SigMonD (Signal Monitor Daemon) is the installer, lifecycle manager, and coordinator for the HamSCI SDR observation suite.
Sigmond manages a family of independent clients that share the DASI2 station hardware. DASI2 is the NSF-funded Distributed Array of Small Instruments, Track 2 project — a station is a ka9q-radio receiver on an RX888 SDR (GPSDO-disciplined, TS-1 time-injected) plus a magnetometer. Each client records a different signal type — WSPR, FT8/FT4, HF time standards, magnetic field, CODAR, HFDL, beacon TEC, meteor scatter — and sigmond handles the parts that require coordination between them: installation, service lifecycle, log aggregation, diagnostics, timing-authority distribution, the shared SQLite sink, and resource (CPU) arbitration.
Two ways, both current:
- A dedicated mini-PC, from the USB image. The image turns the machine into a Proxmox host with the station in a VM; you answer a few questions and it does the rest. Follow sigmond-appliance INSTALL.md.
- A Linux machine you already run (bare metal). Sigmond installs from these repositories onto your own Debian. Follow docs/install-bare-metal.md.
Who are you? Station host → Operator guide · Scientist wanting a signal recorded → Scientist guide · Developer → CONTRIBUTING.md. Full map: docs/README.md.
The DASI2 hardware kit (DXE antenna → RX888 SDR, disciplined by an
LB-GPSDO and time-injected by a TS-1, plus a magnetometer) feeds
ka9q-radio (radiod), which multicasts per-channel RTP streams.
Every RF client subscribes via ka9q-python;
mag-recorder is the one client that reads its sensor directly (USB-I²C)
rather than through radiod.
- Core clients:
hf-timestd(timing authority — publishes the RTP↔UTC offset and tier that the other RF clients label against),wspr-recorder,psk-recorder,mag-recorder,meteor-scatter,hamsci-physics,hfdl-recorder,codar-sounder. - Additional clients:
hf-tec,superdarn-sounder.
Decoded spots and products land in a shared SQLite sink
(/var/lib/sigmond/sink.db), and hs-uploader ships them upstream. WSPR,
FT4/FT8 and MSK144 spots go to wsprdaemon.org, which forwards copies to
wsprnet.org and pskreporter.info; GRAPE and magnetometer products go to
PSWS / HamSCI.
hfdl-recorder and codar-sounder send their timing rows to wsprdaemon.org
the same way. hfdl-recorder can also feed airframes.io directly through
dumphfdl; that feed stays off until the operator turns it on. Sigmond is the
oversight layer wrapping the whole software stack: install, configure,
lifecycle, CPU-pinning, and the shared sink.
The diagram source is
docs/architecture.svg(editable vector, drops straight into slides); a raster copy is atdocs/architecture.png.
Hardware:
- An HF antenna
- An RX-888 (or other SDR supported by ka9q-radio)
- A GPS disciplined oscillator (e.g. Leo Bodnar GPSDO) providing 10 MHz + PPS
- A Linux computer (Debian 12+ or Ubuntu 22.04+)
Full list with models, what is optional, and what B4 runs: docs/hardware/shopping-list.md
Software prerequisites:
- Python 3.11 or later
- git
- systemd
- chrony (for time-standard work)
These are the bare-metal steps with their details; docs/install-bare-metal.md puts them in order with what each does to your machine. For the USB image, see Install a station.
Sigmond installs at /opt/git/sigmond/sigmond/, peer to its managed
components. Clone and install from there:
sudo mkdir -p /opt/git/sigmond
sudo chown $USER /opt/git/sigmond
git clone https://github.com/HamSCI/sigmond /opt/git/sigmond/sigmond
cd /opt/git/sigmond/sigmond
./install.shinstall.sh creates a sigmond system user that owns
/opt/git/sigmond/*, adds you to the sigmond group so you can edit
sources as yourself, sets up /opt/git/sigmond/sigmond/venv, and symlinks smd to
/usr/local/bin/smd. Open a new shell after install (or
newgrp sigmond) to pick up the group membership.
