input field to open a context picker. You can pin pages (Production World Model, Ontology Explorer, Incidents, Simulations, Memory Explorer, and others), services (workloads configured for RCA), incidents from your alert system, stored simulation runs, or prior chat sessions. Pinned items stay attached until you remove them, and the assistant considers them when answering.Effort Levels
Use the slash command/effort followed by a level to control how deeply the assistant reasons:
Higher effort levels let the agent take more reasoning steps, which produces more thorough answers but takes longer.