Workflow

How it works

At a high level, a task moves through the same general path every time — the specific resource it lands on depends on what the task needs and what's available.

1

You submit a task

Through the web app, desktop client, mobile client, or API/CLI — one place to describe what you need done.

2

GemmaNode analyzes the task

The request is reviewed at a high level to understand its general nature — coding, research, generation, analysis, and so on.

3

Available resources are considered

GemmaNode looks at what's currently available across the resource pool — local hardware, cloud compute, AI providers, and agent frameworks.

4

Work is assigned to a suitable resource

Based on general factors like capability, availability, and cost. The exact selection logic is internal — this page describes what happens, not how it's computed.

5

Independent work can run concurrently

When resources allow, multiple agents or parts of a task can execute in parallel rather than strictly one at a time.

6

Results come back to you

Output is collected, consolidated, and returned through whichever client you started from.

Difficult or uncertain tasks can optionally be escalated for manual review by a stronger model rather than forced through an automated path that isn't working. This is a deliberate fallback, not the default route for routine work.