Skip to main content

SerenGanglion

Seren's reflexes on the M5Stack LLM module. A ganglion is a cluster of nerve cells outside the brain that runs reflexes without asking it, and a wake word is a reflex. Sibling to Sinew, Meninges, Soma and Psycheion.

Status, 20 Nov 2026: design, plus the StackFlow wire. seren_ganglion.wire speaks the framing M5's units use between themselves, and sys_client registers with sys; both are tested against a fake sys. Nothing has run on a module yet.

Where it runs

The LLM module (AX630C) is its own computer inside the StackChan: Ubuntu 22.04 (arm64), kernel 4.19, about 1 GB of RAM, 32 GB of eMMC, M5's StackFlow units, and no network of its own. It talks to the CoreS3 over one UART (JSON per line, up to 1.5 Mbaud). What's known about the module, and how it was found, is in SerenSoma's firmware/docs/appendages.md under "The LLM module as ears".

Chad, 20 Nov: yes. A ganglion is a cluster of nerve cells outside the brain that runs reflexes without asking it, and a wake word is a reflex. It sits beside Sinew, Meninges, Soma and Psycheion.

A package, not a distro. Forking M5's image means its Windows-only reflash with the module out of the stack for every mistake, and owning a kernel 4.19 BSP with closed NPU blobs through every M5 update. Everything we want lives above the kernel. So it is one .deb, seren-ganglion, installed beside M5's StackFlow packages, and apt treats it like theirs.

It is a StackFlow unit (work_id: "ganglion"), so it speaks StackFlow's JSON-per-line to the core like every other unit, and the core's locked llm route needs only one more name on its list. bashexec, push, upgrade and rmmode stay off the robot's API, as now.

What it does:

  • Listening, wired inside the module. Wake word, VAD and Whisper chained on the module; the core gets one finished event (heard, with its text and timing). Only text crosses the UART.
  • Face sounds. Plays the MP3s the core forwards (Psycheion's forward-on-play; audio.play takes wav, mp3 or pcm, not Opus), about 64 KB a line at most.
  • The three LEDs, its health, its versions.
  • Its own upkeep: the update Chad's module needed on 19 Nov as a script anyone can run, telnet on port 23 switched off at install, and a refusal to run on a StackFlow older than the one tested.

Its shapes live in Sinew beside the robot's, with the same contract test, so the two ends can't drift. Sinew can run on the module: its floor is Python 3.10 and Ubuntu 22.04's is 3.10, so that floor becomes a promise (a line in Sinew's README). Two things first:

  • Check the module's Python really is 3.10 (one command over adb).
  • import seren_sinew must need only the standard library. Today the package's front door pulls in starlette for the request-logging middleware (Psycheion, 20 Nov); load that only when asked, and the ganglion carries one small pure-Python wheel.

The mic light is the core's, never the ganglion's

Psycheion, 20 Nov; Chad: "yes yes yes". A unit that crashes or hangs could leave a mic open and never say so, which is the hidden microphone again. So:

  1. The ganglion opens a mic only when the core tells it to. Never on its own, and never by itself after a restart.
  2. The core doesn't take its word for it. Every few seconds it asks sys.lstask; if anything that captures audio is running (ours, or M5's audio, kws, vad, asr, whisper), the light is on.
  3. No answer counts as listening. A module that stops replying keeps the light on until it answers.

The ganglion's own "mic open" event is still wanted, for the face's listening look, because it's fast. It never decides the light.

Connections: however the person brings it

Chad, 20 Nov: the module has a microSD slot and a USB-C port, and the Ethernet board "doesn't really go on for the real stuff". So nothing assumes a network.

  • Talking to the robot is always the UART through the core. It's the one link every StackChan with a module has. The ganglion listens on no network port.

  • Getting software and models on takes any of:

    Way How
    microSD A folder of .debs and models on the card; the installer finds it
    USB-C adb push from a computer, or the installer run from that computer
    Ethernet, if someone has it apt from M5's repo as usual
    The UART Small things only; 1 MB is about 10 s at 1.5 Mbaud, base64 included

    Whichever way, the installer checks the same checksums before it installs anything.

Open

  • Whether a StackFlow unit can be a Python process or has to be C++ on their framework. StackFlow's own units are C++; how ours registers with sys decides what the package is written in.
  • Whether the microSD slot mounts in its Linux at all, and where.
  • Whether its USB-C can be a network to a computer (a USB gadget), separate from adb. Untested; the 19 Nov adb reverse was refused, and that's a different mechanism.

Getting it on

What a person installs is one .deb, and the module needs no pip, no venv and no network.

  • CI builds it on each tag and attaches it to the GitHub release: seren-ganglion_<version>_arm64.deb.
  • It carries everything inside, unpacked under /opt/seren/ganglion/:
    • seren_ganglion itself;
    • Sinew (pure Python, imported with the standard library alone; see tests/test_sinew_on_the_module.py);
    • pyzmq built for the module, with ZeroMQ's library bundled inside it.
  • It always ships its own pyzmq, whatever the module already has. That's 673 KB against 25 GB free, and in exchange every module runs the same ZeroMQ, whether M5's image has a copy Python can use or not.
  • A systemd unit, seren-ganglion.service, starts it beside M5's units, with its state in /var/lib/seren-ganglion. Installing it also switches off the telnet server on port 23.

