Skip to content
Merged
12 changes: 10 additions & 2 deletions codex-rs/core/src/event_mapping.rs
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@ use tracing::warn;
use uuid::Uuid;

use crate::user_instructions::UserInstructions;
use crate::user_shell_command::is_user_shell_command_text;

fn is_session_prefix(text: &str) -> bool {
let trimmed = text.trim_start();
Expand All @@ -31,7 +32,7 @@ fn parse_user_message(message: &[ContentItem]) -> Option<UserMessageItem> {
for content_item in message.iter() {
match content_item {
ContentItem::InputText { text } => {
if is_session_prefix(text) {
if is_session_prefix(text) || is_user_shell_command_text(text) {
return None;
}
content.push(UserInput::Text { text: text.clone() });
Expand Down Expand Up @@ -197,7 +198,14 @@ mod tests {
text: "# AGENTS.md instructions for test_directory\n\n<INSTRUCTIONS>\ntest_text\n</INSTRUCTIONS>".to_string(),
}],
},
];
ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: "<user_shell_command>echo 42</user_shell_command>".to_string(),
}],
},
];

for item in items {
let turn_item = parse_turn_item(&item);
Expand Down
1 change: 1 addition & 0 deletions codex-rs/core/src/lib.rs
Original file line number Diff line number Diff line change
Expand Up @@ -81,6 +81,7 @@ mod function_tool;
mod state;
mod tasks;
mod user_notification;
mod user_shell_command;
pub mod util;

pub use apply_patch::CODEX_APPLY_PATCH_ARG1;
Expand Down
157 changes: 128 additions & 29 deletions codex-rs/core/src/tasks/user_shell.rs
Original file line number Diff line number Diff line change
@@ -1,28 +1,35 @@
use std::sync::Arc;
use std::time::Duration;

use async_trait::async_trait;
use codex_protocol::models::ShellToolCallParams;
use codex_async_utils::CancelErr;
use codex_async_utils::OrCancelExt;
use codex_protocol::user_input::UserInput;
use tokio::sync::Mutex;
use tokio_util::sync::CancellationToken;
use tracing::error;
use uuid::Uuid;

use crate::codex::TurnContext;
use crate::exec::ExecToolCallOutput;
use crate::exec::SandboxType;
use crate::exec::StdoutStream;
use crate::exec::StreamOutput;
use crate::exec::execute_exec_env;
use crate::exec_env::create_env;
use crate::parse_command::parse_command;
use crate::protocol::EventMsg;
use crate::protocol::ExecCommandBeginEvent;
use crate::protocol::ExecCommandEndEvent;
use crate::protocol::SandboxPolicy;
use crate::protocol::TaskStartedEvent;
use crate::sandboxing::ExecEnv;
use crate::state::TaskKind;
use crate::tools::context::ToolPayload;
use crate::tools::parallel::ToolCallRuntime;
use crate::tools::router::ToolCall;
use crate::tools::router::ToolRouter;
use crate::turn_diff_tracker::TurnDiffTracker;
use crate::tools::format_exec_output_str;
use crate::user_shell_command::user_shell_command_record_item;

use super::SessionTask;
use super::SessionTaskContext;

const USER_SHELL_TOOL_NAME: &str = "local_shell";

#[derive(Clone)]
pub(crate) struct UserShellCommandTask {
command: String,
Expand Down Expand Up @@ -78,34 +85,126 @@ impl SessionTask for UserShellCommandTask {
}
};

let params = ShellToolCallParams {
let call_id = Uuid::new_v4().to_string();
let raw_command = self.command.clone();

let parsed_cmd = parse_command(&shell_invocation);
session
.send_event(
turn_context.as_ref(),
EventMsg::ExecCommandBegin(ExecCommandBeginEvent {
call_id: call_id.clone(),
command: shell_invocation.clone(),
cwd: turn_context.cwd.clone(),
parsed_cmd,
is_user_shell_command: true,
}),
)
.await;

let exec_env = ExecEnv {
command: shell_invocation,
workdir: None,
cwd: turn_context.cwd.clone(),
env: create_env(&turn_context.shell_environment_policy),
timeout_ms: None,
sandbox: SandboxType::None,
with_escalated_permissions: None,
justification: None,
arg0: None,
};

