Models & Capabilities

Use your existing agents for conversations and background tasks, and configure optional Huabu services separately.

Bring your own agent

Connect an agent you already use in Settings > Agent. Huabu provides the visual workspace, Space context, conversation interface, and change review; the agent's harness provides its model access, authentication, tools, and permissions. No built-in model setup is required to start working with a connected agent. See External Agents for supported ACP connections.

Setting	Used for	Required
Agent Profile	Conversations and work on the Space using your existing agent.	Connect an agent, or explicitly configure Built-In Pi instead.
Utility tasks	Huabu-owned background work such as titles, summaries, keywords, and image labels.	A configured and available Utility Agent.
Services	Optional Huabu integrations for OCR, image generation, web search, and video transcripts.	Only for the integration you want to use.

Utility tasks

Under Settings > Agent > Utility tasks, select an external Profile for automatic titles, summaries, keywords, image labels, and other supported background tasks. On first setup, Huabu selects an available external Profile automatically when no Utility Agent configuration has been saved. Review this selection if you already have settings from an earlier version.

A supported external Profile can use a separate model preference for Utility tasks when its harness exposes model selection. Otherwise it uses the harness default. The Utility Agent does not change which Profile an existing conversation uses.

The selected Utility Agent must remain connected and authenticated. Huabu does not silently switch to Built-In Pi when that agent is unavailable. Background tasks cannot wait for interactive tool approval. Automatic Huabu Memory curation currently requires Built-In Pi as the Utility Agent; the built-in Skill authoring flow also remains separate from external agents.

Optional Built-In Pi

If you prefer Huabu's built-in Chat and Agent, expand Built-In Pi under Settings > Agent and open its configuration. Choose a provider, complete its credentials or sign-in, and select a Chat Model or Azure OpenAI Deployment. These settings are not required for external Agent Profiles.

Built-In Pi has its own Chat Model and Utility Model settings. For providers with multiple models, the conversation can select a different model without changing the Settings default; Azure OpenAI uses the configured Deployment. To use Pi for background work, explicitly select it under Utility tasks.

Image understanding

Image input depends on both the connected agent's ACP support and the selected model's vision capability. The same requirement applies to external Utility tasks that inspect images. Huabu does not fall back to a built-in model when an external agent cannot process an image. Built-In Pi uses its own model configuration for image input.

Image generation

Huabu's Image Generation integration is separate from Agent Profiles. Configure it in Settings > Services when you want to use that integration. Connecting an agent does not configure Huabu's service credentials. An agent may also provide image tools of its own, subject to its harness and permissions.

Web search and video transcripts

Capability	What it enables	Setup
Web Search	Provides the Huabu web-search integration; connected agents may also have their own search tools.	Configure the search service under Settings > Services.
YouTube Transcripts	Lets Huabu import transcript text when processing YouTube material.	Configure the transcript service under Settings > Services.
OCR	Adds text-recognition evidence when interpreting handwritten input.	Configure the OCR service under Settings > Services.

Connecting an agent is not a guarantee that every model or tool capability is available. Huabu services and an agent's own tools have separate setup, accounts, and usage charges. Skills provide instructions; they do not supply missing tools or credentials.