Data

Tools

Indexes

Applications

Experiments

Awesome

An open API service indexing awesome lists of open source software.

https://github.com/psycho-baller/audora

real-time AI communication coach that helps you articulate your vision in high-stakes situations
https://github.com/psycho-baller/audora

convex expo nativewind react react-native react-router turborepo

Last synced: 26 days ago
JSON representation

real-time AI communication coach that helps you articulate your vision in high-stakes situations

Awesome Lists containing this project

README 

          
# 🎙️ Audora (LinkMaxxing)



**Your Personal Communication Coach - Powered by AI**



[![License](https://img.shields.io/badge/license-GPL%203.0-blue.svg)](LICENSE)

[![Convex](https://img.shields.io/badge/backend-Convex-orange)](https://convex.dev)

[![React](https://img.shields.io/badge/web-React%20Router%20v7-blue)](https://reactrouter.com)

[![React Native](https://img.shields.io/badge/mobile-React%20Native%20%2B%20Expo-purple)](https://expo.dev)



---



## 🌟 What is Audora?



Audora is an **on-device speech reflection app** that helps you become a more intentional, articulate, and connected communicator. It's like having a personal communication coach in your pocket.



### The Problem We're Solving



Loneliness doesn't come from being alone—it comes from not being understood. Many people struggle to express what they really think or feel, so conversations stay on the surface. From filler words to unfocused rambles, most of us don't realize how much we get in our own way.



### Our Solution



Audora helps you **understand and improve how you speak**, with the ultimate goal of unlocking deeper relationships through better conversations.



---



## ✨ Features



### 🎯 Real-Time Speech Analysis



- **Filler word detection** (um, uh, like, you know)

- **Pacing analysis** (words per minute, pause patterns)

- **Repetition tracking** (repeated words and phrases)

- **Weak phrasing identification** (vague language, weak sentence starters)

- **Clarity & confidence scoring**



### 🗣️ Conversation Recording & Transcription



- **Privacy-first** recording with mutual consent

- **QR code sharing** for easy participant onboarding

- **Speaker diarization** (identifies who said what)

- **Real-time transcription** with Speechmatics

- **Batch processing** for imported audio files



### 🧠 AI-Powered Insights



- **Conversation summaries** generated by GPT-4

- **Fact extraction** (key points from each speaker)

- **Contextual suggestions** for better phrasing

- **Relationship insights** to strengthen connections



### 📱 Cross-Platform



- **Web app** (React Router v7)

- **Mobile app** (React Native + Expo)

- **Shared backend** (Convex)



---



## 🚀 Quick Start



**New team member?** Start here:



👉 **[Onboarding Checklist](./docs/ONBOARDING.md)** - Complete setup checklist for new team members (2-3 hours)



**New to React, React Native, or Convex?** Read this:



👉 **[Complete Beginner's Guide](./docs/GETTING_STARTED.md)** - Step-by-step setup with explanations



**Already familiar with the stack?** Quick setup:



```bash

# Install dependencies

pnpm install



# Set up Convex backend

cd packages/backend

npx convex dev



# Configure environment variables (see docs/GETTING_STARTED.md)



# Run everything

pnpm dev

```



**Useful commands:**



- `pnpm reset` - Clear caches and reinstall (when things go wrong)

- `pnpm clean` - Remove all node_modules and cache folders



---



## đź“š Documentation



### For Beginners



- **[Getting Started Guide](./docs/GETTING_STARTED.md)** - Complete setup for beginners

- **[Architecture Overview](./docs/ARCHITECTURE.md)** - Understanding the tech stack

- **[Troubleshooting](./docs/TROUBLESHOOTING.md)** - Common issues and solutions



### For Developers



- **[Documentation Index](./docs/README.md)** - Complete documentation guide

- **[Web Quick Start](./docs/QUICK_START_WEB.md)** - Run web app in 10 minutes

- **[Mobile Quick Start](./docs/QUICK_START_MOBILE.md)** - Run mobile app in 10 minutes

