Mental Wellness Chatbot: A Supportive Conversational AI Prototype
A non-clinical chatbot prototype designed for supportive check-ins, reflection prompts, and careful conversational boundaries.
- Python
- LLM / NLP
- Conversational UX
- Conversational AI prototype builder
- Prototype
- 2024
- NLP & AI
- Case study only
Overview
The Mental Wellness Chatbot is a non-clinical prototype — a supportive check-in companion for everyday reflection, not therapy, diagnosis, or crisis support. It explores how a conversational assistant can hold a careful, encouraging tone within clear limits.
Problem
People often want a low-pressure way to check in on how they're doing. But a sensitive topic raises the stakes: tone, boundaries, and knowing when to step back matter more than features. The challenge was designing something supportive without overstepping.
Conversation design
The conversation centres on gentle check-ins and reflection prompts, with a consistent, warm tone.
- Supportive check-in and reflection prompts
- A steady, non-judgemental tone
- Encouraging language without false reassurance
Safety and boundaries
Safety came before capability. The prototype is explicitly non-clinical: it does not diagnose, treat, or handle crises, and it is not a substitute for professional care.
- No diagnosis, treatment, or medical advice
- Clear, repeated framing that it is not therapy
- Directs serious concerns toward qualified professionals and emergency support
- Responsible fallback behaviour when a conversation moves out of scope
Technical approach
The build focuses on conversational flow and guardrails rather than raw capability.
- Prompt and flow design around a safe scope
- Guardrails on what the assistant will and won't do
- Python with an LLM/NLP layer for the conversation
Limitations
- Non-clinical and not a substitute for professional care
- Cannot handle crises or emergencies
- Best shown as a case study given the sensitivity of the domain
- Conversational models can still respond imperfectly
What it demonstrates
- Designing for a sensitive domain with care
- Tone, boundary, and fallback design
- Responsible framing of an AI assistant's limits
- Conversational UX thinking
Stack
- Python
- LLM / NLP
- Conversational UX
Proof assets
Some proof assets use dummy data or are shared as private walkthroughs to protect sensitive systems and records.
- Planned
Conversation screenshots with dummy prompts
Example exchanges using dummy prompts.
- Planned
Flow diagram
Conversation flow and boundaries.
- Case study only
Safety boundary notes
How scope and fallbacks are defined.
- Planned
Sanitized demo
A demo running on dummy prompts.
Availability
Case study onlyBest presented as a case study; no live build is published.
Next steps
- User testing with a small, consenting group
- Expand the safety boundary documentation
- Add sanitized conversation examples