let tool_call = ToolCall {
tool_name: USER_SHELL_TOOL_NAME.to_string(),
call_id: Uuid::new_v4().to_string(),
payload: ToolPayload::LocalShell { params },
};
let stdout_stream = Some(StdoutStream {
sub_id: turn_context.sub_id.clone(),
call_id: call_id.clone(),
tx_event: session.get_tx_event(),
});

let router = Arc::new(ToolRouter::from_config(&turn_context.tools_config, None));
let tracker = Arc::new(Mutex::new(TurnDiffTracker::new()));
let runtime = ToolCallRuntime::new(
Arc::clone(&router),
Arc::clone(&session),
Arc::clone(&turn_context),
Arc::clone(&tracker),
);
let sandbox_policy = SandboxPolicy::DangerFullAccess;
let exec_result = execute_exec_env(exec_env, &sandbox_policy, stdout_stream)
.or_cancel(&cancellation_token)
.await;

if let Err(err) = runtime
.handle_tool_call(tool_call, cancellation_token)
.await
{
error!("user shell command failed: {err:?}");
match exec_result {
Err(CancelErr::Cancelled) => {
let aborted_message = "command aborted by user".to_string();
let exec_output = ExecToolCallOutput {
exit_code: -1,
stdout: StreamOutput::new(String::new()),
stderr: StreamOutput::new(aborted_message.clone()),
aggregated_output: StreamOutput::new(aborted_message.clone()),
duration: Duration::ZERO,
timed_out: false,
};
let output_items = [user_shell_command_record_item(&raw_command, &exec_output)];
session
.record_conversation_items(turn_context.as_ref(), &output_items)
.await;
session
.send_event(
turn_context.as_ref(),
EventMsg::ExecCommandEnd(ExecCommandEndEvent {
call_id,
stdout: String::new(),
stderr: aborted_message.clone(),
aggregated_output: aborted_message.clone(),
exit_code: -1,
duration: Duration::ZERO,
formatted_output: aborted_message,
}),
)
.await;
}
Ok(Ok(output)) => {
session
.send_event(
turn_context.as_ref(),
EventMsg::ExecCommandEnd(ExecCommandEndEvent {
call_id: call_id.clone(),
stdout: output.stdout.text.clone(),
stderr: output.stderr.text.clone(),
aggregated_output: output.aggregated_output.text.clone(),
exit_code: output.exit_code,
duration: output.duration,
formatted_output: format_exec_output_str(&output),
}),
)
.await;

let output_items = [user_shell_command_record_item(&raw_command, &output)];
session
.record_conversation_items(turn_context.as_ref(), &output_items)
.await;
}
Ok(Err(err)) => {
error!("user shell command failed: {err:?}");
let message = format!("execution error: {err:?}");
let exec_output = ExecToolCallOutput {
exit_code: -1,
stdout: StreamOutput::new(String::new()),
stderr: StreamOutput::new(message.clone()),
aggregated_output: StreamOutput::new(message.clone()),
duration: Duration::ZERO,
timed_out: false,
};
session
.send_event(
turn_context.as_ref(),
EventMsg::ExecCommandEnd(ExecCommandEndEvent {
call_id,
stdout: exec_output.stdout.text.clone(),
stderr: exec_output.stderr.text.clone(),
aggregated_output: exec_output.aggregated_output.text.clone(),
exit_code: exec_output.exit_code,
duration: exec_output.duration,
formatted_output: format_exec_output_str(&exec_output),
}),
)
.await;
let output_items = [user_shell_command_record_item(&raw_command, &exec_output)];
session
.record_conversation_items(turn_context.as_ref(), &output_items)
.await;
}
}
None
}
Expand Down
108 changes: 108 additions & 0 deletions codex-rs/core/src/user_shell_command.rs
Original file line number Diff line number Diff line change
@@ -0,0 +1,108 @@
use std::time::Duration;

use codex_protocol::models::ContentItem;
use codex_protocol::models::ResponseItem;

use crate::exec::ExecToolCallOutput;
use crate::tools::format_exec_output_str;

pub const USER_SHELL_COMMAND_OPEN: &str = "<user_shell_command>";
pub const USER_SHELL_COMMAND_CLOSE: &str = "</user_shell_command>";