- **[Setup Instructions](./docs/SETUP_INSTRUCTIONS.md)** - Additional configuration notes



---



## 🏗️ Project Structure



This is a **monorepo** containing multiple related projects:



```

audora/

├── apps/

│   ├── web/          # Web application (React Router v7)

│   └── expo/         # Mobile application (React Native + Expo)

├── packages/

│   ├── backend/      # Convex backend (database + API)

│   ├── ui/           # Shared UI components

│   ├── tailwind-config/  # Shared styling

│   └── feature-home/ # Shared features

└── docs/             # Documentation

```



---



## 🛠️ Tech Stack



### Frontend



- **Web**: React Router v7, TailwindCSS, shadcn/ui

- **Mobile**: React Native, Expo, NativeWind

- **Shared**: TypeScript, Turborepo, pnpm



### Backend



- **Platform**: Convex (backend-as-a-service)

- **Auth**: Clerk

- **Database**: Convex (real-time, serverless)

- **File Storage**: Convex Storage



### AI Services



- **Transcription**: OpenAI Whisper, Speechmatics

- **Analysis**: OpenAI GPT-4, Vercel AI SDK

- **Knowledge Graph**: Zep Cloud

- **Voice Calls**: VAPI



---



## 🎯 Use Cases



### For Professionals



- Improve presentation skills

- Master client communication

- Prepare for important meetings

- Track communication progress over time



### For Social Butterflies



- Make the best of each interaction

- Remember key facts about people

- Build deeper relationships

- Reflect on conversations



### For Language Learners



- Practice speaking clearly

- Reduce filler words

- Improve pacing and confidence

- Get instant feedback



---



## 🌍 The Story Behind Audora



We've always believed that the strongest force in the world isn't data, code, or money—it's the connections we make with each other. As human beings, we're wired to link, to bond, to feel seen. But the modern world makes that hard. We're flooded with noise, trapped in surface-level talk, and robbed of real presence.



LinkMaxxing (now Audora) was born out of a simple but urgent question: *How can we help people truly connect?*



We discovered that loneliness doesn't come from being alone—it comes from not being understood. Many people struggle to express what they really think or feel, so conversations stay on the surface. From filler words to unfocused rambles, most of us don't realize how much we get in our own way.



So we built a tool that gives people the power to understand and improve how they speak, with the ultimate goal of unlocking deeper relationships through better conversations.



## Competition



There are dozens of platforms that promise communication transformation like getfluently.app (a YC-backed company), but there are several problems that are left unsolved for several audiences:



- Young professionals don’t have the time to sit and talk to an AI (which is what most platforms do)

- They also don’t have the time to watch coaching videos that promise more articulate speech.

- They also don’t have the ability to pay for a personalized coach that can help them speak better



  - What if you combine all these 3 features into one platform that autonomously listens to your conversations so that an AI coach can assess your performance and provide instantaneous feedback

- Most platforms are focused on non-native English speakers. But there’s an untapped market for:



  - busy professionals:



    - who need to become master communnicators to excel at their job

    - who talk to A LOT of people everyday and would benefit from improving their communication skills

  - people who love socializing and want to make the best of each interaction they have with someone

- A lot of these platforms are money hungry and get a lot of negative reviews saying they kept on getting billed despite cancelling their plan. They don’t allow free trials or their free version is very limited -> we can be open-source



Many virtual meeting bots exist which help storing and remember everything that was shared during these meetings, but physical meetups are served only by expensive AI wearables that raise significant privacy concerns. We believe we can bring forth a privacy-first approach for remembering all your in-person interactions with 3 key features enhancing it:



1. **Memory Vault that grows with you:** A secure, evolving vault that stores summaries, key facts, reflections, and relational cues from your conversations. As you engage with more people, it reveals how your relationships connect and evolve, helping you understand, manage, and strengthen them over time.

2. **Contextual Reflection:** Lets users revisit past conversations and dig deeper: analyzing specific moments, identifying speech patterns to improve, and exploring what topics or insights could spark stronger future conversations.

