cloud-orchestrator

Multi-agent workflow orchestration coordinates complex tasks by splitting them into specialist sub-agents and managing their planning, delegation, critique, and synthesis. This approach uses a shared scratchpad state to organize cross-domain work and traceable handoffs.

Model routing assigns each phase of a complex workflow to the most suitable Ollama Cloud model. It dynamically matches specific tasks like reasoning, coding, research, or critique to the best available model for specialized execution.

Adversarial review is needed for non-trivial work to catch gaps, errors, and weak assumptions. It adds a mandatory critic pass to the workflow orchestration pipeline, ensuring the synthesized output from specialist agents is robust.

Workflow orchestration is specifically designed for projects with three or more phases and cross-domain work. It decomposes complex jobs like a product launch brief into a coordinated pipeline spanning research, architecture, implementation, and QA.

To break down a cross-domain project, pass it to an orchestration workflow that automatically generates a structured task plan. It then delegates phases to specialist sub-agents, executes them, applies critique, and merges the results into one final deliverable.