StackChan experiments

Public notes and tooling for experiments with the M5Stack StackChan desktop robot.

Each experiment lives in its own folder under experiments/. Nothing here should contain Wi‑Fi passwords, API keys, or other secrets. Secrets stay in gitignored local/ dirs on your machine (or Doppler).

Experiments

#	Folder	Status	Summary
01	experiments/01-ai-stackchan-ex-realtime	Done (poor UX)	AI_StackChan_Ex WS+PCM — half-duplex, no AEC; lessons only
02	experiments/02-minimal-realtime-aec	Research → build	Continuous mic + AEC + OpenAI Realtime via Espressif WebRTC openai_demo + CoreS3 board support

Hardware baseline

• Device: Official M5Stack StackChan (CoreS3 + serial servos)
• Cable: USB‑C data cable, Mac ↔ CoreS3 (for flash + serial only)
• Not ESPHome: these firmwares are mostly PlatformIO / Arduino (or other stacks per experiment)

Repo conventions

experiments/
  NN-short-name/
    README.md          # human walkthrough for this experiment
    docs/              # deeper notes, decisions, gotchas
    config/            # example YAML / env templates (no secrets)
    patches/           # diffs against upstream, if any
    scripts/           # clone / apply / flash helpers
    local/             # gitignored: real secrets, private notes
    upstream/          # gitignored: cloned upstream sources + build tree

Safety

• Do not commit files under local/ or real SC_SecConfig.yaml values.
• Prefer Doppler (or similar) for API keys; inject at flash time via scripts.
• Public repo: assume every committed byte is world-readable.