A calm digital space where thoughts feel safe to be shared.
🌿 Your companion Nira is here to listen.
Built with emotionally-aware AI, real-time UI adaptation, and human-centered design.
👉 Try TalkSpace Live: https://talkspace-ai.streamlit.app/
TalkSpace is designed as a calm digital environment for conversation.
Unlike traditional chatbots focused on quick answers, TalkSpace focuses on:
- helping users feel heard
- maintaining natural, human-like dialogue
- offering gentle, supportive responses
The goal is not just interaction -- but emotional presence.
🌿 AI Companion – Nira
- Conversational agent designed for warmth and empathy
- Supports dual modes: Supportive Friend / Counselor
🧠 Emotion Awareness
- Detects emotional signals (loneliness, stress, anxiety, etc.)
- Adapts tone and response dynamically
🎨 Emotion-Aware UI
-
Sidebar reflects real-time emotional state
-
Includes:
- 🌿 “Right now” summary
- 🧘 Mood ring (color-based state)
🚨 Crisis Detection System
- Identifies high-risk emotional signals
- Overrides normal flow with supportive responses
- Updates UI to reflect critical state
💬 Conversation Intelligence
- Maintains short-term conversational memory
- Context-aware responses using recent messages
🗂 Multi-Conversation Support
- Create, switch, and persist multiple chats
- Backed by Supabase
🌬 Calm Tools
- Guided breathing
- Grounding interactions
🌱 Gentle Reminders
- Soft affirmations for a supportive environment
TalkSpace uses a hybrid approach combining machine learning with conversational logic.
User Input → Emotion Detection → Response Strategy → LLM Generation → Memory Update → UI Sync
| Component | Description |
|---|---|
| Emotion Detection | Identifies emotional signals using keyword logic |
| Response Strategy Engine | Selects response type based on emotion + intensity |
| LLM Layer (OpenRouter) | Generates natural, human-like responses |
| Conversation Memory | Maintains short-term conversational context |
| Crisis Detection Layer | Overrides flow for high-risk situations |
| UI Layer (Streamlit) | Real-time emotional visualization |
- Built a real-time emotion tracking system that synchronizes UI and conversation state
- Implemented state-driven UI rendering using Streamlit session management
- Developed a priority-based response system (crisis overrides normal flow)
- Integrated LLM-based response generation with contextual memory
- Engineered multi-conversation persistence using Supabase
- Optimized for natural conversation flow by reducing repetition
| Category | Tools |
|---|---|
| Language | Python |
| Framework | Streamlit |
| ML Library | Scikit-learn |
| Database | Supabase |
| LLM API | OpenRouter |
git clone https://github.com/Keertilata20/talkspace-ai
cd talkspace-ai
pip install -r requirements.txt
streamlit run app.py- animated mood ring based on intensity
- deeper emotional understanding
- improved long-term memory
- voice-based interaction
- enhanced calming experiences
TalkSpace is not a medical or diagnostic tool.
If you are experiencing severe emotional distress, please consider reaching out to a trained professional or someone you trust.
Built with care while exploring:
AI • conversation design • human-centered technology
🌿