Goose Orchestrator FAQs

Question 1

Can I customize the AI models and their behavior?

Accepted Answer

Yes, Goose Orchestrator allows full customization of worker roles. You can change assigned models, adjust temperature settings (e.g., 0.1 for math, 0.9 for creative), add new roles, or remove existing ones directly through Goose's chat UI.

Question 2

What AI model providers does Goose Orchestrator support?

Accepted Answer

It offers multi-provider support, integrating with local models via Ollama, as well as remote models through OpenAI-compatible APIs and Anthropic.

Question 3

How does it handle complex multi-step tasks?

Accepted Answer

The orchestrator LLM automatically decomposes complex prompts into a dependency graph of sub-tasks. These sub-tasks are then assigned to specialized worker roles and can be executed in parallel where independent, leading to an assembled final response.

Question 4

What is Goose Orchestrator?

Accepted Answer

Goose Orchestrator is a multi-model extension for Goose that intelligently routes user prompts to specialized worker AI models, optimizing task execution with VRAM-aware hot-swapping and parallel processing capabilities.

Question 5

How does Goose Orchestrator manage VRAM for local models?

Accepted Answer

It employs a VRAM-aware model pool with LRU (Least Recently Used) eviction to keep local models within a configurable memory budget. It also features speculative preloading to anticipate and load the next needed worker model.

Goose Orchestrator

Goose Orchestrator

Key Features

Use Cases

Key Features

Use Cases