3. **Adaptive Learning:** An autonomous communication coach that continuously learns from your reflections and interactions. It  tracks your speaking patterns, learns your strengths and blind spots, and refines its feedback to guide you toward more confident, intentional, and impactful conversations



## What it does



LinkMaxxing is an on-device speech reflection app that helps you become a more intentional, articulate, and connected communicator. It listens as you speak, with your full consent, and delivers personalized feedback on:



- Filler word usage

- Pacing and pauses

- Redundancy and repeated words

- Weak phrasing and vague language

- Context-aware rewording suggestions (vocab words suggestion)

- Sentence starters and structure

- Conciseness vs rambling

- clarity & articulation score

- confidence



But more than that, it helps you *see* how your words affect your relationships. Are you clear? Do you sound engaged? Are you deepening connection, or just filling silence?



It’s like a mirror for how you talk; so you can maxx out how you link.



## How we built it



We built LinkMaxxing with one guiding principle: everything should work **on-device**. That meant starting with Whisper-tiny for local transcription, pairing it with timestamp-aware NLP pipelines, and layering custom rule-based analysis and algorithms (filler detection, pacing, repetition) on top.



The transcription happens in realtime. So as you speak, the device analyses the conversation



For smarter insights, like better phrasing or compression, we used OpenAI’s GPT-5 with user-controlled opt-in. That gave us a hybrid architecture: fast, privacy-first feedback on-device, and deeper coaching from the cloud only when needed.



We designed a mobile-first UX that focuses on clarity: clear flags, clean summaries, and a coaching tone that builds confidence.



For every conversation you have, you can go on a phone call or chat with an AI to reflect on how it went:



- what you learned about yourself

- about them

- what you can talk about for next time

- Key things to keep in mind next time you chat with them



---



## Challenges we ran into



- **False positives in feedback:** Not every "just" or "like" is a problem. Context is king, so we had to tune aggressively.

- Graph database not accurately connecting people

- **Conciseness scoring**: It’s surprisingly hard to know when someone is being "too wordy" without just asking a model to summarize it.

- **Balancing critique with encouragement**: Feedback that feels robotic or overly critical makes users bounce. Tone matters.

- **Latency vs insight**: LLMs are slow. We had to make sure most of the experience felt instant, and reserved LLM calls for high-value moments

- Connected different cutting-edge technologies and made them work together



---



## Accomplishments that we're proud of



- We created a working speech reflection engine that leverages the latest and greatest AI tech while being mindful of people's privacy

- We built a system that grows with people

- We turned vague concepts like “speaking clearly” or “being concise” into real-time, trackable metrics.

- We created a system that respects privacy. Like a therapist that would never share personal info to anyone else



---



## What we learned



- transcribe conversations between people with diarization IN REALTIME!

- People crave deeper relationships more than performance metrics — but they’ll use metrics if it gets them there.

- Small, well-timed insights are more powerful than long reports.

- Nobody wants to be judged — they want to feel seen and supported.

- On-device AI is hard, but it's the future if you care about trust.

- Self-awareness is rare. Tools that help us listen to ourselves can change how we listen to others.



---



## What's next for LinkMaxxing



We're just getting started.



Next, we’re:



- Expanding our emotional tone detection layer — to help users see how their tone aligns with their intention

- Have CRM features to remember key info from conversations

- Building a “conversation challenge” system, lightweight prompts to help people practice linking better

- Creating a private memory system that tracks your growth, not just per session, but across time

- Exploring integrations with IRL communities (student clubs, events, teams) to help people reflect *together*, not just alone

- Maybe designing a wearable mode for ambient conversation reflection (without screens)



We believe LinkMaxxing can become a core part of how people grow, not just as communicators, but as humans who want to connect more deeply.



Built with ❤️ using React Router v7, Convex, Clerk, Polar.sh, ZepCloud, Speechmatics, Vapi, GraphDB, VectorDB, and OpenAI