Running under Proxmox?
install.shdetects KVM guests and offers to configure the Proxmox host (PCIe passthrough, CPU isolation, vfio, hookscript) before installing sigmond — one prompt, one host reboot, automatic resume. Seedocs/installation-guide.md"Running under Proxmox VE? Read this first."
One command installs, configures and starts a whole station on this machine. It works the same on bare metal and inside a VM:
smd bringup dasi2 --callsign AC0G --grid EM38ww --reporter AC0G/S \
--psws-station-id S000418 \
--psws-instrument hf-timestd=367 \
--psws-instrument mag-recorder=401Use your own callsign, grid square, reporter ID and PSWS ids. The PSWS flags
matter only if you upload to PSWS: one station ID for the site, and one
instrument ID per product under it (the RX888 for hf-timestd, the
magnetometer for mag-recorder). Bring-up writes them to
/etc/sigmond/site-profile.toml and into each recorder's config;
smd config psws sets them later.
dasi2, the default profile, gives you a local radiod, hf-timestd (GRAPE),
WSPR, FT4/FT8, meteor scatter, HFDL and CODAR timing, and the magnetometer. base stops at radiod
and timing; client decodes from radiods on other machines; see "Decode-only host"
below. Bring-up skips, with a warning, anything
whose hardware it cannot find on the USB bus.
Bring-up also sets up remote access when it runs on bare metal. It runs
sigmond-rac activate, which registers the machine with the gateway and
starts its tunnel. A Proxmox appliance VM skips this step, because its host
carries the machine's remote access. If registration fails, bring-up still
finishes, and a retry timer keeps trying.
The machine's remote-access name puts the callsign and the hostname together
in upper case, as in AC0G-SIGMA or AC0G-B2. A hostname that already starts
with <CALL>- stands alone. Once sigmond-rac activate records the name,
the name sticks. A hostname of localhost, or one equal to the callsign,
gives no default; pass --site.
smd remote status shows whether remote access is on, and smd remote on or
smd remote off switches it. Type these as yourself, never with sudo. The
switch changes system files, so smd asks for root through sudo on its own and
may ask for your password.
A machine that runs only radiod and ka9q-web never runs bring-up. On such a machine, run this once:
sudo /opt/git/sigmond/sigmond/scripts/rac/sigmond-rac activate --site AC0G-B1The design explains the channels, the ports and the exit codes.
Add --dry-run to see every step first. Add --non-interactive to accept
radiod's defaults instead of answering its questions.
If the station uploads to PSWS, put each instrument's ID in
/etc/sigmond/site-profile.toml and render it into the client configs:
[psws.instruments]
"hf-timestd" = "<GRAPE instrument id>"
"mag-recorder" = "<magnetometer instrument id>"smd config renderA decode-only host runs no radiod of its own. It decodes from radiods on other machines on the LAN. Name each radiod it should use; the option may repeat:
smd bringup client --callsign AC0G --grid EM38ww \
--remote-radiod ac0g-b1-a-status.local --remote-radiod ac0g-b1-b-status.localBring-up browses the LAN for radiods and prints every one it finds, with
its host and advertised name. An unattended install (--non-interactive,
or no terminal) binds only the radiods you name, and with none named it
stops after printing the list, so you can choose from it. A named radiod
that does not answer stops bring-up before it installs anything.
On a terminal with no radiods named, bring-up proposes the ones whose advertised name starts with your callsign, so a radiod that advertises "AC0G @EM38ww B1-a longwire" belongs to AC0G. The first word must equal the callsign; a radiod that advertises a reporter ID such as "AC0G/S RX-888 Mk2" never gets proposed. Check the proposal before you accept it: every radiod on your LAN that advertises your callsign gets proposed, whichever machine it runs on. Bring-up prints one line for each radiod it would keep, with the radiod's host, its advertised name and the reason, then asks before it binds them.
