Repository navigation
feat(runtime): configure an ACP agent's model and reasoning from Quartet - #82
Merged
Merged
Conversation
An ACP agent advertises its session config options — model, reasoning level, mode — on session/new, and takes session/set_config_option to change them. The ACP runtime now captures those options, reapplies the operator's saved choices to every session (a fresh session starts at the agent's defaults), tracks config_option_update the agent pushes on its own, and exposes a setter. Choices persist per identity in config.json and are relayed into the bridge snapshot, so a model or effort picked once survives restarts.
…board The runtime view was a read-only note telling you to configure the agent elsewhere. It now renders a selector per option the agent discloses, bound to the live snapshot, and writing a choice back through /api/runtime/config. Options appear once a room exists to read them from; anything the agent does not expose stays configured in the agent, as the footer still says.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
The options only exist once a session does, so the runtime screen opened to an empty page telling you to go open a room first. It now asks the bridge to probe — open one throwaway session, read the options (applying saved selections), close it — when the screen is shown, with a loading line while it runs and a retry if the agent could not be reached.
The four places that had to agree on the preset list — the config type union, the ACP_PRESETS record, the wizard's CHOICES, and the hand-written error string — collapse into one catalog in runtime-catalog.ts, so adding a built-in is a single entry. A machine can register its own agents in ~/.quartet/runtimes.json (name → command/args/label/installHint); they appear in the wizard, answer to --runtime <name>, and carry an install hint, with built-in and reserved names protected. An agent that does not speak ACP still needs its own TurnRunner; this is only about the ACP shortcuts.
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What & why
ACP agents (Claude, Codex) expose their own session settings — which model, what reasoning level — but until now Quartet treated all of that as off-limits, showing a read-only note telling you to configure the agent by hand. This adds a model/reasoning picker to the runtime view.
Quartet discovers the choices from the agent itself (ACP session config options), so the dropdowns show exactly what the connected agent supports — no hardcoded model list. Opening the runtime screen probes the agent (a throwaway session) to populate the options without making you open a room first. A choice is saved to the identity and reapplied to every session the agent starts, so it survives restarts, and it also tracks changes the agent makes on its own.
How to verify
--runtime claude/--runtime codex) and open Your runtime: it briefly probes the agent, then shows model and reasoning selectors.identities/<id>/config.jsonunderruntime.configOptions).Depends on nothing beyond the adapters already installed. Builds on #81 (merged).
🤖 Generated with Claude Code