Hacktoberfest Weekend Challenge: Build for a Friend

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

Demo

Code

Chirantan2110 / lingobuddy

💬 LingoBuddy: Local-First AI Language Companion & Interactive Tutor

LingoBuddy is a submission-ready, offline-first AI conversational language tutor. Powered by local open-weight LLMs via Ollama, local speech recognition, and structured SQLite memory, LingoBuddy provides realistic dialogues, pedagogical error detection, automated recurring mistake tracking, and tailored next-session quizzes without sending private audio or data to commercial third-party cloud APIs.

---

🌟 Key Features

• 🦙 100% Local Open-Weight LLMs (Ollama): Seamlessly connects to llama3.2, mistral, gemma2, or llama3. Includes an intelligent built-in simulation fallback so the app runs instantly even before pulling large models.
• 🎙️ Voice & Speech-to-Text: Real-time microphone input with Web Speech API and backend faster-whisper / speech_recognition upload endpoints.
• 🔊 Native Pronunciation TTS: Built-in audio playback using native speech synthesis for authentic target language pronunciation.
• 🎯 3 Pedagogical Correction Modes
  • Gentle: Emphasizes flow and confidence; only corrects severe blockers.
  • Balanced…

View on GitHub

How I Built It

I use antigravity and devrelay

Why Does Open Innovation Matter?

LingoBuddy is built around open AI because language practice can be personal and sometimes embarrassing. A learner should not have to send their private conversations and voice recordings to a third-party AI service just to practice speaking.

My Agent Session

I used an AI coding agent throughout the development of LingoBuddy to design the architecture, implement the local AI integration, build the conversation and correction logic, and iterate on the user experience.

Prize Categories