Each kept radiod gets one instance of each per-reporter recorder the
profile carries. The client profile carries two: wspr-recorder (WSPR)
and psk-recorder (FT4/FT8). It also installs hfdl-recorder and
codar-sounder, which run one instance per radiod named in their own config
files; smd start starts one for each. Bring-up names the instances after the radiod:
ac0g-b1-a-status.local gives the reporter ID AC0G/B1A. With a single
radiod, --reporter still sets the name. With several, --reporter
counts as an error, because one name cannot cover two instances.
hf-timestd takes its timing from one radiod. With one kept radiod it uses
that one. With several, name it with --timing-radiod <status name>.
Without that, bring-up installs hf-timestd, leaves it idle and says so.
Bring-up records each radiod with smd admin sources bind, and you can run
it yourself:
smd admin sources bind ac0g-b1-a-status.local --reporter AC0G/B1A [--timing]The verb writes the radiod into /etc/sigmond/coordination.toml. For each
recorder enabled here (or each one named with --clients), it makes the
instance, selects the radiod as its source and points the instance's config
at it. --timing binds hf-timestd to the same radiod. Running it twice
changes nothing.
No verb renames an instance yet. To give one a new name, bind its radiod again under the new reporter ID, start the new instance, then retire the old one:
smd admin sources bind ac0g-b1-a-status.local --reporter AC0G/LOOP \
--clients psk-recorder
smd admin instance enable psk-recorder AC0G/LOOP
smd admin instance disable psk-recorder AC0G/B1A
smd admin instance remove psk-recorder AC0G/B1AThe old name's entry stays in coordination.toml until you delete it by hand.
The [[receiver]] list in /etc/sigmond/site-profile.toml is the one record
of which radiod each reporter ID listens to. Bring-up, smd check and
smd admin instance list all read it.
[[receiver]]
reporter_id = "AC0G/B1A"
radiod = "ac0g-b1-a-status.local"
injector = true
antenna = "20 m longwire"
[[receiver]]
reporter_id = "AC0G/B1B"
radiod = "ac0g-b1-b-status.local"Read the list back, then apply it:
smd admin receiver list
smd admin receiver applyapply binds every declared receiver. It adds and updates. It never
removes an instance, a sources file or recorded data. It sets an instance's
upload flag only when the receiver decides it, through upload = or
same_antenna_as, and only once the client has a config. Otherwise the flag
stays as smd config upload left it.
A receiver with no clients key runs one instance of each per-reporter
recorder (wspr-recorder, psk-recorder, meteor-scatter) that this host enables
or has installed. clients = ["psk-recorder"] narrows it to the ones you
name. apply --clients a,b sets that default set for one run; bring-up
passes its profile's recorders this way.
Two commands write the list for you. smd admin sources bind <status> --reporter <ID> records the receiver, then binds it. To add one receiver
for one client, run:
smd admin instance add <client> <reporter> --radiod <status>Several receivers may share one radiod. Each receiver after the first on a
radiod must name same_antenna_as, the reporter ID of a receiver on that
radiod. Such a receiver keeps its spots local unless you set
upload = true. At most one receiver per radiod carries injector = true,
and at most one on the whole host, because hf-timestd binds to that
receiver's radiod.
With a list declared, bring-up plans from it. It needs no --remote-radiod
or --reporter, because the list already names every radiod and reporter ID.
Bring-up accepts flags that match the list, so you can re-run the same
command after a failure. It refuses flags that disagree with the list; edit
the list instead.
Several receivers put several instances of each recorder on one host. Before
it installs anything, bring-up adds up their cost against the CPUs left over
after radiod. When the host falls short, bring-up stops and says by how much.
Narrow a receiver's clients = [...], or pass --ignore-capacity. Under
--dry-run the shortfall prints as a warning and the plan still shows.
