Sensor Pipeline

The sensor pipeline ingests data from external sources, triages events by urgency, correlates them into stories, and produces daemon commands.

Architecture

  Sources                  Triage              Stories         Daemon
  -------                  ------              -------         ------
  +-----------+
  |  RSS      |--+
  +----------+  |
  |  Email    |--+      +------------+     +------------+   +---------+
  |  (JMAP/   |  +----->|  Triage     |---->|  Story      |-->| Command  |
  |  Google)  |--+      |  (per-type  |    |  Builder    |   | Producer |
  +----------+  |      |   scoring)  |    |  (dedup,    |   | -> Kafka |
  |  Webhook  |--+      +------------+    |   merge)    |   +---------+
  +----------+  |                         +------------+
  |  Weather  |--+
  +----------+  |
  |  Audio    |--+
  +----------+  |
  |  Image    |--+
  +----------+  |
  |  MCP      |--+
  +----------+

Sources

Each source implements the Sensor trait with name(), modality(), and run():

Source	Modality	Description
RSS	Text	Poll RSS/Atom feeds
Email (JMAP)	Text	Google Workspace / JMAP email monitoring
Webhook	Text	HTTP webhook receiver
Weather	Text	Weather data polling
Audio	Audio	Audio input processing
Image	Image	Image input processing
MCP	Text	MCP server events

Triage

Per-modality classifiers score urgency and relevance of incoming sensor events. Events below the threshold are dropped; high-urgency events are fast-tracked.

Stories

The StoryCorrelator aggregates related events:

• Deduplication — prevents the same event from triggering multiple actions
• Merge — combines related events into a single story
• Action production — completed stories produce DaemonCommand entries sent to Kafka

Configuration

[daemon.sensors]
# Sensor sources are configured per-type
# See daemon configuration for full reference

Feature Flag

Sensors require the sensor feature flag (which implies daemon):

cargo build --features sensor