1. Two Complementary Assistance Tools
Accelerated Logic AI provides two unintrusive companion utilities to guide your workspace experience:
A dynamic visual status indicator that reflects real-time agent pipeline activity and offers 1-click contextual prompt suggestions.
An in-app searchable FAQ center with direct action buttons that jump straight to relevant settings panels or doc guides.
2. Robot Companion — Visual Mood States
The Robot Companion widget changes its expressive animations according to workspace state:
3. Contextual Prompt Suggestions
Clicking the Robot Companion opens a popover card with 1-click prompt recommendations based on your current workspace context (e.g., "Make mobile responsive", "Add dark mode toggle", "Export to GitHub").
4. Help Companion & Interactive FAQ Accordions
Open the Help Center from the header (? icon) to search categorized FAQs:
Categorized Filters: Filter by Getting Started, Models & Workflow, Features & Storage, and Troubleshooting.
Direct Action Navigation: Accordion items feature direct action buttons (e.g., "Go to Model Organizer") that jump immediately to relevant settings tabs.
5. How Robot & Help Work Together
The Robot Companion alerts you to real-time status transitions (e.g. showing an Alert pulse if Ollama disconnects), while the Help Center provides step-by-step diagnostic solutions and automated shortcuts to resolve the issue.
6. Privacy & Customization
The companion and Help Center display workspace status and help in your browser. Actions that connect a model or open an external service follow the data handling described in the Privacy Policy.
You can hide the Robot Companion at any time via the companion popup or Settings → Display.
7. Troubleshooting Companions
Robot Hidden? Re-enable it in Settings → Display → Robot Companion Widget.
Mobile Layout: On ultra-narrow screens (<350px), the Robot Companion auto-hides to conserve composer screen space.
8. The Help Center FAQ — What It Covers and How to Query It
The Help Center is a curated, categorized index — not a general-purpose chatbot. It rewards precise queries:
- Query with symptoms, not stories: "ollama disconnected", "drive 401", "white screen sandbox" — the search matches FAQ keywords, so the telegraphic error string beats a paragraph.
- Category filters narrow fast: switch to Troubleshooting when you have an error, Features & Storage when you have a "where did my data go", Models & Workflow for behavior questions.
- Action buttons jump straight to the fix: many accordion answers embed a button (e.g. Go to Model Organizer, Open Export) that closes Help and lands you on the exact settings panel — use it instead of navigating manually.
- Coming up empty? The FAQ is finite by design. If nothing matches, your fastest path is the docs (you are in them), then Settings search (Find a setting..., Part 20), then GitHub issues for the open-source workspace.
9. Robot Mood States During Multi-Agent Runs
The mood is a compressed view of pipeline health. Reading it correctly during a long run:
| Mood | What the pipeline is doing | Should you act? |
|---|---|---|
| Idle → Thinking | Orchestrator decomposing your prompt | No — decomposition is normally a few seconds |
| Thinking (sustained) | Planner/critic loop, possibly multiple Infinity attempts | Peek at the activity drawer; if the same finding repeats 3 times, stop and tighten the prompt |
| Coding | Developer agent(s) streaming code | No — let it finish; interrupting here strands half-written files |
| Success | Finalizer committed output | Review, rate, or iterate normally |
| Alert | Provider error, fallback trip, or blocked tool | Yes — open the drawer or the error card; the Alert pulse is a prompt to look, not to panic |
10. When the Robot Is Wrong — Signals, Not Summons
The Robot reports pipeline events, not your feelings and not absolute truth. Three situations where its signal deserves skepticism:
- Alert ≠ broken. A single transient 429 on a free route may self-heal via the built-in routing. Check the outcome before re-sending — duplicating a prompt after a one-off blip just doubles the token spend.
- Success ≠ correct. A green bounce means the run completed, not that the code works or the facts are right. Verification is still your job (Parts 7, 10, 19).
- Idle ≠ ready for anything. An idle robot says nothing about whether your local model is still resident or your quota is exhausted — start the prompt and read the first trace entry to confirm health.
If the widget itself is distracting during long flows, hide it via Settings → Display — the Help Center remains available from the header either way.