huly-platform/services/worker

Time Machine Service

The Time Machine service is an autonomous, generic service responsible for handling delayed events (timers). It replaces the previous Temporal implementation with a database-backed polling mechanism and Kafka-based communication.

How it works

1. Commands: The service consumes commands from the TimeMachine Kafka topic.
2. Storage: Scheduled events are stored in a PostgreSQL table time_machine.delayed_events.
3. Polling: The service periodically polls the database for expired events.
4. Events: When an event expires, the service sends the stored data to the specified topic via Kafka and removes the record from its database.

Kafka Interactions

Consumed (Incoming)

Topic: TimeMachine (timeMachine) Message Type: TimeMachineMessage

Type	Description
schedule	Schedules a new timer or updates an existing one. Requires id, targetDate, topic, and data.
cancel	Removes scheduled timers. The id supports pattern matching via ILIKE (e.g., prefix_%).

Produced (Outgoing)

Topic: Dynamic (specified in schedule command) Message Type: Arbitrary JSON (stored in data)

When a timer expires, the service relays the exact data payload to the target topic.

Environment Variables

Variable	Default	Description
DB_URL	postgres://localhost:5432/huly	Connection string for the PostgreSQL database.
POLL_INTERVAL	5000	Polling interval for expired events in milliseconds.
QUEUE_CONFIG	-	Kafka bootstrap servers configuration.
QUEUE_REGION	cockroach	Platform region configuration.

Database Schema

The service automatically initializes its own schema if it doesn't exist:

• Schema: time_machine
• Table: delayed_events
• Columns: id (text), workspace (uuid), target_date (int8), topic (text), data (jsonb).
• Primary Key: (id, workspace)