Sources & lineage

familiar didn’t spring from nothing — it’s the newest branch of a family tree of open projects, and this page is the map back to its ancestors. (For the complete license-level credits of everything vendored or linked, see the repository README.)

The StackChan family — the body’s ancestors

stack-chan/stack-chan — the original. Shinya Ishikawa’s “JavaScript-driven M5Stack-embedded super-kawaii robot” started the whole palm-sized-companion genre; every StackChan descends from this design and its open hardware.
m5stack/StackChan — M5Stack’s official open-source StackChan: the hardware poppet runs on and the base application poppet/main/ is built from.
BrettKinny/StackChan — the fork whose firmware work (the dotty / firmware branches) poppet directly derives from: the HAL, avatar, servo, and personality layers grew from here.

The XiaoZhi family — the mind’s ancestors

78/xiaozhi-esp32 — the MCP-based ESP32 voice-assistant firmware. poppet’s conversational core is a pruned in-tree fork of it (v2.2.4, at poppet/vendor/stackchan-esp32/) with all cloud endpoints removed.
howecheung/StackChan-XiaoZhi — a kindred ancestor: StackChan body, xiaozhi mind, with a focus on companionship (touch, motion, mood lights, servos). Itself derived from mo-hantang/Stackchan-HtSz. familiar walks the same StackChan-marries-xiaozhi path.
xinnan-tech/xiaozhi-esp32-server — the self-hosted backend for xiaozhi devices. grimoire replaces it for this project: same seat at the table, new wire protocol, LAN-only by design.

The engines

The heavy lifting inside grimoire is other people’s brilliant work:

whisper.cpp — hearing (ASR), run in-process.
Kokoro via Kokoro-FastAPI — the voice (TTS).
Any OpenAI-compatible LLM server — the thinking; we run ours through llama-swap.
microWakeWord (Open Home Foundation) — the wake-word architecture behind “Hey Artemis”, with ESPMicroSpeechFeatures as the audio frontend.

And this site

The docs’ ink-and-parchment look riffs on the ensō documentation theme, steered a few shades witchier.