|
The daemon answers with whichever model llama-server has loaded and
never asks for one, so it never evicts a model another client is using.
The Assistant settings page now lists the router's models with their
state. Load saves one as the preferred model, in Prefs, and asks the
router to load it now, polling /v1/models until it reports loaded;
Prefer marks the loaded model as preferred without swapping. The swap
happens only on that click, so the daemon's rule holds.
AGENTS.md records the router's load endpoint and its two refusals, both
measured: it lives on the server root, not under /v1.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
|
Two settings pages for desktop-assistant.
Assistant edits the daemon's config.toml: server, temperature, tool
rounds, thinking, the system prompt, and the [env] values its tools
read; a switch per tool renames tools/<name>.toml to .toml.off, which
the daemon's glob skips. The daemon reads all of it at startup only, so
the page says when a change is pending and Restart runs assistant.sh
restart, showing its status when the daemon does not come back.
The file is edited line by line, like autostart.lua, so the comments
the example ships with survive: unset keys stay commented out in place.
Every write parses the whole new file with tomllib first and replaces
the old one atomically, so no edit can leave a config the daemon will
not start with. Defaults are read from assistant.py --defaults rather
than copied into QML.
Ask presets edits the canvas's canned questions, which Prefs already
stored for exactly this page: label, question, and the kinds each is
offered for. They apply live.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
