How nishuzumi/gemini-teacher works
Project type and purpose
• This is a Python-based command-line “English Speaking Assistant” that:
• Captures microphone audio
• Sends it to Google Gemini AI for speech recognition & evaluation
• Provides grammar/ pronunciation feedback
• Optionally uses ElevenLabs for text-to-speech output
• It’s neither a web app nor microservice cluster, but a single-process CLI tool that integrates with external AI services.
File structure
• Top-level files:
• .env.example / environment config
• requirements.txt / Python dependencies
• starter.py / the single entry-point and core logic
• No nested directories; all logic lives in or is imported by starter.py.
• External configuration (API keys) is managed via .env.
README for additional insights
• Describes real-time audio capture, AI-driven pronunciation assessment, grammar correction, scenario dialogues, and optional TTS.
• Lists dependencies: Python 3.11+, PortAudio/PyAudio, Gemini API, ElevenLabs API.
• Usage flow: record → send → receive feedback → display/improve.
Detail
• Modules, even if implemented in one file
• Environment config as a separate component feeding API clients
• Continuous feedback loop
• Do not forget optional paths like speech synthesis