meshdaemon - Meshtastic Async Daemon

Multi-source async daemon that:

• Polls REST endpoints on different intervals,
• Listens to UDP multicast for Meshtastic (via mudp),
• Sends messages it has not heard before to the mesh

We use this in Stockholm to broadcast weather and PSA warnings. This script can be run by multiple nodes and any warnings heard before broadcast will be suppressed. This way multiple users can run the script without drowning the mesh in warnings.

Uses Sveriges Radio's API for Important Public Announcements / Viktigt meddelande till allmänheten (3.0) to extract alerts for a given region and broadcast to a local Meshtastic® network.

Setup

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Configure

Environment variables (with defaults):

Variable	Default	Purpose
MESHTASTIC_MCAST_GRP	224.0.0.69	UDP multicast group for mesh packets
MESHTASTIC_MCAST_PORT	4403	UDP multicast port
MESHTASTIC_NODE_ID	!112	Node identifier advertised by daemon
MESHTASTIC_LONG_NAME	VMA ROBOT	Long display name
MESHTASTIC_SHORT_NAME	VMA	Short display name
MESHTASTIC_CHANNEL	MediumFast	Channel name for outgoing packets
MESHTASTIC_KEY	AQ==	Channel key
MESHTASTIC_HOP_LIMIT	5	Hop limit for text + nodeinfo packets
MESHTASTIC_MAX_BYTES	200	Max UTF-8 bytes per text chunk
MESHTASTIC_MAX_MESSAGES	2	Max chunks per outbound alert
NODEINFO_INTERVAL_SECS	43200	Nodeinfo broadcast interval (12h)
SEEN_TTL_SECS	86400	Dedupe memory TTL for seen messages
VMA_URL	Sveriges Radio URL	VMA API endpoint
VMA_INTERVAL	60	VMA polling interval in seconds
VMA_GEOCODE	01	VMA geocode filter
SMHI_URL	SMHI URL	SMHI warning endpoint
SMHI_INTERVAL	60	SMHI polling interval in seconds
SMHI_GEOCODE	1	SMHI area id filter
WARMUP	1	1 = mark first fetch as seen only, do not transmit
MAX_RETRIES	3	HTTP retries per fetch cycle
BASE_BACKOFF	2	Exponential backoff base seconds
MAX_RESTART_INTERVAL	60	Worker restart throttle window
RESTART_HISTORY	5	Worker failures in window before stop

Runtime behavior

• Warmup: on source startup, first fetched messages are inserted into dedupe state with seen_only=True; this suppresses startup floods.
• Dedupe strategy: All instances use the same deterministic CRC32 hash of normalized message text to generate packet IDs when broadcasting. When receiving packets from the mesh, we store the packet ID in the dedupe cache. Since all instances send with content-based IDs, receiving an identical message (from any source node) will have the same packet ID, triggering dedupe and preventing relay floods.
• Failure policy: each worker restarts after crashes; if crashes reach RESTART_HISTORY within MAX_RESTART_INTERVAL, the task fails and the daemon shuts down.

Logging style

• Use component tags in uppercase (for example [ASYNC], [UDP], [VMA], [SMHI]).
• Prefer action-first, present-tense phrasing (for example Task completed: ..., Source started (...)).
• Keep retry logs consistent as Request failed ...; retrying in ....
• Use ... started / ... stopped wording for lifecycle logs.
• Keep per-message logs explicit (Message ready: ...) and aggregate counters concise (Messages fetched: N).

Run

python -m app.main

---

Meshtastic® is a registered trademark of Meshtastic LLC. Meshtastic software components are released under various licenses, see GitHub for details. No warranty is provided - use at your own risk.