Huabu 0.9.1
Highlights: the desktop app now updates itself, each conversation can use its own model, and connected external agents can read and edit your Space directly.
✨ New
- Desktop auto-update — Huabu checks for new versions and prompts you; you decide when to update.
- Per-conversation model settings — pick a model and reasoning effort for each conversation (with Auto); the choice is saved per thread. The model list now shows descriptions, pricing, and context-window size, supports live OpenAI model discovery, and lets you sign in to OpenAI Codex via ChatGPT OAuth.
- External agents can read & edit your Space directly — a connected external agent can now query and modify the current Space on its own (read the outline/nodes, search, snapshot sketches, and create/edit/connect/move/delete nodes) without routing through Huabu's built-in agent. Changes stay scoped to the current Space and conversation, and show up in the change-review card so you can review or undo them.
- Agent identity — choose a custom avatar/icon when creating an agent (saved on its Profile); question nodes now show the agent's avatar and a status badge.
🐛 Fixes
-
External agent (ACP) controls — the model menu no longer flattens every model into one entry per reasoning level; the reasoning-effort, Plan, and Fast controls work again. Switching an agent to full-access now requires an explicit risk confirmation.