Models & Capabilities

Understand which model Huabu uses for each kind of work and configure optional web, image, and video capabilities.

Choose your models

Open Settings > Huabu Agent to configure models and optional capabilities. Chat and Agent require a Provider, its credentials or sign-in, and a Chat Model or Azure OpenAI Deployment.

Setting	Used for	Required
Chat Model or Deployment	Conversations, Agent work, and running Skills.	Yes
Utility Model	Memory, Skill creation and updates, labels, summaries, keywords, and intent suggestions.	No
Image Generation	Creating or editing images through Agent.	No

Chat Model

The Chat Model powers built-in Chat and Agent conversations. It reasons over your request, reads Space material, chooses tools, and plans Space changes. Running a Skill to complete a task also uses this model. Chat acts as the fallback when the Utility Model is not configured or cannot process required image input.

When a Provider exposes multiple models, Settings labels the selection Default model. Built-in Chat and Agent conversations use it until you switch models above the chat input. The new choice is saved for that conversation only and does not change the Settings default. A Provider or account that exposes one model shows Model instead. Azure OpenAI uses the single Deployment configured in Settings, so there is no conversation-level model choice. Models with configurable reasoning also show a Reasoning control above the chat input.

For a multi-model Provider, keep a capable model with a sufficiently large context window as your Default model, then switch individual conversations when a task benefits from a different model.

Utility Model

The optional Utility Model handles frequent, lightweight background work: Memory curation, Skill creation and updates, intent suggestions, image and Frame labels, and content labels, summaries, and keywords. A faster or less expensive model can reduce latency and cost for these tasks.

Leave Utility Model set to Automatic (cheapest) and Huabu selects the least expensive eligible model that it can confirm for your Chat Model provider. If availability cannot be confirmed, including with a Codex subscription login, Huabu uses the Chat Model. For predictable capability and cost, choose a specific Utility Model instead.

Image understanding

Chat and Agent can inspect images when the selected Chat Model supports image input. Huabu also uses image input when it labels image nodes or interprets visual context. If a configured Utility Model cannot accept images, Huabu falls back to the Chat Model for that image task.

Image generation

Image Generation is separate from the Chat and Utility Models. Configure an Azure OpenAI image deployment in Settings > Huabu Agent to let Agent create images or use Space images and sketches as visual references. Built-in Chat is read-only and does not generate images in the Space.

Web search and video transcripts

Capability	What it enables	Setup
Web Search	Lets Chat and Agent find current information and cite the source URLs.	Add a Tavily API key under Settings > Huabu Agent > Other Capabilities.
YouTube Transcripts	Lets Huabu import transcript text when processing YouTube material.	Add a RapidAPI key under Settings > Huabu Agent > Other Capabilities.

Skills provide reusable instructions; they do not add new tools. Web search, image generation, and transcript loading must be configured as capabilities separately.