MindScape AI
AI-driven emotional wellness and mental-health companion.
TL;DR
An intelligent emotional wellness companion combining NLP sentiment analysis with a hybrid dialog engine (Rasa + heuristic fallback), rule-based clinical scoring, and crisis detection with adaptive UI alerts.
- Context Processing
- <1.2s TTI (LangChain Pipeline + Vector Index)
Problem
Wellness chatbots fail exactly when they matter most — the model errors out, or it misses language that signals a crisis and replies with small talk.
Approach
- 01
Combined a Rasa dialog engine with a heuristic fallback so the assistant always responds, even when the model layer is unavailable.
- 02
Scored sentiment with TextBlob and layered rule-based clinical scoring on top of the raw signal.
- 03
Added crisis-phrase detection that switches the interface into an adaptive alert state.
- 04
Visualized burnout trends in Plotly, with voice input/output via SpeechRecognition and pyttsx3.
Key architectural decision
Chose a hybrid Rasa + heuristic fallback engine over pure LLM APIs to guarantee zero-latency response during crisis detection and reduce API cost overhead.
Impact
- Zero-downtime conversational path through the hybrid engine design.
- Real-time burnout analytics turn scattered check-ins into a visible trend line.
- Crisis detection escalates the UI instead of burying the signal in chat history.
What shipped
Hybrid dialog engine
Rasa primary path with deterministic heuristic fallback.
Clinical scoring layer
Rule-based scores over NLP sentiment output.
Burnout analytics
Plotly trend views over longitudinal check-ins.
Stack
- Python
- Streamlit
- TextBlob
- Rasa
- Plotly
- SpeechRecognition
- pyttsx3