GenerateSaaSChat
The side panel's persisted, on-device Chat tab - a multi-turn conversation with the user's chosen model, run by the companion daemon.
The right side panel's Chat tab is a persisted, multi-turn conversation with the user's chosen model. Every turn runs on the user's machine through the companion daemon, driving their own connected coding CLI - nothing goes to your backend, and their own subscription pays.
What the Chat tab gives the user
- Multi-turn, with memory: the conversation remembers prior turns, so the model has the context of everything said so far.
- Per-chat model: a model picker on the panel, sourced from the connected coding CLIs (the same set the Models tab lists). A new chat opens on the device default and can be re-pointed per conversation.
- Streaming: the assistant's reply streams into the panel as it is produced; a Stop button ends a turn early.
- Compact and clear: Compact summarizes the older turns so a long conversation keeps fitting the model's context window; Clear starts a fresh conversation while keeping the session.
Sessions
A conversation is a session: a persisted, ordered list of turns, stored locally on the user's machine and never sent to your backend. Sessions are created on the first message and survive app restarts, so a conversation picks up where it left off. The session switcher lists them, and each can be renamed or deleted.
When the companion is offline
The Chat tab runs through the companion daemon, so if the daemon is not running the tab says so and asks the user to bring it online rather than failing silently. A chat also needs a connected coding CLI to run: connect one first (see Models).
Tasks
Builder-declared AI tasks the end-user assigns a model and reasoning effort to, per task and per device - stored on their own machine.
Terminal and side panel
The right side panel's two built-in tabs - Chat and Terminal - how to choose which ship and which opens first, and what the Terminal gives the user's own CLI.