Getting Started
Help Assistant & Public Help Centre Chat
Last updated 19 May 2026
Help Assistant & Public Help Centre Chat
Two AI help surfaces share a similar pattern but serve different audiences. Help Assistant is the in-app chat for logged-in operators with venue context. Public Help Centre Chat is the anonymous version on the marketing /help page. Both are free.
Help Assistant: in-app — sidebar Help launcher (web) or More → Help (mobile). Public Help Centre Chat:
https://openchairpro.com/help.
TL;DR
- Help Assistant: in-app, authenticated, venue-aware, sessions persist per user/venue. Thumbs up/down feedback.
- Public Help Centre Chat: on
/help, anonymous, RAG over published help articles. Source links + follow-up chips. - Both free, both run on
gpt-5.4-miniwith prompt-leak detection. - All plans for both.
Help Assistant (operators)
The in-app AI help chat for logged-in operators. Knows your venue context — your services, your team size, what plan you're on, what features are enabled.
Where to find it
- Web: Help launcher in the sidebar Workspace Hub, or the floating support button on most pages
- Mobile: support chat sheet in the More tab → Help
What it knows
- Your venue's name, country, industry
- Your tier (FREE / Trial / PRO)
- Features you have enabled (Calendar Sync, AI Concierge, Storefront tracking, etc.)
- The help article catalogue
- Common settings paths
It does not know your customer data, booking history, or financial data — that's for the Coach (your business advisor) or Intelligence Chat (your data Q&A). Help Assistant is about how to use OpenChair, not about your data.
Sessions
Sessions persist per (user, venue) — up to 20 sessions retained. Open the help launcher tomorrow and your last conversation is still there. Switch venues and you get a fresh session set for the new venue.
Feedback
Each AI response has thumbs up / thumbs down buttons. Feedback flows to our team via Langfuse and informs prompt iteration. Useful when:
- The answer was wrong → thumbs down
- The answer was unclear → thumbs down + add a note
- The answer was perfect → thumbs up (calibrates our quality signal)
We use these signals to improve help quality, not for individual support tickets — for live support, contact us via email or chat with a human.
Examples of what to ask
- "How do I set up a deposit rule for first-time clients?"
- "Where do I configure my brand voice persona?"
- "What's the difference between Smart Waitlist and Concierge gap-fill?"
- "Why isn't my dedicated SMS number provisioning?"
- "How does FREE plan caps work?"
Security and safety
- Prompt-injection guards via canary token + Slack alert on leak detection
- Your venue data isn't sent off-platform
- The system prompt explicitly forbids the AI from revealing internal infrastructure details
Public Help Centre Chat (unauthenticated)
A second chat surface lives on /help (marketing site, not the app). Available to anyone — prospects evaluating OpenChair, customers of operators looking for help with bookings, or operators who aren't signed in.
Where to find it
https://openchairpro.com/help — the marketing help centre. The chat is a chat-bubble component embedded in the page.
What it knows
- The published help-article catalogue (every doc under
docs/help/) - Generic OpenChair feature knowledge
It does not know any venue or user context — it's anonymous, no signed-in identity.
How answers are sourced
The chat uses retrieval-augmented generation (RAG) over the help article corpus. Every answer:
- Is grounded in specific help articles
- Includes source links to the relevant docs
- May suggest follow-up chips for related questions
Rate limits
To prevent abuse, the public chat is IP-rate-limited — typical users won't hit the limit, but scripted abuse triggers a temporary slow-down.
Examples of what to ask
- "Can I take bookings on my own website?"
- "How does pricing work?"
- "What's the difference between FREE and PRO?"
- "Do you support Tap to Pay in New Zealand?"
If you're an existing operator and want venue-specific help, use the in-app Help Assistant instead (it knows your context).
Privacy and data handling
| Surface | Auth | Stores |
|---|---|---|
| Help Assistant | Logged-in user | Per (user, venue) session history — up to 20 sessions, retained until you clear them |
| Public Help Centre Chat | Anonymous | No persistent session; IP for rate-limiting only |
Both surfaces use the same Slack-alert prompt-leak detection so any internal prompt content leaking back through the model is flagged immediately.
Tier
Both surfaces are all plans / free. Help isn't a paid feature — we want every operator and every prospect to get answers quickly.
When to use which
| Use | Surface |
|---|---|
| "How do I configure X in my venue?" | Help Assistant (in-app, knows your setup) |
| "What does OpenChair do?" | Public Help Centre Chat |
| "Why isn't my specific feature working?" | Help Assistant (then escalate to support if AI can't resolve) |
| Browsing while evaluating OpenChair | Public Help Centre Chat |
| Customer of an operator looking for help with their booking | Probably contact the venue directly — the public chat doesn't know about specific bookings |
What neither surface replaces
- Live human support — for billing disputes, account recovery, or anything urgent, contact human support via email (
support@openchairpro.com) or in-app support panel. - Coach — for business advice grounded in your venue data (Coach).
- Intelligence Chat — for analytics questions about your real data (Intelligence Chat).
- Operator Concierge — for AI-proposed actions on your venue (Concierge).
Help surfaces are about using OpenChair. The other AI surfaces are about running your business with OpenChair.
Common mistakes
| Problem | What to check |
|---|---|
| Help Assistant said something wrong | Thumbs down with a note. We use these signals to improve. For urgent issues, contact human support. |
| Public Help Centre Chat doesn't have the context I need | Use the in-app Help Assistant when signed in — it knows your venue. The public chat is anonymous on purpose. |
| Want to find a specific help article | Use the Help Centre search at /help or browse by category. The chat is for natural-language questions; search is for known topics. |
| Help Assistant keeps suggesting features I'm not on | Confirm the AI knows your plan — if it keeps suggesting PRO features when you're on FREE, that's a sync issue; report it via thumbs down. |
| Want to disable help notifications | The help surfaces don't send notifications — they're chat surfaces only. |
FAQ
What's the difference between Help Assistant and the public Help Centre chat?
Help Assistant is the in-app chat available to logged-in operators — it knows your venue context and can answer questions about your specific setup. The public Help Centre chat is on the marketing /help page; it's anonymous, no venue context, just RAG-grounded answers over our help articles. Operators usually want Help Assistant; prospects and unauth'd visitors use the public chat.
How do I open Help Assistant?
Web: the help launcher button in the sidebar Workspace Hub, or the floating support button on most pages. Mobile: the support chat sheet in the More tab. Sessions persist per user per venue — open it tomorrow and your last conversation is still there.
Do I get charged Sparks for help chats?
No. Help Assistant and the public Help Centre chat are both free. Sparks apply to operator-facing AI features (Concierge actions, campaign drafts, brand voice, etc.), not to help support.
Is the AI's answer reliable?
The AI is grounded in our help articles and your venue context. It includes source links so you can verify. Both surfaces let you give thumbs up / thumbs down feedback that flows to our team to improve the answers.
Can the AI take actions for me?
No — Help Assistant and the public chat are read-only. They answer questions but don't change anything in your venue. For AI that takes actions, use the Concierge.
Will the AI know about features that just shipped?
The help catalogue is updated as features ship — usually within a day or two. If you've just seen something new in a release note and the AI doesn't know, give it a day or report via thumbs down.