Parallel Cloud Agent Execution
Parallel Cloud Agent Execution
As of v1.0.142, SaaS Factory's core orchestration loop supports genuine parallel agent execution via multi-slot concurrent implementation. Multiple implementation tasks can now run side-by-side within a single cycle, closing the gap between the platform's architectural design and its runtime behaviour.
How It Works
The platform's unified orchestration loop (unified-loop.ts) manages work across one or more implementation slots. Each slot represents an independent unit of concurrent execution — an agent thread that can pick up, implement, and complete a task independently of other slots.
Prior to v1.0.142, the platform only ever activated a single slot at a time, even though the slot-based architecture was designed to support many. This release enables the runtime to spin up multiple slots concurrently.
Health-Gated Scaling
Parallel execution is gated on project health to ensure stability:
- Non-critical health → Multiple slots are enabled. The orchestrator schedules tasks across concurrent slots, increasing throughput.
- Critical health → A single slot runs sequentially. This prevents compounding failures and keeps recovery focused and deterministic.
The health gate means parallelism is applied opportunistically — you get speed when the project is stable, and caution when it is not.
What This Changes in Practice
- Higher throughput during healthy operation: More implementation tasks complete per cycle, accelerating feature velocity.
- No change to critical-state behaviour: Projects experiencing failures continue to recover under a single-slot sequential model.
- No configuration required: Slot scaling is automatic based on the platform's own health signals.
Architecture Note
The slot-based model in unified-loop.ts was designed from the start to accommodate multiple concurrent implementation threads. The v1.0.142 release activates this capability at the platform orchestration level — the concept was always present; the runtime now reflects it.