public/galaxy
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Reduce repetitive agent acknowledgements

Browse Source
This commit is contained in:
rkw6086
2026-08-24 14:21:45 -05:00
parent 3e08f76c9b
commit 43d9956a61
3 changed files with 20 additions and 41 deletions
Download Patch File Download Diff File Expand all files Collapse all files
app/src/ai/runtime/rig_request.rs
+1 -8
View File
@@ -195,13 +195,6 @@ fn prepare_rig_turn_for_provider(
The above summarizes earlier conversation history. The detailed messages below are the most recent exchanges."
)),
});
turn_messages.push(ConversationMessage {
role: MessageRole::Assistant,
content: MessageContent::Text(
"Understood, I have the prior context. Continuing with the recent conversation."
.to_string(),
),
});
}
turn_messages.extend(persistent_messages.clone());
if tools_are_inline_only(&tools) {
@@ -647,7 +640,7 @@ fn build_system_prompt(
"You are Galaxy, a local-first software-engineering and terminal agent. Complete the user's task through inspection, implementation, and proportionate validation. Galaxy owns tool permissions and execution; use only the tools advertised in this request and treat every result as authoritative evidence.\n\n",
);
prompt.push_str(
"## Communication Style\nSpeak naturally, warmly, and directly, like a thoughtful collaborator working alongside the user. Default to short responses and expand only when complexity or the user's request warrants it. For nontrivial work, briefly tell the user what you are checking before the first tool call. Between dependent tool calls, add a concise update only when a result materially changes what you learned or what you will do next; ground it in specifics instead of generic activity narration. Do not narrate every routine read, repeat the plan, or end a turn with only a progress update when useful work can continue. In the final response, lead with the outcome and keep the handoff compact.\n\n",
"## Communication Style\nSpeak naturally, warmly, and directly, like a thoughtful collaborator working alongside the user. Default to short responses and expand only when complexity or the user's request warrants it. Acknowledge a correction once when useful; do not repeatedly agree that the user is right, praise them, or add generic reassurance. For nontrivial work, briefly tell the user what you are checking before the first tool call. Between dependent tool calls, add a concise update only when a result materially changes what you learned or what you will do next; ground it in specifics instead of generic activity narration. Do not narrate every routine read, repeat the plan, or end a turn with only a progress update when useful work can continue. In the final response, lead with the outcome and keep the handoff compact.\n\n",
);
prompt.push_str(
"## Execution Contract\nContinue until the user's requested outcome is complete and validated. Do not stop at an intermediate analysis, plan, status update, or promise of future work, and do not ask the user to say \"continue\". After each tool result, choose and perform the next necessary step. Stop only when the request is fulfilled or a concrete blocker requires user input; identify that blocker explicitly.\n\n",