On a station that already runs instances, smd config render --init builds
the list from them. It reads each instance's sources file and makes one
receiver per reporter ID. It prints what it wrote, and it names every
instance it left out and why: an empty or unreadable sources file, a key that
is not a radiod, a client that takes no per-reporter instance, or a reporter
whose instances point at two radiods. If the adopted
list fails validation, it writes nothing and says what to fix. A profile
that already declares receivers stays as it is. --dry-run prints the
result and writes nothing.
The steps below build the same station a piece at a time, for when you want to choose each part yourself.
smd component list --catalog━━━ catalog — known clients ━━━
Servers (2)
✓ ka9q-radio ka9q-radio SDR daemon — receives and multicasts RF channels
✓ ka9q-web Web interface for ka9q-radio (radiod status UI)
Clients (8)
✓ codar-sounder Opportunistic ionospheric sounder using CODAR chirp transmissions contract=0.8
✓ hf-tec HF PRN-coded beacon recorder for ionospheric specification contract=0.8
✓ hf-timestd HF time-standard analyzer (WWV/WWVH/CHU/BPM) contract=0.8
✓ hfdl-recorder HFDL (High Frequency Data Link) recorder — one dumphfdl subprocess per band contract=0.8
✓ mag-recorder RM3100 magnetometer recorder + PSWS uploader contract=0.8
✓ meteor-scatter Meteor-scatter ping recorder/decoder (jt9 --msk144) contract=0.7
✓ psk-recorder FT4/FT8 spot recorder for PSKReporter contract=0.8
✓ wspr-recorder WSPR/FST4W audio recorder (period-aligned WAVs) contract=0.8
Infra (3)
✓ gpsdo-monitor Leo Bodnar GPSDO health monitor + mDNS advertiser
✓ igmp-querier Robust IGMP querier for radiod multicast on home switches
✓ sigmond-rac Remote Access Channel (frpc reverse tunnel)
Setting up a new host end to end? See the greenfield install runbook for the full ordered walkthrough (identity → radiod → wisdom → tuning → clients → validate).
Install whichever clients you need. Each client's installer handles everything: creates a service user, Python venv, config template, and systemd units.
smd install hf-timestd
smd install psk-recorderRun smd as yourself, never under sudo. It asks for root itself when a
step needs it, and it refuses to run under sudo.
Use --dry-run to preview what would happen without making changes:
smd install wspr-recorder --dry-runEach client owns its own config file. After installation, edit the config for your station:
| Client | Config file |
|---|---|
| radiod | /etc/radio/*.conf |
| hf-timestd | /etc/hf-timestd/timestd-config.toml |
| psk-recorder | /etc/psk-recorder/psk-recorder-config.toml |
| wspr-recorder | /etc/wspr-recorder/wspr-recorder.toml |
| mag-recorder | /etc/mag-recorder/mag-recorder-config.toml |
| codar-sounder | /etc/codar-sounder/codar-sounder-config.toml |
| hfdl-recorder | /etc/hfdl-recorder/hfdl-recorder-config.toml |
| hf-tec | /etc/hf-tec/hf-tec-config.toml |
| meteor-scatter | /etc/meteor-scatter/meteor-scatter-config.toml |
Sigmond's own coordination config lives at /etc/sigmond/topology.toml.
Copy the example to get started:
sudo cp /opt/git/sigmond/sigmond/etc/topology.example.toml /etc/sigmond/topology.toml
sudo vi /etc/sigmond/topology.tomlEnable or disable components by setting enabled = true or false.
