A voice-interactive chatbot for Japanese language learning and translation, powered by a locally-hosted LLM via Ollama. All inference runs on-device — no external API calls, no data sent to third-party services.

Built around an Ollama-backed inference engine (phi3.5:latest), the assistant supports two operational modes — Translation and Conversational Chat — both exposed through a voice I/O pipeline using SpeechRecognition + pyttsx3. A lightweight RAG layer performs keyword-matched retrieval against a local Markdown knowledge base to ground factual responses about Japanese language and culture.

Demo

Project Screenshots

‹›

Output 1

Technology Stack

Python

Ollama