Getting the .deb onto the module is however the person brings it (microSD, adb push, Ethernet), then dpkg -i seren-ganglion_<version>_arm64.deb.

The wheels by hand, for working on it. Ganglion and Sinew are pure Python, so a desktop's pip fetches files that work on the module as they are. pyzmq isn't: fetch it for the module's platform (Psycheion, 20 Nov; checked the same night, it gives one file, pyzmq-26.4.0-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl):

pip download seren-ganglion seren-sinew --no-deps -d ganglion-wheels
pip download pyzmq --only-binary=:all: --platform manylinux2014_aarch64 --python-version 3.10 --no-deps -d ganglion-wheels

--no-deps because Sinew's declared dependencies (starlette and its family) are never imported here. On the module, a wheel is a zip: python3 -m zipfile -e <wheel> /opt/seren/ganglion/lib needs nothing that isn't already there.

Still to read off the module (one adb shell, read-only), before the .deb is written:

Check Why
python3 --version The 3.10 floor
ldd --version The pyzmq wheel needs glibc 2.17 or newer; Ubuntu 22.04 has 2.35
ls -la /tmp/rpc.* /tmp/llm/ StackFlow's sockets match the code we read, and who may connect: connecting to a unix socket needs write permission on its file, so read the w bits for group and other, not just the names
systemctl list-units 'llm*' How M5 starts its units, so ours starts the same way
lsblk; mount | grep -i mmcblk1 Whether the microSD slot appears, and where

Who it runs as

Psycheion, 21 Nov: the ganglion parses whatever the core sends, so a parsing bug as root is a root bug on the box we locked bashexec away from. M5's units run as root, and their sockets may admit only root.

  • If the sockets let a group write, the .deb makes a ganglion system user in that group, and the unit runs as it.

  • If they're root-only, it runs as root, boxed by systemd, and this section says so until it's fixed:

    NoNewPrivileges=yes
    CapabilityBoundingSet=
    ProtectSystem=strict
    ReadWritePaths=/var/lib/seren-ganglion /tmp
    ProtectHome=yes
    PrivateNetwork=yes
    RestrictAddressFamilies=AF_UNIX
    

    No capabilities, a read-only system apart from its own state and the sockets, and no network at all: path-based unix sockets still work across a private network namespace, so StackFlow is reachable and nothing else is. PrivateTmp stays off, because the sockets live in /tmp. These need checking against the module's systemd before they are trusted.

  • We don't change M5's units to loosen their sockets: a drop-in that changes how their root processes create files would reach well past ours.

Layout

Where What
SerenGanglion/seren_ganglion/wire.py StackFlow's framing: pairs, register_unit's answer, unit URLs. Standard library only
SerenGanglion/seren_ganglion/sys_client.py RPC to sys or any unit over ZeroMQ (pyzmq, imported only here)
Sinew, seren_sinew.ganglion (to come) The shapes the ganglion and the core exchange, contract-tested like the robot's
SerenSoma The core's half: forwarding to ganglion, the lstask check behind the mic light, the allow list

To come here: the unit itself (serve setup, work, exit), the .deb (debian/), and the installer that takes microSD, USB-C, Ethernet or the UART.

Tests

cd SerenGanglion
pip install -e ".[test]"   # the lean shape: as close to the module as a runner gets
python -m pytest

[dev] adds pyzmq for the socket tests, which run against a fake sys on tcp because Windows has no ipc sockets. Python 3.10 is the floor because it's the module's.

Metadata

Release files for seren-ganglion 0.2.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for seren-ganglion 0.2.0
File Size Uploaded Provenance Actions
seren_ganglion-0.2.0.tar.gz 15.4 kB With provenance Details for seren_ganglion-0.2.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for seren-ganglion 0.2.0
File Interpreter ABI Platform Actions
seren_ganglion-0.2.0-py3-none-any.whl Python 3 none any Details for seren_ganglion-0.2.0-py3-none-any.whl

Total release size: 24.9 kB

Release files / seren_ganglion-0.2.0.tar.gz

Download URL seren_ganglion-0.2.0.tar.gz
Size 15.4 kB
Tags Source
SHA-256 checksum
How to use checksums
abc07f1ec0d449fd9853acae4a19f2d1930aa46898601db64977c12db08a7aee
BLAKE2b-256 checksum
How to use checksums
69a75d519c569f153783bf156ddfe4a4be91b7d73b7be779c2dc70a1c1f8b008
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 21, 2026.

Transparency log

Release files / seren_ganglion-0.2.0-py3-none-any.whl

Download URL seren_ganglion-0.2.0-py3-none-any.whl
Size 9.4 kB
Tags Python 3
SHA-256 checksum
How to use checksums
5f1bb2579b495bee6ed83497aa15c375566ad9c5ba16e578e06e230e03b73702
BLAKE2b-256 checksum
How to use checksums
c8561985915059d32d9793b078fd894d5fbb6c57e6f8a2f484f3c7a913a33005
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 21, 2026.

Transparency log

Security for seren-ganglion 0.2.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for seren_ganglion-0.2.0.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 21, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page