Skip to content
ContextUnity

ContextWorker

ContextWorker is the “Hands” of the ContextUnity ecosystem. It provides durable workflow infrastructure, scheduled jobs, and sub-agent execution — all built on Temporal.io.

Durable Workflows

Temporal.io workflows that survive restarts and failures, with automatic retries and state persistence.

Scheduled Jobs

Cron-based recurring tasks — harvesting, enrichment, catalog sync, data retention.

gRPC Service

Trigger and monitor workflows from other services via ContextUnit protocol.

Sub-Agent System

Isolated execution environments for agent-generated code with Brain recording.

Architecture

src/contextworker/
├── __main__.py          # CLI: gRPC service (default) or --temporal
├── config.py            # WorkerConfig (Pydantic settings)
├── service.py           # gRPC WorkerService
├── server.py            # gRPC server setup
├── schedules.py         # Temporal schedule management
├── interceptors.py      # gRPC interceptors
│
├── core/
│   ├── registry.py      # WorkerRegistry, ModuleConfig, plugin discovery
│   └── worker.py        # Temporal client setup and worker creation
│
├── subagents/           # Sub-agent execution framework
│   ├── executor.py      # SubAgentExecutor (orchestrator)
│   ├── isolation.py     # IsolationManager, IsolationContext
│   ├── brain_integration.py  # Step recording to Brain
│   ├── local_compute.py # Sandboxed Python execution
│   ├── rlm_tool.py      # RLM tool integration
│   └── monitor.py       # Execution monitoring
│
└── jobs/
    └── retention.py     # Data retention policies

Key Design Principle

Worker contains NO business logic. It provides infrastructure only.

Business logic lives in domain packages:

  • Commerce harvesting → ContextCommerce (registers workflows)
  • AI enrichment → ContextRouter (runs agents)

Modules register themselves via the WorkerRegistry and are discovered at startup.