Instead of editing config files manually, you can use the interactive TUI:
smd tuiThe TUI provides:
- Install screen — browse and install components from the catalog
- Topology screen — enable/disable components with live validation
- Overview screen — system health dashboard
- Logs screen — view and filter service logs
- CPU affinity screen — configure CPU core isolation
- CPU frequency screen — monitor and control CPU frequencies
- Environment screen — discover and probe network peers (KIWISDRs, GPSDOs)
- GPSDO screen — monitor Leo Bodnar GPSDO health
- Validate screen — cross-client harmonization checks
- Lifecycle screen — start/stop/restart services
- Apply screen — reconcile services with current config
- List (Software versions) screen — per-component status (git ref, upstream divergence, version policy) with Update All / per-component update buttons; replaces the old separate Update screen
- Backup/Restore screens — backup and restore configuration
Keybindings: t = topology, i = install, v = validate, q = quit
smd startThis starts all enabled components in the topology. To start a single component (the component name is now a positional argument; an optional second positional selects a single instance):
smd start psk-recordersmd statusShows systemd unit state for every managed service, plus inventory data from each installed client: version, channel count, active frequencies, and any issues.
smd's verbs are grouped the same way smd --help and the TUI present
them. Run smd <verb> --help (or smd admin <verb> --help) for the full
options on any verb.
Lifecycle — operate on running services. Without a component name they act on all enabled components; a positional component (and optional instance) narrows the scope.
| Command | Description |
|---|---|
smd start [<component> [<instance>]] |
Start managed services |
smd stop [<component> [<instance>]] |
Stop managed services |
smd restart [<component> [<instance>]] |
Restart with reset-failed |
smd reload [--via=auto|systemd|socket] [<component>] |
Reload via SIGHUP or restart |
smd status [<component>] |
Service health + client inventory |
smd enable <name>... |
Set enabled = true in topology |
smd disable <name>... |
Set enabled = false and stop the units |
smd apply |
Reconcile running services with current config |
Install & components
| Command | Description |
|---|---|
smd install [<client>] |
Install + configure a client, or run the full-suite install |
smd bringup |
Guided station bring-up from a catalog profile (installs dormant when hardware is absent) |
smd adopt <name> |
Adopt detected hardware or a discovered source — the only verb that starts anything; asks first (--dry-run, --yes) |
smd component list [--catalog] |
Per-component status (lifecycle + git ref + version policy); --catalog lists what's installable |
smd component update [<name>] |
Pull latest and reapply per topology version policy (root) |
smd component add|remove <name> |
Clone/remove a component repo and its topology entry |
smd config show|migrate |
Inspect or migrate coordination config |
smd config init|edit <client> |
Run a client's first-run wizard or edit flow |
smd update [--apply] [--latest] |
Plan an update (dry run by default); --apply performs it. --latest takes every component to the tip of its main branch and pins the station there; it serves test stations |
smd version |
Each component's commit, against the image and upstream |
smd sink status|upload|fill|off |
The site sink switch: whether this station sends its data |
smd config reporters [--uploader-call NAME] |
The name this machine uploads under |
smd remote status|on|off |
Remote access for the fleet admin |
Observe & diagnose — most diagnostics now live under the admin group.
| Command | Description |
|---|---|
smd watch <target> |
Live tail for wspr, psk, hfdl, codar, hf-tec, mag, ka9q, radiod, uploads, verifier |
smd admin log [<client>] [--files] [--level L] |
Follow journal (or --files for file logs); --level sets the client's level via coordination.env + SIGHUP |
smd admin diag [net|cpu-affinity|cpu-freq] |
Network / CPU diagnostics; cpu-freq --radiod-max-mhz N --other-max-mhz N --save keeps the caps across reboots |
smd admin validate |
Cross-client harmonization rules |
smd admin environment list|probe|describe |
Situational awareness of network peers (KIWISDRs, GPSDOs, NTP) |
smd admin verifier|storage|sources|timing|radiod|instance|rac … |
Specialized maintenance — see smd admin --help |
smd tui |
Launch the interactive TUI configurator |
CLI v2 note. The verb surface was reorganized into
component/admin/configgroups. The old flat verbssmd list,smd log,smd diag,smd validate, andsmd environmentno longer exist — usesmd component listandsmd admin {log,diag,validate,environment}instead.smd installandsmd applyremain available at the top level. Lifecycle scope is now positional (smd start psk-recorder); the old--components X,Yflag still works but is deprecated. Full mapping in docs/CLI-V2-SPEC.md.
smd statusFor each installed client, status shows the systemd unit state plus data from the client's contract inventory: version, git commit, number of channels, active modes, and any reported issues.
