PidgeIoT System Architecture

The data flow from constrained edge devices, through our own transport terminators and Cloudflare's global compute network, down to managed PostgreSQL storage.

PidgeIoT platform architectureUsers and IoT devices connect through Cloudflare's edge to the dovecote Worker, either directly over HTTPS and WebSocket or through the loft and pigeonhole transport terminators on our VPS; dovecote routes to per-device Durable Objects and mirrors state to PostgreSQL; telemetry rides a queue to history storage or a user-defined forwarding endpoint.

Bring Your Own (optional)

Off-Edge Services

Cloudflare Edge Compute

Endpoints

Kratos session cookie

HTTPS / WebSocket, Ed25519-signed 69-byte token

CoAP over DTLS/UDP or TLS/TCP

MQTT over TLS

device routes over HTTPS

device routes over HTTPS

proxy_to_pigeon_do

X-User-Id

firmware upload / Range download

best-effort sync

enqueue telemetry

consumer write (history)

optional, per device

User (WASM Dashboard)

IoT Device(s)

fancier Worker
(Static WASM Dashboard Assets)

dovecote Edge Router
(HTTPS / WebSocket Ingestion)

Pigeon Durable Object
(1:1 per device, SQLite, source of truth)

Cloudflare Queue
(Telemetry buffer)

R2 Bucket
(Content-addressed firmware)

Hyperdrive

Ory Kratos
(Dashboard sessions, VPS via Cloudflare Tunnel)

loft
(CoAP terminator, DTLS/UDP + TLS/TCP :5684, VPS)

pigeonhole
(MQTT broker, TLS :8883, same VPS)

Managed PostgreSQL
(Cross-device mirror, telemetry history, identities)

InfluxDB-line-protocol endpoint
(per device)