Overview
The AI Assistant is a single configuration that powers every AI feature in the app: chat on the client booking site, chat in the client portal, the report summaries you see above every dashboard, and vision-based hints for uploaded clinical images.
You pick the provider (OpenAI, Anthropic, Ollama, LMStudio, or a custom endpoint), the model, the language, and the tone. The same settings apply everywhere the AI is used.
Activation
- 1Dashboard → Add-ons → AI Assistant → Activate.
- 2Open Add-ons → AI Assistant. Toggle Enabled at the top of the settings page.
- 3Configure Provider, Model, and the rest as described below, then Save.
Provider & model
- OpenAI
- Base URL fixed to OpenAI's endpoint. Paste your API key. Models load automatically from your account.
- Anthropic
- Same as OpenAI but for Claude. API key required.
- Ollama
- Local. Point the Base URL at your local Ollama server (default http://localhost:11434/v1). No API key. Click Refresh to reload the list of installed models.
- LMStudio
- Local. Similar to Ollama — point at the LMStudio server URL.
- Other
- Any OpenAI-compatible endpoint. Useful for proxies or custom deployments.
Vision support
For Clinical Imaging AI hints and photo analysis, pick a model that lists vision support. The model dropdown shows context window and modality so you can tell at a glance.
Language & tone
- Dialect — the language the AI replies in. English, Modern Standard Arabic, Egyptian Arabic, Saudi Arabic, and more.
- Tone — friendly, professional, casual. Tweaks the personality without retraining.
Limits & custom prompt
- Max Messages Per Session (1–100) caps the conversation length per chat window. 30 is a sensible default.
- System Prompt Suffix is a free-text field appended to the AI's system prompt. Use it to add house rules — "Never quote prices; always redirect pricing to the front desk," for example.
Where the AI shows up
- Client booking site — a chat widget for FAQs and rescheduling.
- Client portal — logged-in clients chat for self-service.
- Reports — every report has a Generate AI Summary button at the top.
- Clinical imaging — the vision model analyses uploaded X-rays and photos (Clinical Imaging add-on must also be on).
- Medical records — AI-suggested diagnoses on draft records (Medical Records add-on must also be on).
Tips
- Self-host Ollama if you can't send data to a cloud provider — the local provider option keeps everything on your server.
- Use the System Prompt Suffix to encode your brand voice and any compliance rules.
- Start with a small model and upgrade if you feel the responses are shallow. Bigger models are slower and more expensive without always being better for short business queries.