Follow the systemd journal for a client:
smd admin log psk-recorderTail the client's file logs (spot logs, decode output):
smd admin log psk-recorder --filesChange a client's verbosity without restarting:
smd admin log psk-recorder --level DEBUGThis writes PSK_RECORDER_LOG_LEVEL=DEBUG to /etc/sigmond/coordination.env
and sends SIGHUP to the client's systemd units. The client re-reads the
environment and adjusts its logger.
To set a default level for all clients (no SIGHUP — restart or SIGHUP each client to apply):
smd admin log --level WARNINGsmd admin diagChecks:
- Network reachability (wsprnet.org, wsprdaemon services)
- Dependency versions against pinned commits
- Per-client self-validation (
<client> validate --json) - Service health
smd admin diag net # unprivileged; uses /proc/net/igmp state
smd admin diag net # adds passive raw-socket listen for queriersClassifies your network for multi-host radiod safety and tells you
whether to stay host-local (ttl=0), enable a querier, or avoid a
Wi-Fi path. See docs/networking.md for the full
guide.
smd admin validateRuns harmonization rules across all enabled clients: CPU core isolation, frequency coverage vs. radiod sample rate, radiod reference resolution, and timing chain verification.
The additional clients (hf-tec, superdarn-sounder) are installed at your discretion. Installing a client enables it — there is no separate enable step:
# See what's available
smd component list --catalog
# Install it — this also enables it in topology
smd install hf-tec
# Configure it for your station (first-run wizard, or $EDITOR fallback)
smd config init hf-tec
# Start it
smd start hf-tec
# Verify
smd statusThe forward path is just install → configure → start — there is no
separate enable step. smd install enables the component, and naming a
component to smd start enables it too if it was installed-but-disabled. The
only off-switch you need is smd disable <name>, which takes a component
offline reversibly (stops its units and clears the topology flag); smd start <name> brings it back. Pass smd install --no-enable to install without
enabling (e.g. to stage config first).
Clients fall into two groups. Core clients are the DASI2 grant
station's primary instruments — they install and run by default (the
dasi2 profile: hf-timestd, wspr-recorder, psk-recorder, mag-recorder,
meteor-scatter, hamsci-physics, hfdl-recorder, codar-sounder, plus the
gmag-webui and station-web pages).
Additional clients extend the same station to other HF science and are
installed at your discretion — either all at once during guided
bring-up (smd bringup dasi2 --with-optional) or one at a time
(smd install <name>, which installs and enables it). See
Adding a discretionary client later.
