A personal collection of AI projects and experiments — agents, RAG, multi-agent systems, prompt engineering, tools/evals, text-to-speech, voice assistants, and more.
Each subfolder is a standalone project with its own dependencies and virtual environment (no shared root-level environment). uv is the standard tool used across all projects for dependency management and running code.
From inside any project folder:
uv run <entrypoint>.pyor, for a Streamlit app:
uv run streamlit run app.pyuv automatically creates the project's .venv and installs dependencies from pyproject.toml
on first run — no manual pip install needed.
See each project's own README.md for what it does, prerequisites, and exact run instructions.
- clone-voice — voice cloning / text-to-speech using Coqui TTS (xtts_v2)
- voice-notes — record a voice note, transcribe it, and clean it up into a formatted note using a LangChain agent (OpenAI, or a local model if no API key is set)
- jev-dino — exploratory script: have TypeSafe AI's Jev model play the Chrome dino-runner game, run via OpenRouter
- documentation-helper — RAG over a documentation site: crawl with Tavily, embed with OpenAI, store in Pinecone, and answer questions with cited sources