Preserve agent question answers and plan output

This commit is contained in:
Ryan Ward
2026-08-25 12:56:25 -05:00
parent b29f35e0fb
commit 78b60af988
5 changed files with 90 additions and 6 deletions
+1 -1
View File
@@ -24,7 +24,7 @@ Environment variables:
- `WS_SERVER_URL` - WebSocket endpoint (default: `ws://localhost:8080/graphql/v2`)
### Testing
- During interactive bug-fix verification, make the requested code changes first and run only `cargo run --bin galaxy-oss` for the user to verify. Leave the app running for the user; do not stop it or treat the command timeout as a failure. Do not run `cargo fmt`, `cargo check`, `cargo clippy`, presubmit, or other validation commands until the user confirms the fix.
- During interactive feature and bug-fix verification, make the requested code changes first and run only `cargo run --bin galaxy-oss` for the user to verify. Leave the app running for the user; do not stop it or treat the command timeout as a failure. Do not run `cargo fmt`, `cargo check`, `cargo test`, `cargo clippy`, presubmit, or other validation commands until the user confirms the behavior. Run formatting, tests, and linting only as the final cleanup step after interactive approval.
- `cargo nextest run --no-fail-fast --workspace --exclude command-signatures-v2` - Run tests with nextest
- `cargo nextest run -p galaxy_completer --features v2` - Run completer tests with v2 features
- `cargo test --doc` - Run doc tests
+1 -1
View File
@@ -807,7 +807,7 @@ fn build_system_prompt(
match mode {
RigRequestMode::Normal => {}
RigRequestMode::Plan => prompt.push_str(
"## Plan Mode\nInspect and produce an implementation-ready plan. Do not edit files or perform state-changing actions.\n\n",
"## Plan Mode\nInspect and produce an implementation-ready plan. Do not edit files or perform state-changing actions. Research as needed, then finish by calling `create_plan` to write the plan with the built-in planning tools. If a plan document already exists for this task, call `edit_plan` instead. Do not return the plan only as prose, and do not claim completion until the plan tool succeeds.\n\n",
),
RigRequestMode::Orchestrate => prompt.push_str(
"## Orchestration Mode\nDelegate only independent, bounded work where parallelism materially helps, then synthesize the results.\n\n",
+36
View File
@@ -200,6 +200,42 @@ fn normal_turn_advertises_plan_creation_and_corrects_false_unavailability_claims
assert!(prompt.contains("do not implement it until they approve"));
}
#[test]
fn plan_turn_requires_a_plan_tool_result() {
let mut params = RequestParams::new_for_test();
let mut input = user_query("Design the implementation");
let AIAgentInput::UserQuery {
user_query_mode, ..
} = &mut input
else {
unreachable!("user_query returns a user query input");
};
*user_query_mode = UserQueryMode::Plan;
params.input = vec![input];
let prepared = prepare_rig_turn(
&config(),
params,
vec![ToolType::CreateDocuments, ToolType::EditDocuments],
Vec::new(),
);
let prompt = prepared.request.system_prompt.expect("system prompt");
assert!(prepared
.request
.tools
.iter()
.any(|tool| tool.name == "create_plan"));
assert!(prepared
.request
.tools
.iter()
.any(|tool| tool.name == "edit_plan"));
assert!(prompt.contains("finish by calling `create_plan`"));
assert!(prompt.contains("call `edit_plan` instead"));
assert!(prompt.contains("Do not return the plan only as prose"));
}
#[test]
fn no_tools_turn_flattens_historical_tool_protocol_messages() {
let mut params = RequestParams::new_for_test();
+32 -1
View File
@@ -294,13 +294,44 @@ impl AIAgentActionResultType {
| Self::FetchConversation(_)
| Self::StartAgent(_)
| Self::SendMessageToAgent(_)
| Self::AskUserQuestion(_)
| Self::WaitForEvents(_) => self.to_string(),
Self::AskUserQuestion(result) => ask_user_question_result_content(result),
Self::RunAgents(result) => result.model_content(),
}
}
}
fn ask_user_question_result_content(result: &AskUserQuestionResult) -> String {
match result {
AskUserQuestionResult::Success { answers } => {
let answers = answers
.iter()
.map(|answer| match answer {
AskUserQuestionAnswerItem::Answered {
question_id,
selected_options,
other_text,
} => serde_json::json!({
"question_id": question_id,
"selected_options": selected_options,
"other_text": other_text,
}),
AskUserQuestionAnswerItem::Skipped { question_id } => serde_json::json!({
"question_id": question_id,
"skipped": true,
}),
})
.collect::<Vec<_>>();
serde_json::json!({ "answers": answers }).to_string()
}
AskUserQuestionResult::SkippedByAutoApprove { question_ids } => serde_json::json!({
"skipped_by_auto_approve": question_ids,
})
.to_string(),
AskUserQuestionResult::Error(_) | AskUserQuestionResult::Cancelled => result.to_string(),
}
}
fn command_result_content(command: Option<&str>, output: &str, exit_code: i32) -> String {
let command = command
.map(|command| format!("Command: {command}\n"))
+20 -3
View File
@@ -1,9 +1,26 @@
use super::{
AIAgentActionResultType, RequestCommandOutputResult, RunAgentsAgentOutcome,
RunAgentsAgentOutcomeKind, RunAgentsLaunchedExecutionMode, RunAgentsResult, StartAgentResult,
StartAgentVersion,
AIAgentActionResultType, AskUserQuestionAnswerItem, AskUserQuestionResult,
RequestCommandOutputResult, RunAgentsAgentOutcome, RunAgentsAgentOutcomeKind,
RunAgentsLaunchedExecutionMode, RunAgentsResult, StartAgentResult, StartAgentVersion,
};
#[test]
fn ask_user_question_model_content_includes_selected_answers() {
let result = AIAgentActionResultType::AskUserQuestion(AskUserQuestionResult::Success {
answers: vec![AskUserQuestionAnswerItem::Answered {
question_id: "scope".to_string(),
selected_options: vec!["Provider-neutral pane plus agent tool".to_string()],
other_text: String::new(),
}],
});
let content: serde_json::Value = serde_json::from_str(&result.model_content()).unwrap();
assert_eq!(
content["answers"][0]["selected_options"][0],
"Provider-neutral pane plus agent tool"
);
}
#[test]
fn shell_execution_error_is_failed_but_not_cancelled() {
let result =