| Core client | What it does | Repo |
|---|---|---|
| radiod | ka9q-radio SDR daemon — receives RF and multicasts IQ channels | ka9q/ka9q-radio |
| hf-timestd | HF time-standard analyzer (WWV/WWVH/CHU/BPM) — also the suite's timing authority (publishes the RTP↔UTC offset + tier other clients label against) | HamSCI/hf-timestd |
| wspr-recorder | WSPR/FST4W recorder + decoder — uploads to wsprdaemon.org, which forwards to wsprnet.org | HamSCI/wspr-recorder |
| psk-recorder | FT4/FT8 spot recorder — decodes; wsprdaemon.org forwards its spots to PSKReporter | HamSCI/psk-recorder |
| mag-recorder | RM3100 magnetometer recorder (USB-I²C, non-radiod) — daily datasets to PSWS | HamSCI/mag-recorder |
| meteor-scatter | Meteor-scatter ping decoder (jt9 --msk144) → local sink + hs-uploader |
HamSCI/meteor-scatter |
| hfdl-recorder | HFDL (High Frequency Data Link) recorder — one dumphfdl subprocess per band. Sends ground-station squitter timing to wsprdaemon.org; the airframes.io feed stays off unless you turn it on |
HamSCI/hfdl-recorder |
| codar-sounder | Opportunistic ionospheric sounder using CODAR chirp transmissions | HamSCI/codar-sounder |
| hamsci-physics | Ionospheric science products (TEC/dTEC, TID, GRAPE/PSWS) from hf-timestd observations | mijahauan/hamsci-physics |
| Additional client | What it does | Repo |
|---|---|---|
| hf-tec | HF PRN-coded beacon recorder for ionospheric specification (TEC) | HamSCI/hf-tec |
| superdarn-sounder | Passive SuperDARN coherent-radar monitor (detection / identification) | HamSCI/superdarn-sounder |
| Component | What it does | Repo |
|---|---|---|
| igmp-querier | IGMPv2 querier daemon — keeps multicast streams alive on LANs without a router | HamSCI/igmp-querier |
| sigmond-rac | Remote Access Channel — frpc reverse tunnel to gw2.wsprdaemon.org for admin SSH/web behind NAT | HamSCI/sigmond-rac |
| gpsdo-monitor | Leo Bodnar GPSDO health monitor + mDNS advertiser | HamSCI/gpsdo-monitor |
The RF clients use ka9q-python
to receive RTP streams from radiod; mag-recorder is the exception — it
reads its magnetometer directly over USB-I²C, not through radiod. Each
client runs in its own Python venv and manages its own systemd services.
Decoded spots / products are written to the shared SQLite sink
(/var/lib/sigmond/sink.db) and shipped upstream by hs-uploader.
Why a station is worth more as part of a network: each DASI2 node already carries several GPS-co-registered ionospheric instruments —
hf-timestd(dTEC),hf-tec(coded-beacon absolute TEC),codar-sounder(oblique sounding / MUF),hfdl-recorder(dense opportunistic paths), andmag-recorder(the geomagnetic driver) — so a single node is already a multi-modal observatory. Replicated across the map, the mesh becomes a distributed-aperture instrument that can image travelling ionospheric disturbances, sense TEC gradients by differencing receivers that share a transmitter, and feed tomographic TEC — and it is resilient to the steady loss of the HF time standards it listens to (CHU included). See docs/STATION-NETWORK-CAPABILITIES.md.
Sigmond follows a layered architecture:
- Catalog — static registry of known clients (
etc/catalog.toml). Answers "what could be installed on this host?" - Installer — clones a client repo and delegates to its
install.sh. Each client's installer is authoritative; sigmond never duplicates it. - Topology — per-host config (
/etc/sigmond/topology.toml). Declares which components are enabled. - Lifecycle — resolves systemd units from each client's
deploy.toml, expands templated units, discovers instances, and drives start/stop/restart/reload. - Logging — aggregates logs across clients via the client contract's
log_pathsandlog_levelfields. - Harmonization — cross-client validation rules that catch conflicts (CPU overlap, multicast collisions, missing dependencies).
Each client conforms to the HamSCI client contract,
which defines a standard interface: inventory --json, validate --json,
deploy.toml, systemd unit conventions, and logging discipline.
smd core is stdlib-only; the TUI and test suite need a dev venv:
uv sync --extra tui --extra dev # creates .venv/ with textual, rich, pytestFor prereqs, clone layout, running smd from the tree, tests, and docs
checks, see docs/contributor/dev-setup.md ★.
- Authors: Michael Hauan (AC0G), Rob Robinett (AI6VN)
- License: TBD
- Repo: https://github.com/HamSCI/sigmond
- Part of: HamSCI — Ham Radio Science Citizen Investigation
