Automation Rules

The rule engine lets NeoMind respond automatically without human intervention: device data crosses a threshold → auto-notify, send commands, trigger an AI Agent. Rules are defined in JSON and created via Web UI, CLI, REST API, or AI Chat.

The Automation page has two tabs: Rules and Transforms. This doc covers Rules; for data transforms see Data Transforms.

Prerequisites

• At least one device onboarded (rules reference device metrics as data sources)
• Notification channels configured (if using notify actions)

Interface Overview

Click Automation (branch icon) in the left nav to open the automation page. The default tab is Rules:

The page displays all rules in a table, each row containing:

Column	Description
Rule Name	The name you set
Trigger	Data Change / Schedule / Manual
Condition	Text preview of the condition (e.g. temperature > 30)
Actions	List of actions in this rule (Notify / Execute / Trigger Agent)
Status Toggle	Enable / disable switch
Actions Menu	Edit, delete, execute now

The Import / Export button in the top right lets you bulk import/export rule JSON.

Rule Structure

A rule has four parts — name, trigger, condition, and actions (optional duration and cooldown):

{
  "name": "High Temperature Alert",
  "trigger": { "trigger_type": "data_change" },
  "condition": {
    "condition_type": "comparison",
    "source": "device:sensor-01:temperature",
    "operator": "greater_than",
    "threshold": 30
  },
  "actions": [
    { "type": "notify", "message": "Temperature too high: {value}°C", "severity": "critical" }
  ]
}

Complete JSON field reference

{
  "name": "Sustained High Temperature",
  "trigger": { "trigger_type": "data_change" },
  "condition": {
    "condition_type": "comparison",
    "source": "device:sensor-01:temperature",
    "operator": "greater_than",
    "threshold": 30
  },
  "actions": [
    { "type": "notify", "message": "Temperature above 30°C for 5 minutes", "severity": "critical" }
  ],
  "for_duration": 300000,
  "cooldown": 60000
}

Other Creation Methods

Option 1: Web UI

Step 1: Open the Rule Builder

In the Rules tab, click the Create button to open the full-screen rule builder:

Fill in the top of the builder:

Field	Description
Name	Display name for the rule
Description	Optional, explains the rule's purpose
Trigger	Select the trigger type (see below)

Step 2: Configure the Trigger

Trigger Type	Description	Use Case
Data Change	Auto-evaluates when referenced metrics have new data	Real-time alerts, threshold monitoring
Schedule	Triggers on a cron expression	Scheduled reports, periodic checks
Manual	Only triggered manually (API / CLI / UI button)	Debugging, on-demand execution

Cron uses 6-field format: sec min hour day month weekday. "0 */5 * * * *" = every 5 minutes.

The data_change trigger automatically extracts referenced data sources from the condition — no need to specify sources manually.

Step 3: Configure the Condition

Conditions determine when a rule fires. Three types are supported:

Comparison condition — compares a value to a threshold:

Operator	Meaning	Threshold Field
greater_than	Greater than	threshold (number)
less_than	Less than	threshold (number)
greater_equal	Greater than or equal	threshold (number)
less_equal	Less than or equal	threshold (number)
equal	Equal	threshold (number/boolean)
not_equal	Not equal	threshold
contains	Contains (string)	threshold_value (string)
starts_with	Prefix match	threshold_value
ends_with	Suffix match	threshold_value
regex	Regex match	threshold_value

source uses DataSourceId format {type}:{id}:{field}, e.g. device:sensor-01:temperature.

Range condition — fires when value is within a range:

{ "condition_type": "range", "source": "device:sensor-01:temperature", "min": 20, "max": 25 }

Logical condition — AND / OR / NOT nesting multiple sub-conditions:

{
  "condition_type": "logical",
  "operator": "and",
  "conditions": [
    { "condition_type": "comparison", "source": "device:sensor-01:temperature", "operator": "greater_than", "threshold": 30 },
    { "condition_type": "comparison", "source": "device:sensor-01:humidity", "operator": "less_than", "threshold": 20 }
  ]
}

Step 4: Configure Actions

Actions execute when the condition is met. A rule can have multiple actions, executed in order.

Action	type value	Description
Send notification	notify	Message template supports {value}, {source_id} interpolation
Execute command	execute	Send control command to a device or extension
Trigger Agent	trigger_agent	Call an AI Agent for deep analysis

notify action — severity values: info, warning, critical, emergency

{ "type": "notify", "message": "Temperature too high: {value}°C", "severity": "critical" }

execute action — send a device control command:

{ "type": "execute", "target": "humidifier-01", "target_type": "device", "command": "power_on", "params": { "level": 3 } }

trigger_agent action — call an AI Agent:

{ "type": "trigger_agent", "agent_id": "diagnostic", "input": "sensor-03 is offline, please diagnose" }

Step 5: Duration and Cooldown

Field	Description
For Duration	Condition must be continuously met for this duration before firing, filtering out sensor jitter. The JSON field is in milliseconds (e.g. 5 minutes = 300000; the UI offers seconds / minutes / hours)
Cooldown	Minimum interval between triggers. The JSON field is in milliseconds, default 60000 (60 seconds)

Click Save to save the rule.

