Week 6 #

Project name: VoiceDiary

Code repository: link

VoiceDiary is an AI-powered voice journaling tool that analyzes tone, emotion, and key themes in spoken entries. It generates personalized emotional insights and well-being suggestions based on recorded reflections.

Team #

Team member	Telegram alias	Innopolis Email	Responcibilities
Dziyana Melnikava	@meldilen24	dz.melnikava@innopolis.university	PM, Frontend
Anastasia Kuchumova	@n_rngk	a.kuchumova@innopolis.university	Frontend, UX/UI
Dzhamilia Fatkullina	@jam11a	d.fatkullina@innopolis.university	ML
Elina Kuzmichyova	@lin_anile	e.kuzmichyova@innopolis.university	ML
Olesia Novoselova	@doiwannaknoww8	o.novoselova@innopolis.university	Backend
Danil Davydyan	@chocop	d.davydyan@innopolis.university	Backend

Links #

Deployment: Link
API Docs: Swagger, README.md
Design&Demo: Link

Project Overview #

Voice Diary is a voice-based journaling application designed to make emotional self-tracking more natural, engaging, and insightful. Unlike traditional text-based mood trackers, this app lets users record short voice messages and uses AI to analyze vocal tone and content to determine the “Emotion of the Day”, provide personalized recommendations, and identify emotional patterns over time.

This app is intended for users who are interested in monitoring their emotional wellbeing but want a more interactive and personalized experience. It leverages voice as a natural medium for expression and uses machine learning models to extract emotional insights.

✅ Solves the Problem of: #

Low engagement in mood tracking apps
Lack of personalization in emotional wellness tools
Time-consuming journaling habits

Features #

The following features are fully implemented in this version of the app:

🎙️ Voice-based Emotion Detection
Record a voice note and get real-time emotion classification.
📊 Emotional Pattern Tracking
View your recent emotional history and observe trends over time.
🧠 Personalized Recommendations
Receive daily suggestions based on your emotional state and patterns.
✍️ Manual Emotion Input
Override or set your emotion manually if desired.
📁 Secure Diary Storage
Each entry is saved securely in your personal log for future reflection.

Tech Stack #

Backend: #

Go with Gin framework — RESTful API services
Python with FastAPI — AI model serving for emotion classification
PostgreSQL — Voice entry and metadata database

Frontend: #

React.js — Interactive and responsive user interface
Redux — State management for UI and API interactions

AI/ML: #

Audio Pipeline

Whisper Large V3 fine-tuned + j-hartmann → Emotion classification
Whisper Large V3 → Low-latency transcription Analysis Layer
Samsum → Concise daily summaries
OpenHermes-2.5 → Psychological insights

🚀 Setup Instructions #

git clone https://github.com/IU-Capstone-Project-2025/VoiceDiary.git
mv .env.example .env
docker-compose up --build # in root directory of the repository

Presentation Draft #

Link

Individual commitments #

Team member	Telegram alias	Contribution	Link
Dziyana Melnikava (Lead)	@meldilen24	Fixed bugs: secure connection (usage of 443 port, wrote SSL certificates, wrote nginx service for Docker), api for user info update reconnected, New features: added record list displaying, updated design, connected insights and updated form for them	Commits
Anastasia Kuchumova	@n_rngk	Finished CI, updated the calendar, added entries list feature, added documentedation of components	Commits
Dzhamilia Fatkullina	@jam11a	Tested input texts, Improved insight generation LLM prompt, Fixed output cleaning & JSON extraction bug, Prepared draft presentation & Week 6 report	Commits
Elina Kuzmichyova	@lin_anile	finalized testing of emotion by test model, built two versions of emotion orchestration—basic and one with a custom covariance matrix, tested by tweaking transitions (limiting neutral jumps)	Commits
Olesia Novoselova	@doiwannaknoww8	REST endpoints in Go service for: Record deletion, Profile update, User existence check during registration, Refactored database schemas	Commits
Danil Davydyan	@chocop	Fixed insights generation in Go backend, Disabled DB save for onboarding page entries, Fixed Docker Compose (production setup with Nginx), Updated Go tests	Commits

Plan for Next Week #

ML #

Further improve psychological insights prompt
Add emotion aggregation logic to handle days with excessive entries
Finalize README.md

Frontend #

Improve connection safety
Improve UX/UI design
Test and fix bugs
Finalize README.md

Backend #

Model inference optimization
API endpoint refactoring (new contract compliance)
Finalize handler logic (full consistency check)
Verify test coverage & passing status
Finalize README.md

Confirmation of the code’s operability #

We confirm that the code in the main branch:

[✓] In working condition.