Repository navigation
fix(chat): the agent's own model and effort on the chips, and no resume on coming back - #56
Merged
Merged
Conversation
…esume it Once the agent had written its session, the conversation's row was the history session's, and a click on it only looked for a terminal: it offered to start the agent again in one. A history row whose conversation is open is that conversation's, as it is a terminal's, also for the context menu, the keys that act on the session under the cursor and the Den.
A conversation that asked for neither said Default for both. The agent is now asked what a turn would run with: Claude Code answers a second request for its settings as it weighs them, Codex's configuration names them or its list marks its own choice, Grok marks a model. What a turn says it answers with (claude-opus-5-5) is shown by the entry a person knows (Opus 5.5), and the lists say what Default is.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Two things found using Leon's own interface (Settings ▸ Experimental) with Claude Code.
The chips said "Default"
A conversation that asked for no model and no effort showed
DEFAULT (RECOMMENDED)andDEFAULT EFFORTinstead of what the agent runs with.AgentOptionsgainsdefault_modelanddefault_effort, andAgentModelgainsresolves_to(the model an entry stands for).get_settings, whoseappliedis the model and the effort after every settings file and default was weighed. Checked against Claude Code 2.1.296 ({"model":"claude-opus-5-5","effort":"medium"}here).modelandmodel_reasoning_effortof its configuration, else the model its list marks and that one'sdefaultReasoningEffort.grok modelsmarks(default). It says nothing of its effort before a session, so that chip still says Default.claude-opus-5-5) is shown as the entry a person knows (Opus 5.5), and the lists say what "Default" is.opencode and Cursor say nothing of their defaults outside a session; their chips are as before.
Coming back to a conversation asked to resume it
Once the agent had written its session, the conversation's row in the tree was the history session's. A click on it only looked for a terminal of that session, found none, and offered "Resume this session? … starts its agent in a terminal".
Shell::running_of_historynow answers for a terminal or a conversation, in the click, the context menu, the keys that act on the session under the cursor, the resume paths and the Den.Tested
scripts/check.shpasses. New tests inleon-turn(each agent's defaults, the script with a stand-in), inchat::complete(what is shown) and inui::tests_chat(the row brings the conversation back). The chips themselves were not looked at in a running window.