Command Center

The Command Center is the unified home screen of SETI AI and your real-time AI workspace — the room where you talk to your power system. Instead of hunting through menus and logs, you ask a question in plain language and your AI operations team answers, grounded in your own documents and live telemetry, with the sources cited.

It is built as a node canvas: every feature is a draggable widget ("node") you arrange to suit how you work. It also acts as the shell for the rest of the app — the top tab bar switches between the canvas and the other major surfaces.

Related pages: Settings · AI Agents · Knowledge Base · Notifications

Top navigation (tab bar)

The tabs across the top switch the whole main view.

The Admin tab stays hidden until the page confirms your role via GET /api/auth/me; admins also see their name in the top-right user badge. Embedded views load on first click (lazy) and stay loaded after that.

Top-bar controls (right side)

The palette (left sidebar)

The palette holds every node type you can place. It has a search box (filter by label), collapsible category sections (collapsed by default), and a collapse toggle for the whole sidebar. To place a node, drag it onto the canvas or single-click it (spawns at canvas center). A node type can only be on the canvas once; trying to add a second copy shows a warning toast.

At the bottom of the sidebar:

Your layout is also auto-saved to the browser (localStorage seti-cc-canvas-v9) on every change, so it persists without Save/Load. Save/Load are for backing up or moving a layout between machines — set up one view for daily operations, another for incident response, another for a customer demo, and switch in seconds.

Palette categories and nodes

Agents

• Agent Status
• Host System

Communication

• Chat with Bob
• Chat with Bill
• Group Chat
• Command Chat (legacy tabs)
• Discord Feed

Knowledge

• Knowledge Base
• Agent Memory

The canvas

• Pan: drag empty canvas space. Zoom: mouse wheel, or the on-canvas zoom toolbar — Zoom In (+), the percent readout, Zoom Out (−), and Fit View (frame all nodes). Zoom ranges 15%–300%.
• Move a node: drag its header. Resize: drag the grip at the bottom-right (most nodes; Host System is fixed-size).
• Delete a node: right-click → Delete Node, or select it and press Delete/Backspace.
• Layout (nodes, positions, sizes) is saved automatically.

Nodes (widgets) in detail

Agent Status LIVE

Side-by-side cards for the two agents (see AI Agents):

• Bob — Senior Ops Lead (command + approval)
• Bill — Research Specialist (analytics + intelligence)

Each card shows an online dot, the active model, and metric cells (GW gateway port, API port, Uptime). Per-agent controls: a Restart button and a model dropdown to hot-swap which Ollama model the agent runs. The model list comes from /api/ollama/tags (embedding models filtered out, cloud models first); live status from /api/status.

Host System .15

Live server vitals — CPU, RAM, Disk as labeled bars, refreshed from /api/status.

Command Chat (legacy tabs) BOB

A single chat window with three channel tabs — Bob, Group, Bill — so you can switch who you're talking to without opening separate panels. The badge updates to the active channel. This is the older combined chat UI; the dedicated panels below are the newer style.

Chat with Bob BOB

Dedicated chat for Bob only — commands, approvals, "is it safe to…", running workflows, fleet status. Type and press Enter; replies render with light markdown (bold, italics, code). Right-click an agent reply to Copy Text. Posts to /api/chat/bob.

Chat with Bill BILL

Dedicated chat for Bill only — analytics, degradation trends, datasheet lookups, monitoring pipelines. Posts to /api/chat/bill.

Group Chat (Bob + Bill) GROUP

One message goes to both agents; each replies in its own bubble (Bob first, then Bill) — you see the operational decision and the analysis behind it side by side. Posts to /api/chat/bob and /api/chat/bill.

Why it matters: the agents are not a generic chatbot. They are pre-loaded with your SETI Power Pack specifications, safety thresholds, and standard operating procedures. Ask "is 57°C on Pack 2 a problem?" and the answer is measured against your hard limits. Each chat panel keeps its own scoped history, so switching panels never mixes conversations, and recent context carries across sessions. A typing indicator shows while an agent responds; timeouts and errors surface inline.

Discord Feed LIVE

Mirrors recent activity from your connected Discord channel onto the canvas, so the team's alert-and-chatter stream lives alongside chat and analytics. Configure Discord in Settings → Discord (see Notifications).

Knowledge Base RAG

The retrieval-augmented (RAG) document workspace — ground the agents in your own documents. See Knowledge Base.

• Upload zone: drag-and-drop or click to upload. Accepted: .pdf, .txt, .md, .docx, .py, .json. Each file is chunked and embedded into a private vector store (/api/upload).
• Ask Your Docs: type a question and hit search. This calls /api/rag/ask, which runs a semantic search over your documents and feeds the top passages to the LLM to produce a written, grounded answer — not a pile of raw hits.
• Cited sources: every answer lists the source filenames it drew from. If nothing relevant is found, the agents are instructed to say so rather than fabricate a citation — critical for an industrial system.
• Stats chips: live doc count and chunk count (/api/rag/stats), plus a list of indexed documents.

The same knowledge base powers the chat agents directly: when you ask Bob or Bill about a product or datasheet, they search the KB first and cite the source filename in their reply.

Agent Memory MEM

A three-column board of recent long-term memories — Bob, Bill, and Team (shared) — pulled live from /api/memory. Lets you see what the agents currently remember and act on, making their accumulated institutional knowledge transparent.

The Command Palette (Cmd+K)

Press Cmd+K (or Ctrl+K on Windows/Linux) anywhere to open a global fuzzy-search modal that jumps you to anything in the platform. It searches across:

• Pages — the top tabs (Command Center, Node Editor, Power Analytics, SETI STACK, Admin)
• Widgets — spawn any node type onto the current canvas
• Saved canvases — your stored layouts (/api/canvas/list)

Start typing and the list filters instantly — matches are ranked exact → prefix → substring → subsequence. Arrow keys navigate, Enter selects, Esc closes, Cmd+K toggles. For a keyboard-first operator it is the fastest way around the entire product.

Standalone Chat page

Separate from the canvas, SETI AI also has a simple full-screen Chat page (chat.html) for talking directly to a raw Ollama model:

• Model selector + Ollama endpoint field (auto-detected), with a connection status dot.
• Streaming responses, auto-resizing input, Enter to send.
• Header nav to Editor, Dashboard, Admin (admins only), and Logout.

This page bypasses the Bob/Bill agent layer and RAG grounding — a direct model chat, useful for quick model testing.

How a conversation is grounded (RAG, end to end)

1. You ask a question in chat (or the KB panel).
2. The system runs a semantic search over your uploaded documents.
3. The most relevant passages are passed to the agent along with its built-in expertise (SETI Power Pack specs, safety thresholds, your fleet topology).
4. The agent composes an answer and cites the source documents.
5. If nothing relevant is found, the agent says so rather than guessing.

The result: answers you can trust and trace — the difference between a chatbot and a command center.

Quick reference — API endpoints

At a glance

Related pages

• Settings
• AI Agents
• Knowledge Base
• Node Editor
• Notifications