Semantic Triggers

React to Meaning, Not Just Data

Register conceptual patterns. When new data matches semantically — not just syntactically — fire webhooks, notifications, or agent actions in real time.

HOW IT WORKS

Concept-Based Event Detection

Semantic triggers use a two-stage evaluation: vector pre-filtering narrows candidates, then exact cosine similarity verifies the match. Triggers support cooldown debouncing to prevent alert fatigue. Actions include webhooks (HMAC signed), agent notifications, write events, and human review flags.

-- Register a semantic trigger
INSERT INTO _hatidata_triggers
  (concept, threshold, action)
VALUES (
  'port congestion delay APAC',
  0.7,
  'webhook:supply-chain-alert'
);

-- Fires automatically when new data
-- semantically matches the concept.
-- No polling. No cron jobs.

WHY IT MATTERS

Why Agents Need This

Beyond Rule-Based Alerts

Traditional triggers match exact values. Semantic triggers match meaning — catching novel patterns that rigid rules would miss.

Zero Polling Overhead

Triggers evaluate on data ingestion. No cron jobs, no scheduled queries, no wasted compute scanning for changes.

Cooldown Debouncing

Configurable cooldown periods prevent alert fatigue. A trigger that fires every second isn't useful — one that fires when it matters is.

COMPARISON

Legacy Approach vs ANDI

Legacy Approach

ANDI Approach

Rule-based exact-match alerts
Semantic concept matching
Polling / cron-based detection
Event-driven on ingestion
Custom webhook infrastructure
Built-in with HMAC-SHA256 signing
No alert deduplication
Configurable cooldown debouncing
Separate alerting system
Integrated with the data layer

ANDI SUBSTRATE

Explore the Full Platform

Agent-Native Data Infrastructure.
Production-Ready Today.

Persistent memory. Isolated state. Verifiable reasoning. Through standard SQL.