CLI

# Create a rule (JSON format)
neomind rule create --body '{"name":"High Temp","trigger":{"trigger_type":"data_change"},"condition":{"condition_type":"comparison","source":"device:sensor-01:temperature","operator":"greater_than","threshold":30},"actions":[{"type":"notify","message":"Too hot"}]}'

# List all rules
neomind rule list

# Enable / disable
neomind rule enable <rule_id>
neomind rule disable <rule_id>

# Dry-run rule evaluation (only evaluates whether the condition is met, without triggering actions)
neomind rule test <rule_id> --input '{"temperature": 35}'

# Delete rule
neomind rule delete <rule_id>

REST API

# Create rule (JSON body)
curl -X POST http://localhost:9375/api/rules \
  -H "Content-Type: application/json" \
  -d '{
    "name": "High Temp",
    "trigger": { "trigger_type": "data_change" },
    "condition": { "condition_type": "comparison", "source": "device:sensor-01:temperature", "operator": "greater_than", "threshold": 30 },
    "actions": [ { "type": "notify", "message": "Too hot" } ]
  }'

# View execution history
curl http://localhost:9375/api/rules/<rule_id>/history

AI Chat

Just tell AI Chat:

"Email me when the temperature goes above 30 degrees"

The LLM auto-generates and creates the rule.

Import / Export

The Import / Export button in the Rules tab supports bulk management:

Operation	Description
Export	Export all rules to a JSON file (neomind-rules-YYYY-MM-DD.json)
Import	Upload a JSON file to bulk import rules; rules that fail to import (e.g. invalid format) are counted as "skipped" and reported in the result

Rule Validation

When creating a rule, NeoMind performs context-aware validation:

• Referenced device exists
• Metric name is valid
• Command parameters match the device type definition
• Extension and Agent IDs are checked to be non-empty

Validation failures return detailed error messages listing which field has the problem.

Complete Examples

1. Temperature & Humidity Combined Alert

High temp + low humidity: notify and turn on humidifier:

{
  "name": "Temp-Humidity Combo",
  "trigger": { "trigger_type": "data_change" },
  "condition": {
    "condition_type": "logical",
    "operator": "and",
    "conditions": [
      { "condition_type": "comparison", "source": "device:sensor-01:temperature", "operator": "greater_than", "threshold": 30 },
      { "condition_type": "comparison", "source": "device:sensor-01:humidity", "operator": "less_than", "threshold": 20 }
    ]
  },
  "actions": [
    { "type": "notify", "message": "High temp low humidity: {value}°C", "severity": "critical" },
    { "type": "execute", "target": "humidifier-01", "target_type": "device", "command": "power_on", "params": { "level": 3 } }
  ]
}

2. Scheduled Energy Report

Trigger Agent at 8 AM daily to summarize energy:

{
  "name": "Daily Energy Report",
  "trigger": { "trigger_type": "schedule", "cron": "0 0 8 * * *" },
  "actions": [
    { "type": "trigger_agent", "agent_id": "energy-reporter", "input": "Summarize yesterday energy data and send report" }
  ]
}

3. Sustained Anomaly Triggers Agent

Device offline for 10+ minutes triggers diagnostic Agent:

{
  "name": "Device Offline Diagnosis",
  "trigger": { "trigger_type": "data_change" },
  "condition": { "condition_type": "comparison", "source": "device:sensor-03:online", "operator": "equal", "threshold": 0 },
  "for_duration": 600000,
  "actions": [
    { "type": "notify", "message": "sensor-03 offline for 10 minutes", "severity": "critical" },
    { "type": "trigger_agent", "agent_id": "diagnostic", "input": "sensor-03 is offline, please diagnose" }
  ]
}

Execution History

Click the actions menu on any rule row to view execution history:

• Trigger time
• Whether the condition was met
• How many actions executed
• Result of each action (success/failure/reason)
• Evaluation duration

Integration with Other Modules

Module	Description
Notifications	notify action routes to configured notification channels
AI Agent	trigger_agent action calls an autonomous agent for deep analysis
Devices	execute action sends device commands
Data Transforms	Rules can reference derived metrics from Transforms
Data Push	Rules evaluate data in-platform and trigger actions; Data Push sends data off-platform
AI Chat	Create rules in natural language, LLM auto-generates JSON

Best Practices

• Add for_duration for debounce: Sensor data is noisy; use "for_duration": 120000 (2 minutes) to filter transient spikes (unit is milliseconds)
• Set cooldown to prevent spam: High-frequency data sources need cooldown to prevent alert storms
• Tiered notifications: Regular alerts severity: "info", severe alerts severity: "critical"
• Prefer rules over Agents: Deterministic logic uses rules (millisecond evaluation), fuzzy judgment uses Agents (seconds of LLM analysis)
• Idempotent actions: Design device commands as idempotent (e.g. power_on safe to call repeatedly), preventing side effects from rule retries

Next Steps

• Data Push — Send in-platform data to external systems in real time
• Notifications — Configure channels and filters for notify actions
• AI Agent — Use trigger_agent actions for deep Agent analysis

---

Last updated: 2026-09-09