alex/g3
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(g3-core): disable auto-continue in interactive mode

Browse Source 
Auto-continue was incorrectly triggering when the LLM asked questions
in interactive/chat mode. Now auto-continue only activates when
is_autonomous is true, allowing proper back-and-forth conversation
in interactive mode.

Agent: fowler
This commit is contained in:
Dhanji R. Prasanna
2026-01-07 10:37:30 +11:00
parent a553764e93
commit 48036d01e3
1 changed files with 4 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
crates/g3-core/src/lib.rs
View File

@@ -2658,9 +2658,11 @@ impl<W: UiWriter> Agent<W> {
// This ensures we don't return control when the LLM clearly intended to call a tool // This ensures we don't return control when the LLM clearly intended to call a tool
// Note: We removed the redundant condition (any_tool_executed && is_empty_response) // Note: We removed the redundant condition (any_tool_executed && is_empty_response)
// because it's already covered by (any_tool_executed && !final_output_called) // because it's already covered by (any_tool_executed && !final_output_called)
let should_auto_continue = (any_tool_executed && !final_output_called) // Auto-continue is only enabled in autonomous mode - in interactive mode,
// the user may be asking questions and we should return control to them
let should_auto_continue = self.is_autonomous && ((any_tool_executed && !final_output_called)
|| has_incomplete_tool_call || has_incomplete_tool_call
|| has_unexecuted_tool_call; || has_unexecuted_tool_call);
if should_auto_continue { if should_auto_continue {
if auto_summary_attempts < MAX_AUTO_SUMMARY_ATTEMPTS { if auto_summary_attempts < MAX_AUTO_SUMMARY_ATTEMPTS {
auto_summary_attempts += 1; auto_summary_attempts += 1;