pub fn is_user_shell_command_text(text: &str) -> bool {
let trimmed = text.trim_start();
let lowered = trimmed.to_ascii_lowercase();
lowered.starts_with(USER_SHELL_COMMAND_OPEN)
}

fn format_duration_line(duration: Duration) -> String {
let duration_seconds = duration.as_secs_f64();
format!("Duration: {duration_seconds:.4} seconds")
}

fn format_user_shell_command_body(command: &str, exec_output: &ExecToolCallOutput) -> String {
let mut sections = Vec::new();
sections.push("<command>".to_string());
sections.push(command.to_string());
sections.push("</command>".to_string());
sections.push("<result>".to_string());
sections.push(format!("Exit code: {}", exec_output.exit_code));
sections.push(format_duration_line(exec_output.duration));
sections.push("Output:".to_string());
sections.push(format_exec_output_str(exec_output));
sections.push("</result>".to_string());
sections.join("\n")
}

pub fn format_user_shell_command_record(command: &str, exec_output: &ExecToolCallOutput) -> String {
let body = format_user_shell_command_body(command, exec_output);
format!("{USER_SHELL_COMMAND_OPEN}\n{body}\n{USER_SHELL_COMMAND_CLOSE}")
}

pub fn user_shell_command_record_item(
command: &str,
exec_output: &ExecToolCallOutput,
) -> ResponseItem {
ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: format_user_shell_command_record(command, exec_output),
}],
}
}

#[cfg(test)]
mod tests {
use super::*;
use crate::exec::StreamOutput;
use pretty_assertions::assert_eq;

#[test]
fn detects_user_shell_command_text_variants() {
assert!(is_user_shell_command_text(
"<user_shell_command>\necho hi\n</user_shell_command>"
));
assert!(!is_user_shell_command_text("echo hi"));
}

#[test]
fn formats_basic_record() {
let exec_output = ExecToolCallOutput {
exit_code: 0,
stdout: StreamOutput::new("hi".to_string()),
stderr: StreamOutput::new(String::new()),
aggregated_output: StreamOutput::new("hi".to_string()),
duration: Duration::from_secs(1),
timed_out: false,
};
let item = user_shell_command_record_item("echo hi", &exec_output);
let ResponseItem::Message { content, .. } = item else {
panic!("expected message");
};
let [ContentItem::InputText { text }] = content.as_slice() else {
panic!("expected input text");
};
assert_eq!(
text,
"<user_shell_command>\n<command>\necho hi\n</command>\n<result>\nExit code: 0\nDuration: 1.0000 seconds\nOutput:\nhi\n</result>\n</user_shell_command>"
);
}

#[test]
fn uses_aggregated_output_over_streams() {
let exec_output = ExecToolCallOutput {
exit_code: 42,
stdout: StreamOutput::new("stdout-only".to_string()),
stderr: StreamOutput::new("stderr-only".to_string()),
aggregated_output: StreamOutput::new("combined output wins".to_string()),
duration: Duration::from_millis(120),
timed_out: false,
};
let record = format_user_shell_command_record("false", &exec_output);
assert_eq!(
record,
"<user_shell_command>\n<command>\nfalse\n</command>\n<result>\nExit code: 42\nDuration: 0.1200 seconds\nOutput:\ncombined output wins\n</result>\n</user_shell_command>"
);
}
}
12 changes: 12 additions & 0 deletions codex-rs/core/tests/common/responses.rs
Original file line number Diff line number Diff line change
Expand Up @@ -61,6 +61,18 @@ impl ResponsesRequest {
self.0.body_json().unwrap()
}

/// Returns all `input_text` spans from `message` inputs for the provided role.
pub fn message_input_texts(&self, role: &str) -> Vec<String> {
self.inputs_of_type("message")
.into_iter()
.filter(|item| item.get("role").and_then(Value::as_str) == Some(role))
.filter_map(|item| item.get("content").and_then(Value::as_array).cloned())
.flatten()
.filter(|span| span.get("type").and_then(Value::as_str) == Some("input_text"))
.filter_map(|span| span.get("text").and_then(Value::as_str).map(str::to_owned))
.collect()
}

pub fn input(&self) -> Vec<Value> {
self.0.body_json::<Value>().unwrap()["input"]
.as_array()
Expand Down
Loading
Loading